October 25, 2026
Update
You now edit your WordPress posts directly in the editor
You load existing posts and pages from your WordPress site into the editor, revise them there and write them back into the same post. Address, rankings and backlinks are kept, and before every overwrite you see what changes.
You load existing posts from WordPress
Content Studio › Editor › Sources › Load Page Content › From WordPress
- The “From WordPress” tab shows the posts and pages of your connected site, most recently modified first. You can search and load a post as a new draft with one click.
- Images, galleries and other elements the editor cannot display appear there as a line of their own and stay unchanged when you write back.
- Posts from page builders such as Elementor or Divi can be loaded and checked but not written back, because their content is not stored in the WordPress text field.
You update the existing post instead of creating a second one
Content Studio › Editor › Publish article
- When publishing, you choose whether Rankmio updates the existing post or creates a new one. Updating changes only title and text; address, status, categories, featured image and SEO settings stay.
- Before overwriting, Rankmio shows you every changed paragraph in its old and new version. Paragraphs you have not touched go back unchanged.
- If the post was changed in WordPress after you loaded it, Rankmio warns you and only overwrites after your confirmation. “Undo” brings back the previous state, and WordPress also saves a revision.
October 24, 2026
Update
Your AI agent works along, and your figures say the same thing
AI agents such as Claude Code now connect directly to Rankmio. Your citation rate is the same everywhere, new search queries no longer pull your values down, and the task list only reports what a measurement proves.
Your AI agent works directly with Rankmio
Profile › API keys, connection via https://rankmio.de/mcp
- Agents such as Claude Code or Cursor connect via MCP with one command and ask Rankmio themselves: open tasks, click losses, AI citations, snippet analysis, backlinks. They choose the period themselves and can write their own report from the data.
- You decide per key what the agent may do. With "act" it ticks off tasks, saves keywords, fetches new Search Console data and reports changed pages to Bing; with "spend credits" it also starts audits and drafts.
- Before anything that costs credits, the agent names the price and only books after your confirmation. Prices and rights are the same as in the web interface, and each key spends at most 200 credits a day.
You see one citation rate, the same everywhere
Visibility, Citation Tracker, Dashboard, Chat and agent
- All places count the same answers from the last 30 days, so index, tracker and dashboard no longer show different rates.
- Questions with your brand count fully. A sentence below the figure splits them into brand and topic questions and rates both, for example: "For questions with your brand you are cited in 76 % of answers. That is good."
Questions from Bing Copilot cost nothing in question discovery
GEO › Citation Tracker › Question discovery; GEO › Bing AI
- If you have uploaded your Bing AI report, Rankmio offers the questions from it free of charge for your watchlist. The Bing page takes you there with one click.
Your GEO score only uses current reports
GEO › Audit results; GEO › Google AI report and Bing AI
- Uploaded AI reports from Google and Bing count for up to 60 days after their period ends. After that, Rankmio reminds you to upload a new one instead of rating with old data.
- Rankmio rejects a report that belongs to another domain when you upload it, so it cannot distort your score.
New search queries no longer pull your values down
Visibility index, Recover clicks, Tasks
- If your page is newly shown for many queries far back, that is growth. The search pillar of the index therefore only measures exhaustion up to position 20.
- Click losses and ranking losses only compare queries that occur in both periods. A page no longer counts as a loser just because it is shown for more queries.
Click losses name the query that costs you clicks
Keyword Hub › Recover clicks
- Each page shows the query with the largest loss, says whether it is a brand query, and counts new queries as growth. You see whether you lost ranking, demand or click rate.
Cannibalisation is only reported for real conflicts
Keyword Hub › Opportunities and Cannibalisation; Tasks
- Opportunities only carry the warning if two of your pages really compete. Before, one single impression on a second page was enough.
- Several pages for your own brand name are normal and appear as a note. If none of them is in position 1, Rankmio reports it as a finding.
Ticked-off tasks come back if the problem remains
Tasks; also for AI agents
- If the next crawl, GEO audit or Search Console sync still finds the problem, the task reopens with the note "Still present despite being ticked off". Nothing disappears just because it was ticked off.
Tasks say more precisely what to do
Tasks, AI action plan
- Length tasks name the measured length and the limit, and the orphan-page task names the pages. Redirects no longer count as missing from the sitemap.
- The AI action plan no longer creates "secure your success" tasks for pages that rose, and title tasks only where the snippet analysis sees a case. Ranking losses beyond position 20 are no longer critical.
- Questions to language models no longer justify link recommendations, and a typo in a search query no longer counts as a missing term.
GEO › Audit results; Content Studio › Editor › Citation readiness
- Paragraphs over 200 words and an AI's verdict on your opening appear as notes without deducting points. Bing advises against walls of text but names no number, so Rankmio says openly that it sets the limit.
- The editor lists these points under "Content hints, no points and not a defect".
You also measure snippet changes you wrote yourself
Keyword Hub › Snippets & click rate
- With "Add your own change" you enter the old and new title and description. Rankmio records the state before and measures again after two weeks, free of charge, just as with an applied proposal.
Referring domains have one number everywhere
Backlinks, Dashboard, Visibility index, JSON export
- Rankmio counts the domains from the link database and from your uploaded Google and Bing link reports, each once and without spam. The same number appears in the index, the backlink hub and the export.
October 23, 2026
Update
What has improved since 14 October
The AI recommendations in the Citation Tracker now show you more precisely where you can gain ground with AI systems and take you straight to the matching brief. In the Content Studio, your texts become more reliable and working in the editor gets clearer.
Rankmio now only recommends where you have something to gain
GEO › Citation Tracker › AI recommendations
- You only get recommendations for questions where AI systems have not cited your website once in the last 30 days. Questions where you already appear no longer clutter the list.
You know in advance what your text needs to cover
GEO › Citation Tracker › AI recommendations
- For every recommendation, Rankmio names three to five subtopics and shows in which AI answers they appear today. The recommended length fits the topic instead of a flat 1,500 words.
Rankmio suggests the right page to expand
GEO › Citation Tracker › AI recommendations
- If an existing page should be expanded, Rankmio reads your pages first. It suggests the page whose content fits the question best.
A recommendation turns into a ready-made brief
GEO › Citation Tracker › AI recommendations, continued in Content Studio › Briefs
- One click turns the recommendation into a prepared brief. It takes the question as its topic and brings along today's AI answers and the competitors cited there as material. You no longer have to gather anything yourself.
You can see what Rankmio is working on
GEO › Citation Tracker › AI recommendations
- While Rankmio creates recommendations, a display with a seconds counter shows how long it has been running. If something goes wrong, a plain sentence tells you what happened.
Your texts address readers the way you defined
Content Studio › Personas, applied when generating in the editor
- If you set the informal or formal form of address in the persona, the whole text sticks to it, from the first to the last paragraph. Without a persona, Rankmio writes in the formal form.
Your texts no longer cite sources that are not sources
Content Studio › Editor
- A text about your own page no longer lists that page as a source. The references at the end only show genuine external evidence.
Research finds more relevant sources
Content Studio › Briefs and Editor
- When generating, Rankmio now searches for the topic of your text. For some text types it used to search for the format, such as "step-by-step guide", and then found generic guides instead of material on your topic.
You find your briefs again faster
Content Studio › Briefs
- You can sort the list by "Newest first" or alphabetically. Briefs that have not produced a text yet are marked "Not created yet".
You see at a glance what your text still lacks for citation readiness
Content Studio › Editor › Citation readiness
- Instead of a ring with an average score, the editor now shows how many requirements your text meets, for example "5 of 7". You see right away how many are still open.
Existing pages arrive in the editor complete
Content Studio › Editor › Import URL
- When you load an existing page into the editor, accordions and line breaks are kept. Previously they were lost and words ran together.
From a task you start in the Content Studio with all the facts
Tasks, continued in the Content Studio
- When a task takes you to the Content Studio, it brings the measured values and evidence along. You see there why the task was created and do not have to jump back.
October 21, 2026
Security
PHP2Shell: Security vulnerability closed the same evening
Security vulnerabilities cannot always be prevented. What matters is how quickly and consistently you respond. When the warning about the critical PHP vulnerability PHP2Shell (CVE-2026-109618 and CVE-2026-109619) became known on 21 October, we immediately activated protective measures at Rankmio and installed the available security update.
From the first protection to the closed security hole
- 15:49: First protective measures on our server: The upload progress display was switched off, the nesting depth of form fields was limited, and PHP processes are renewed after every 100 requests.
- 15:52: An additional limit of 8 MB for file uploads.
- Until 20:18: We checked continuously whether the corrected PHP package was available for our server. The package server was in high demand during this time and at times unreachable, so we kept restarting the query.
- 20:18: The corrected PHP package was available for our server.
- 20:29: The update to PHP 8.4.27 was installed. The temporary upload limit could be lifted again.
- The result: Only eleven minutes after the repaired package became available, the corrected PHP version was running on our server. The vulnerability was closed the same evening, about four and a half hours after the PHP developers published the fix.
Security is more than an update
- Rankmio is also reachable exclusively through Cloudflare. This upstream service adds a further layer of protection that can filter known attack patterns.
- Our subsequent check found no traces of an attack related to this vulnerability.
- Our standard: Do not put security warnings on the back burner, but limit risks immediately and install available fixes promptly. Trust does not come from security promises, but from actions that can be traced.
October 14, 2026
Content Studio
The Content Studio now also works without a website of your own
Alongside the website project there is now the content project. It needs no domain and bundles everything you need for texts on one topic. It stands on its own or belongs to one of your website projects and then uses its data.
- Under New Project you now choose between a website project and a content project. A content project needs a name, a language and a country for keyword research; a description and the industry are optional and feed into personas and texts.
- Content projects do not count towards the limit of website projects. Each account can create up to 100.
- You assign a content project to a website project and detach it again in the project overview. A content project never turns into a website project. If you delete a website project, its content projects remain with all drafts.
- Under a website project, the editor uses that website's data: The Search Console, the crawled pages for internal links and the shared competitor register. If no pages have been crawled yet, the editor says so and links to the sitemap crawl.
- In a content project, Rankmio shows only the areas that work without a website: The Content Studio, keyword research and the competitor register. Analysis areas are hidden instead of greyed out.
- The chat assistant advises you on your texts in a content project and does not discuss rankings that cannot be measured there.
- Content projects are shared individually, like website projects. When you create one under a website project, it is shared by default with everyone who can see the website project.
- In the sharing overview, the button "+ content projects" gives everyone who sees a website project its content projects as well.
- Projects that a team member creates now belong to the team lead and are shared with the member. This applies to both kinds of projects; existing projects stay where they are.
October 13, 2026
Update
Re-checking asks Google directly, and the GEO hints target the right paragraph
- The task "Google does not know the current state of a page" can now be re-checked. Rankmio asks Search Console for exactly this page and confirms the task as soon as Google reports the page as indexed.
- Addresses you have submitted to Bing no longer count as unknown until Bing has checked them again. The task disappears and is recorded as resolved.
- Pages submitted to Bing show "Submitted on" with the date instead of the submit button.
- The sitemap crawl now reads every page it finds. Before, it sometimes stopped early and reported pages as vanished that still existed.
- The check "answer first" now judges the first paragraph directly below the H1, as readers and AI systems see it, even when it sits in the page header.
- Overview pages such as glossaries or article lists and FAQ pages no longer receive the hints "answer first" and "list missing". Grids of cards now count as lists.
- An anchor text is only suggested if it describes the target page itself. Search queries with typing errors no longer appear as missing terms.
October 12, 2026
Content Studio
The editor checks citability like the GEO audit and suggests internal links
- The editor checks citability while you write, by the same rules as the GEO audit and free of charge. The same page therefore gets the same score everywhere. Only the assessment of named entities uses AI, for one credit.
- If your CMS outputs the title as H1, you tick a box once per project. Rankmio then checks that the text itself contains no H1.
- A new box in the editor shows which links the page should set and which pages fit its topic. One click places the link at a suitable spot in the text.
- The right-hand panel follows the order of work: the open task first, then the open points of the citability check. Everything already fulfilled can be collapsed.
- The editor now loads the text of a page from every entry point, including page comparison, click-through rate and Search Console. Opening a task a second time continues in the existing draft instead of creating a new one.
- Your schema builds on the page's existing markup instead of replacing it, so publisher, logo and topics are kept. A comparison shows the original and your schema side by side, field by field.
- If a WordPress site already marks up its posts with a plugin such as Yoast, Rankmio only transfers the additions, such as an FAQ. After publishing, it checks whether the schema has arrived on the page.
- Before enriching keywords, Rankmio shows the cost and only charges for terms it has not fetched before. The full check after generating the schema is included in the price.
October 12, 2026
SEO & GEO
Link recommendations become tasks, and keyword selection follows the project's topic
- Link recommendations now target pages with a ranking opportunity and pages that are hardly linked in the content. Each recommendation names the reason and becomes a task in the task list. One click opens the source page in the Content Studio, and re-checking confirms the link once it is set.
- The AI assistant knows the open link recommendations and answers questions about them for each page.
- The search phrases for AI visibility are now created in the Citation Tracker's Discovery. The first run from Search Console data is free for each project, and existing prompts can be imported.
- "Select the best 20" picks keywords that match the topic words of the project and skips very difficult terms and competitor brands. You can see and change the topic words.
- Gap analyses are saved and can be reopened free of charge. Competitors can be chosen from the register.
- Ticks, discarded and own tasks apply to the whole team for each project. Team leads can allow members their own branding.
- Each branded PDF is charged only once per content within 24 hours, so reloading is free. Before the click, the button explains why a PDF is not possible yet.
October 10, 2026
Version 1.5
Rankmio 1.5 brings the key improvements since version 1.0
Version 1.0 was released on 15 June 2026. Since then, Rankmio has advanced most clearly in these areas. On 10 October the AI assistant and further improvements were added.
- The assistant now knows all measurements of a project at once, from rankings and indexing to AI citations and backlinks. Its answers are more accurate as a result, and where data is missing, it says so openly.
- It recognises seasonal keywords from the searches of the last twelve months, so content can be revised in good time before the strongest month.
- The score is half verifiable requirements for AI answers and half measured appearances in AI answers. It shows what a website can improve itself and what is already working.
- The Citation Tracker measures citations in six AI services, including ChatGPT, Perplexity, Google's AI Overview and Bing Copilot. Google's AI report and Bing's Copilot report can be imported as well.
- All observed questions of a project are kept in one watchlist and can be filtered, edited and removed together. How often they are measured is up to you, from weekly to every four weeks, which keeps costs predictable.
- The NotebookLM mode works with your own sources, and the Content Coach backs statements with verified references. Existing pages can be loaded and saved with complete markup.
- The coach in the editor now reads long articles in full. It finds contradictions and repetitions across the whole text, not just at the beginning.
- One click in the "Action required" box shows whether the task has already been solved in the draft.
- Identical issues are merged, similar tasks are bundled and ordered by four fixed levels. Resolved tasks can be re-checked free of charge.
- Outdated blocks or errors that Google still reports appear as a separate task. Archive pages are combined into one task so the list stays clear.
Snippets and click-through rate
- Rankmio shows for which search queries a better snippet brings more clicks, and a free tool checks a page's title and description.
- Competitors can be added manually. The competitor comparison also rates how well each page matches the search term and therefore shows more fairly who is really ahead.
- The gap analysis takes over the pinned competitors automatically. It warns before the same analysis runs a second time unnecessarily.
- Traffic Recovery shows for every page why clicks were lost: lower demand, weaker rankings or a lower click-through rate. This makes clear where work on the page pays off.
- After connecting, the first data arrives right away. Bing search performance has its own tab in the Keyword Hub.
October 5, 2026
GEO
The GEO score has been redesigned, foundation and impact determine the rating
- The score is half foundation and half impact. The foundation counts the requirements for AI answers a page meets, the impact its measured appearances in AI answers. New websites can see their progress this way, too.
- Every criterion of the foundation can be verified on the page, such as crawler access, readable text and a clear structure.
- The impact is based exclusively on real measurements from the Citation Tracker, Google's AI report and Copilot.
- Every open requirement appears as a separate task that can be fixed and re-checked. If a website meets the requirements but is rarely named in AI answers, the bottleneck lies outside the page. Rankmio then points to missing links from other websites and to sources cited in AI answers where the project does not appear.
- Core statement: a hint reports when the first paragraph does not deliver the core statement of the heading.
- Search query and page content: a hint reports search queries whose terms are missing from the page.
- The free check shows the GEO foundation of a page with all requirements, for example whether AI crawlers may fetch it and whether the text is clearly structured. Without signing up, it shows what stands in the way of a citation.
- The foundation is measured more accurately for your own page and every competitor page.
October 3, 2026
Page roles
Page types are recognised and rated more intelligently
- Legal, function, tool and directory pages can be marked. Unsuitable tasks no longer apply to them, and they are excluded from averages.
- Rises and upcoming seasonal demand are shown more clearly, and the texts are easier to understand.
- For every page, the audit shows how often it appears in Google's AI report, in Copilot and in the Citation Tracker.
- It tracks the rankings of competitors and reports gained and lost search terms even more precisely.
- The free check rates HTTPS positively only with a valid certificate.
October 2, 2026
Citation-Tracker
Citation Tracker and competitor comparison measure more reliably
- Queries that could not be carried out are shown as "not measured" and are not charged.
- All domains are measured against the same AI answer.
- Tasks follow the most recent comparison. After changes to the page, older recommendations expire.
October 1, 2026
Content Studio
The Content Studio has been extended
- External sources can be added as a file, address or text, up to 20 per project free of charge.
- From web pages, only the actual article text is filtered and imported.
- It only names sources from the library or from a verified research run. Every address is checked for availability before it is shown.
- The "Search sources on the web" function finds up to five verified references for a draft.
September 30, 2026
Action list
The action list has been extended, resolved tasks can be re-checked
- After ticking off a task, the page can be re-checked free of charge.
- The re-check can be repeated.
- Tasks can be filtered by topic and finding.
- The nightly sync runs more reliably and reports missing permissions if the Search Console sign-in was not fully completed.
September 29, 2026
Task list
Tasks have a consistent urgency
- All findings are assigned to four fixed levels: critical, high, medium and the new level hint. A hint marks cases in which more data is needed before a recommendation.
- Better meta descriptions are only recommended for search queries on page one of Google's results. Only there is a snippet noticed; on page two, only a better position brings more clicks. More in the article Optimising snippets and in the glossary under Snippet.
- Instead of up to three tasks on topic coverage, AI readiness and AI visibility, there is one task per page.
- Requests rejected by the server count as a missing measurement and not as an error of the page.
September 28, 2026
Task list
Identical findings are merged
- Duplicate check: the same issue on the same page appears only once, even if several analyses report it.
- Bundling: many similar findings are combined into one expandable collective task.
- Key figures without a measurement are marked as "not measured".
September 27, 2026
Dashboard
The dashboard loads much faster
Changes from 26 and 27 September 2026.
- Tasks on click-through rates and keyword conflicts now take the position and a minimum number of impressions into account. Hints on search terms that can hardly bring clicks anyway no longer appear.
- Legal texts, the manual and text files no longer receive unsuitable tasks, for example about missing images or short titles.
- The loading time of large projects dropped from around 19 seconds to under one second.
- The distribution of search terms across positions 1 to 10, 11 to 20 and beyond replaces the average position. An average hides whether a page has many terms in the top ten or only ranks moderately.
- The list can be exported with the current filter as a table or a list of addresses.
- Recommendations are checked against the measured values of your own page.
- The audit panel always shows all relevant tasks, even if there is none for the selected page itself.
- The PDF export contains exactly the sections opened in the panel.
- It explains falling impressions by position ranges.
September 25, 2026
Action list
Task bundling can be configured
- Bundling of similar tasks can be switched on or off.
September 24, 2026
Sitemap analysis
Duplicate content is detected more accurately
- Redirects, noindex pages and pages with a canonical reference no longer count as duplicate content. False tasks no longer appear.
- Tasks whose data basis no longer exists appear as "expired" instead of done.
- Search queries for your own brand name no longer create cannibalisation tasks.
- Data is kept for 90 instead of 30 days. Every task names the date of its data basis.
- New titles and descriptions are only suggested when they are actually needed.
- Expired or invalid SSL certificates are reported, with an advance warning 14 days before expiry.
September 23, 2026
GEO-Audit
The GEO audit detects FAQ sections more reliably
- Question-and-answer sections are also recognised when the questions are set as subheadings. Unjustified FAQ tasks no longer appear.
- Only the questions from the "Frequently asked questions" section are included in the FAQ markup.
September 23, 2026
Content Studio
Load an existing page and build the markup it lacks along the way
Plenty of pages have been online for years, rank reasonably well and still carry no markup: no Article, no FAQPage, even though ten questions with answers sit at the bottom. Rewriting them for that reason would be unnecessary. You can load the page into the editor, adjust it there and have the markup built when you save. It costs no credits, because no language model runs — what stands in the text is read.
Saving now marks up the whole piece
- Until now the editor built only the quiz markup when you saved. An exported piece therefore carried a Quiz and nothing else, while a dozen questions with answers stood in the text and the article frame was missing altogether.
- Saving now adds every type the text supports: the article frame with headline, description, author, publisher, preview image, both dates, language and word count; the FAQ from the questions under the heading of that name; and the quiz if a quiz block sits in the text.
- What you have reworked by hand stays untouched. Only a type the draft does not yet carry gets added, so a second save changes nothing. If you would rather build it again after editing the text, the new button "Alles auszeichnen" in the schema card does exactly that.
- Nothing is invented. If a detail is missing, the field stays out — no placeholder image, no made-up author, no estimated date. A gap in the markup is more honest than a value nobody checked.
The export now carries real markup
- When a page is loaded, the studio reads the markup it finds — but files it as a finding rather than copying it into your draft. The reason: the moment you change the text, copied markup keeps asserting the previous state, and markup that says something other than the page is worse than none at all.
- The HTML export read the wrong column and wrote that finding into the head section, raw and without the wrapper it needs. An exported piece carried ten kilobytes of diagnosis between the meta tags. It now emits only markup that belongs to your draft, and both output routes run the same check beforehand.
- Two further details in the same export: the description took the first paragraph, which on an imported piece is the category line, and the author was hard-wired to "Rankmio Redaktion" although the draft knows better.
Read more: what structured data is and why we build it fresh
- A new article sets out what structured data actually is, why Schema.org and JSON-LD are not the same thing — the first is the vocabulary, the second one of the formats in which it is embedded — and what happens when a page is fetched into the editor: Add Schema.org and JSON-LD to existing pages.
- It also covers two points that are easily got wrong. Structured data is not an automatic ranking factor and guarantees no better position. And FAQPage markup leads to an enhanced Google presentation only for certain particularly authoritative sites today — it is still worth doing, but for different reasons.
September 22, 2026
Backlinks
Rankmio detects patterns in your link profile and creates a disavow file on request
A single spam score says little about whether a link is harming your website. It only becomes meaningful when several conspicuous traits occur together: the same link text on several unrelated domains, the same URL pattern on the linking pages, the same week in which they all first appeared. That is exactly what Rankmio now looks for in the backlink hub, and under Technology there is a new tool with which you can build a disavow file for Google Search Console from the findings. Rankmio does not classify any link as harmful; it describes what was observed. On top of that come two new ways to add backlinks from Bing and Search Console, a stricter assessment of your link profile, and a detailed article about what the spam score of backlink tools actually measures.
The backlink list now shows findings
- Above the backlink list under Analyse you will find a new panel listing the patterns Rankmio observed in the loaded links. Five kinds are recognised: the same anchor text across several unrelated domains, the same URL pattern on the linking pages, a cluster of new referring domains within a short period, links from your own other projects, and links with a conspicuous spam score that carry no follow attribute and therefore pass on no weight.
- Each finding states its evidence, how many links and domains belong to it, and how many of those actually pass on ranking weight. One click filters the list below to exactly that group, and every row in the list shows which findings it belongs to.
- Rankmio deliberately passes no judgement here. A pattern is an indication and not proof, because franchise systems, partner programmes and central copy templates produce the very same repetitions. The panel says so, and the decision remains yours.
- A real example from a customer project: nine domains linked to the same site using an identical anchor text ninety-two characters long, all nine links sat on pages with the same URL pattern, and all nine were first seen within three weeks. The domain ratings of those nine pages ranged from 1 to 216, so anyone filtering by that widely used metric would have missed about half the network.
New tool under Technology: link profile check and disavow file
- A disavow file tells Google to disregard certain incoming links when assessing your website. The new tool creates such a file on request. It opens with the findings, not with a download button, and the button only becomes active once you have confirmed three checks: whether Search Console shows a manual action, whether a disavow file already exists for the property, and whether you have compared the candidate domains against your own traffic data.
- For each finding you decide separately whether it enters the file at all, and if so whether as whole domains or only as the individual URLs. Choosing a whole domain shows the reminder that this also covers editorially placed links, which matters for press distributors, directories and platforms with many independent users. Links from your own other projects cannot be selected, because you would be declaring your own references invalid.
- The generated file follows the rules of Google Help exactly: one entry per line, whole domains with the correct prefix and without protocol or www, no subdirectories, and a comment line with reason and date before every group. Entries Google would reject are named rather than quietly bent into shape.
- Right at the top the tool quotes Google itself, which calls the feature advanced and warns that using it incorrectly can harm your performance in Google Search. We offer the function to make the work easier; we classify no link as toxic, and its use happens at your own responsibility.
Backlinks from Bing and Search Console now complete the picture
- The Bing report on referring pages describes one individual link per row, including the source URL and the anchor text. Rankmio had been adding those rows up and discarding the detail. The rows are now kept and flow into the pattern detection, so findings can also arise from links our data supplier does not know at all.
- For the links report of Google Search Console there is no interface, only the export. Rankmio now accepts it in the backlink hub, either as the ZIP archive containing every table or as a single CSV. It reads the linking sites, the link texts and your most linked pages, and shows Search Console as a third source alongside DataForSEO and Bing in the source comparison.
- What Search Console does not supply is stated on the upload card: no follow status, no spam score, no first-seen date. It widens the coverage and does not replace the regular fetch.
- Domains known only from these uploads appear in the disavow tool under their own heading, separated by source, so you can go through them one by one. Upload a list once in the hub and you will find it in the tool.
The assessment of your link profile has become stricter
- The health score of a link profile is capped when a large share of the links is conspicuous. Those caps now take effect earlier, from twenty instead of twenty-five percent of weighted conspicuous links. Where a cap applies, it is named alongside the score, so a capped value cannot be mistaken for a genuinely poor one.
- All existing snapshots were recalculated so that the displayed grades match the current assessment. Four of eleven changed, two of them considerably: profiles containing a machine-generated link network dropped from an A+ to a B.
- Links carrying a nofollow attribute now count only a quarter as much as before, because under the usual treatment they pass on no ranking weight. The previous calculation punished profiles that simply collected many harmless links from URL shorteners and analysis tools.
New article: what the spam score really measures
- Under Insights there is a new article on the disavow file. It explains when Google recommends the file, what happens if you use it wrongly, how the file has to be built, and why Bing removed its equivalent in October 2023.
- At its centre stands the calculation behind the spam score of our data supplier, which is publicly documented. Eighteen technical characteristics of the linking page are examined, among them the encryption, the meta markup and the length of the domain name. A dated but harmless club website reaches seventy points, while a professionally built page from a link-selling network stays low. The article lists all eighteen signals with their point values.
- The article also names a detail many overlook: the disavow tool does not support domain properties, so the file has to be uploaded for a URL-prefix property.
September 14, 2026
Backlinks
The backlink history now shows DataForSEO and Bing together
Until now the history in the backlink hub showed only what DataForSEO had measured. If you had also uploaded a report from Bing Webmaster Tools, its figures were stored in Rankmio but never appeared in the chart. That has changed. Both sources are now shown side by side, and Rankmio additionally combines them into one figure in which each domain is counted only once. On top of that there is a further Bing report you can upload, and two corrections to displays that had been misrepresenting pages.
The history now shows both sources
- In the History tab you will now see three lines instead of one. The first shows how many referring domains DataForSEO knows, the second shows the same figure from Bing Webmaster Tools, and the third shows both sources combined, counting a domain that appears in both of them only once.
- Every line keeps its most recently measured value until new data arrives for that source. If you upload a Bing report without running a new DataForSEO request, only the Bing line moves while the other one continues horizontally. This way the chart no longer shows a drop that occurred simply because not both sources were measured on the same day.
- Our own domain shows why combining the two is worth the effort. For rankmio.de, DataForSEO knows 24 referring domains and Bing knows 19. Eleven domains appear in both, so after removing the duplicates 32 distinct domains remain. Eight of them are found by Bing alone, among them several genuine specialist sites that DataForSEO has not recorded.
- The figure at the top of the page now matches the table below it. Previously it could happen that three different values for the referring domains appeared in three places on the same page, because each place calculated the figure in a different way.
- The history reaches back as far as measurements are available. The combined line, however, only begins on the day a complete domain list was available from both sources for the first time. Older lists from DataForSEO are not retained, and without them there is no way to tell which domains overlapped back then. A note below the chart explains this where you see it.
The third Bing backlink report can now be uploaded
- Bing Webmaster Tools offers three different backlink reports for download. Rankmio was able to read two of them; the third one, the referring pages, was rejected with an error message. That has been fixed.
- This report is the most detailed of the three, because every row describes a single link: the page the link sits on, the text it was given, and the target page on your site. From one file Rankmio now derives all three views for which you previously had to upload two separate files, namely the referring domains, the most frequently linked target pages, and the link texts that were used.
- Rankmio also recognises on its own which character separates the columns in your file. Anyone who opens a CSV file in Excel and saves it again gets a semicolon instead of a comma in German-speaking countries. Files like that were previously rejected with the message that the CSV type had not been recognised, even though there was nothing wrong with their content.
Rankmio now distinguishes between a page Bing does not know and a question Bing did not answer
- In the Bing coverage view Rankmio checks for every address in your sitemap whether Bing holds it in its index. When Bing turns such a request down because too many were made at once, Rankmio previously had no precise information and recorded the page as unknown to Bing. That was a statement about the index which had never actually been measured. Cases like these now appear as a pending query, which is to say an open question. For rankmio.de this affected 129 of 148 addresses.
- The check now runs once an hour instead of once a night, and it takes on the project that has gone without an answer from Bing for the longest time. Previously it worked through the projects in the same order every time, so the first project used up the small number of permitted requests on its own and the remaining ones came away empty-handed.
- Within two days the number of confirmed indexed pages for rankmio.de rose from 9 to 78, and two further projects were checked in full.
The dashboard no longer reports sitemap gaps that are not gaps
- When Rankmio could not read your existing sitemap.xml during a crawl, the basis for the comparison was missing afterwards. Because not a single page had been marked as contained in the sitemap, every page that was found then counted as missing. The dashboard consequently listed tasks such as “Missing from the sitemap” for pages that had been in it all along. For rankmio.de there were 135 such pages, including the home page.
- Rankmio now checks in every one of these places whether the comparison actually took place. If it did not, no gap tasks are created at all, and the results page of the sitemap hub states explicitly that no comparison happened. In that case the chat agent no longer produces a list of gaps either.
September 4, 2026
Update
The visibility index is more precise, and the Content Studio can now tell a story
Two things changed on this day. SERP visibility no longer counts how many search terms a domain appears for; it weighs each term by its demand and by the click probability of your position. And in the Content Studio you get prose instead of an article when you set the content goal to Story.
- Each search term is weighed by its demand and by the click probability of your position, instead of simply being counted. Scores drop across all projects as a result — the same visibility is now expressed on a stricter scale.
- Bing counts as well: its visibility is added to the Google figure, and the dashboard states the Bing share underneath the SERP pillar.
- How the figure is put together, and why Rankmio does it differently from the established indices: Rankmio visibility index.
Narrative mode in the Content Studio
- With the content goal Story you get chapters, scenes and dialogue — without TL;DR, FAQ and CTA.
- Two new templates come with it: Short story and Narrative in chapters, built from seven narrative blocks — chapter, scene, dialogue, flashback, turning point, resolution, epilogue.
- If a file in your library names chapters, Rankmio recognises them and offers to take them over.
- The schema is marked up as ShortStory with the genre Belletristik, so a machine does not read the text as a factual report.
- What it looks like from the idea to the finished text, with every setting: The first story Rankmio told.
September 2, 2026
Content Studio
New tab in the Content Studio: author in the text and in the schema
Under Content Studio, Authors tab, you store per project who writes your texts — with role, author page and profiles such as LinkedIn or ORCID that make the person verifiable. Mark one entry as the default and it is preselected for every new brief, appearing both in the byline and in the structured data.
September 1, 2026
AI visibility
The citation tracker now shows where Google actually cites you — and how certain each finding is
Until now you gave the tracker phrases and learned whether your site appeared in them. Now it also works the other way round: Rankmio asks Google for the queries where your site is named as a source in the AI overview. And every monitored phrase now states how many checks its result is based on.
Where Google already cites you
- You see the queries in which Google names your site in its AI overview. Not the phrases you guessed — the ones where you are actually quoted, including those nobody thought of.
- Each entry shows its monthly search volume. That tells you which citations actually carry reach and which are marginal.
- One click adds a query to your watchlist. Phrases you already monitor are marked, so the list stays complete.
- The lookup costs no credits. It comes from the same data source Rankmio already queries for your rankings.
Every finding now states what it rests on
- Under each phrase you now read: visible in 9 of 20 checks. A single measurement is not a fact — for four out of five visible phrases, the result differs between checks.
- Below five checks the number is marked as too thin. Then a swing cannot be told apart from chance, and you should not act on it yet.
- The configuration shows how many checks your schedule produces. One weekday gives about four per month. If that is too few, the note suggests checking fewer engines and measuring more often instead.
The visibility page shows again which positions your clicks come from
- The distribution across position bands is back, and so is the daily click curve. You can see at a glance whether your traffic comes from the top places or from a wide spread far back.
- A fifth band names what cannot be attributed. Google reports neither query nor position for rarely used searches. Those clicks count, but they belong to no band — and now you can see how many they are.
August 28, 2026
Snippets
A free tool checks your snippet, and a new article explains what matters
The title and description decide whether someone clicks your result. From now on you can have both checked without an account: enter an address, look at the preview, take the suggestion with you. A new article explains when working on a snippet pays off and when it does not. The snippet module has its own menu entry now, and in the content studio Rankmio writes the matching markup for every quiz.
The free snippet check at /seo-tools/snippet-check
- Enter an address and see how your result looks in Google. The preview shows the title and description as Google displays them.
- The check measures whether your title is too long. Google cuts titles off at roughly 580 pixels, and pixels are what count here, not characters — a title full of capital letters runs out of space sooner than one in lower case. The length of the description is checked as well.
- You receive a suggestion for a better title. Before writing it, the tool looks at up to 18 other pages of the same website, so the suggestion fits the site rather than standing out from it.
- You need no account. One check per day is free.
The new article: Google rewrites snippets. Is optimising them still worth it?
- Google replaces many titles with its own wording. The article explains when that happens, when your own title survives, and which pages are worth the effort.
- We checked 36 websites for it. The article shows what came out of that — which mistakes appear again and again, and which of them actually cost clicks.
- At the end there are six common questions and a short quiz. In the quiz you can check whether you would judge a given snippet the same way.
Snippets and click rate now has its own menu entry
- You will find the module in the menu under "Snippets & click rate". Behind it, Rankmio lists the pages of your project where a better title wins clicks — and states for every one of them why.
- In the traffic recovery, the "snippet tuning" area now applies the same standard. Only pages whose figures support the finding appear there. Every row states why the title is the cause here, and a button opens the page straight in the snippet tool.
Quiz markup in the content studio
- If you generate a quiz, the matching Schema.org markup is written along with it. Google can recognise the quiz from it. You no longer have to add anything by hand.
- Article, FAQ and quiz now sit in one shared block. Previously each stood on its own in the page; connected, search engines understand more easily that they belong to the same article.
August 25, 2026
Snippets · Click rate
Changes to the snippet feature that lift your click rate
Your page ranks and gathers impressions, yet receives not a single click? Between a good position and an actual click sit two lines of text: The title and the description Google shows in its results. Rankmio now goes about finding these pages differently. The new feature first checks whether the snippet is the cause of the missing clicks at all, and backs every answer with your project's measured data. Only then does it write a proposal, one you can still edit before applying it.
Every query is examined: Is it the snippet, the ranking or the search intent?
- Every query receives exactly one of seven properties. They are: Snippet, ranking, cannibalisation, SERP layout, search intent, AI prompt or too little data. Six of them deliberately lead to no proposal, because a page on position 47 has a ranking question and not a title question. Saying so helps you more than a new headline does.
- Every deviation is tested for reliability. A query with 30 impressions and no clicks looks dramatic; whether it means anything depends on how likely that result would be at a normal click rate. That probability is calculated, and only queries that pass appear as findings. What is left is short — and every line in it carries weight.
- The expected click rate refers to the actual position. A page averaging 5.6 is measured against what a page on 5.6 achieves, interpolated between the neighbouring values. The curve now runs to position 20, so pages further back are assessed on their own terms.
Three areas, sorted by the moment they take effect
- The first area gathers the pages where a better snippet wins clicks right away. They rank well in the results, their click rate still falls reliably short of expectation, and no other cause explains it. They win clicks on the day you change the snippet.
- The second area gathers what works once the page climbs. For pages further back, two figures now stand side by side: What the snippet brings today, and what it would bring at position 5. A rejection thereby turns into a task with a different timing, because anyone planning to move that page up anyway is better off writing the title along the way.
- The third area shows pages that lack a description of their own. Where none is stored, Google assembles a snippet from the page text and usually takes the first passage that matches the query. This is worth doing regardless of position, and up to ten pages can be handled in one pass.
Brands and proper names now form a group of their own
- Searches for your own name are assessed separately. People searching for a company by name click almost regardless of the wording. Those queries now form their own group, so the assessment of everything else rests on searches where the snippet actually decides.
- You can store several brand terms per project and decide about them yourself. The domain stem is derived automatically. Beyond that Rankmio makes suggestions: Parts of a compound domain name, and terms whose click behaviour looks like a brand. A suggestion becomes a brand only once you confirm it — company name, product name, spellings and typos included.
- Questions addressed to an AI likewise stand apart. Since AI overviews pass their prompts through, whole sentences appear in Search Console data — they have impressions and a position, but nobody typed them into a search box. They now form their own group instead of appearing as keywords to optimise.
What goes into a proposal
- The proposal grows out of the words people actually search with. The real queries of that page go into the brief, with their volumes. So does the search intent: Someone comparing offers needs a different snippet from someone wanting to understand something.
- Length follows the device mix. Google truncates earlier on a phone. A project with 82 percent mobile impressions gets a title written for 50 characters, one with a desktop majority up to 60 — and the character counter shows live where the limit runs.
- Your own winners serve as the model. Where pages in the project click well above expectation — reliably so, and outside brand queries — their titles go into the brief as examples. Patterns that work with this audience beat any general rule.
- Whatever is distinctive in the existing text stays. A concrete number, an admission, a detail no competitor could copy — that survives into the new version. And if the existing text is already good, the proposal says so and presents itself as an alternative rather than a recommendation.
- You can change both before you apply them. Title and description sit in input fields. What you apply is your wording, and it is stored separately from the proposal — so the follow-up measurement compares against the text that actually stands in the results.
You can choose any page and see what happened afterwards
- You can pick out a page for yourself as well. Beyond the recommendations every page of the project is selectable — including one with no impressions at all. The diagnosis runs anyway and costs nothing; only the proposal below it costs credits. If there is no search data, that is the answer, and something can still be written.
- Applied proposals are measured. Rankmio notes the date and compares the click rate before and after — but only at a comparable position. If the page climbed in the meantime, that is stated plainly, because the improvement then belongs to the ranking and not to the title. A result appears once there is enough data behind it, not before.
- The description of every page is now stored. Crawls keep the text of the meta description, not just its length. That is what makes it possible to say what stands beneath the title today — and to tell a page that has none from one that was never checked.
The chat agent has full access to the snippet analysis and answers straight from the chat
- Ask where you rank well but get too few clicks, and the agent answers from the properties described above — including the reasons why a particular page is not a case for a snippet improvement. Proposals can be triggered straight through the agent and cost credits; the agent names the price beforehand.
- It knows what stands on a page. Title, description and structured markup now come with the page data it reads, each distinguishing between "not present" and "never recorded". Where figures come from a crawl that failed to capture a field, they are no longer reported as a finding.
August 21, 2026
Audit · Client report
The site audit: Periods you set yourself, with new charts and an AI report on request
Until now the site audit answered a single question: what is wrong with this website today? For everyday work that is enough. But anyone sitting across from a client needs a second answer — namely what has changed since the last conversation, and how much of it came from their own work. That is what the audit has been extended for: it compares a period you choose yourself and writes a report from it that you can edit and file before passing it on.
Choosing a period and seeing the change
- A period you choose yourself, instead of a fixed window. You set the start and end date. Below them runs a timeline marking every day on which something was actually measured — a snapshot, a GEO audit, a citation scan. Clicking one of those marks moves the start date there, so the comparison rests on a real measuring point rather than an arbitrary calendar day.
- A balance made up of five tiles. Visibility index, clicks, impressions, AI citations and completed tasks sit side by side, each with the change against the equally long period before it, in absolute numbers and in percent. Clicks and impressions combine Google and Bing: a click remains a click, and splitting by search engine answers no question that comes up in a client conversation.
- Every figure states how far its source reaches. If only five months of data exist for a period covering a year, the sum is no longer presented as an annual figure but carries the date since which measurement has taken place. Days without data no longer count as zero. For a report, the difference between a measured zero and a missing measurement is decisive, and it is now made consistently.
The report, in five questions
- You choose the questions before the report is written. Five are available: where do we stand, what has moved, what have we done, what comes next, and what could we not measure. If a question has no data behind it, it stays visible but cannot be selected, and it names the reason — that no measurements exist for the comparison period, for instance. What the report is able to deliver is therefore clear from the outset.
- Every answer sits in its own text field and can be revised. Whatever you change is marked as edited by hand. If a single answer has turned out badly, you can have precisely that one rewritten for one credit instead of regenerating the whole report for five. The remaining sections stay as they are, including your own wording.
- Saving turns the draft into a document. It is kept even when new measurements arrive later; in that case a note simply appears saying the figures have since changed. The revised text goes into the PDF report, unedited drafts do not — a text nobody has read should not be the first thing a client sees. The AI original is kept alongside each section, so it stays traceable what the machine wrote and what you made of it.
August 18, 2026
API · AI visibility
The data export now covers the Google side of AI visibility, too
If you use the JSON export to hand Rankmio data to an AI agent, your own dashboard or a deploy pipeline, AI visibility only ever told you half the story: Bing Copilot was included, Google was missing — even though Rankmio has been collecting those numbers since mid-August. The export now covers both sides, so whatever reads it sees the same picture as your dashboard.
What the export now returns
- New block
citation.google_ai_crossref — total impressions, the 25 most-cited pages, plus breakdowns by country and device and a 90-day trend. It sits right next to the Bing block that was already there, so both sources read the same way.
- Google reports impressions only, no clicks. That is a limitation of the Search Console export, not of Rankmio — the field is documented in the block so nobody reads a missing click count as a zero.
- The data is as fresh as your last upload. Google AI figures come from the Search Console export you feed in under
/geo/gsc-ai; the export block carries no separate schedule.
August 13, 2026
AI visibility · Data quality
AI visibility on Google — plus a cross-check that shows whether you measure the right things
Rankmio now shows how often Google pulls your own pages into AI answers as a source — until now that was visible only inside Search Console. On top of it a cross-check that answers whether your citation tracker watches the right phrases.
AI visibility on Google (new)
- New area under GEO: upload the export from Search Console, and impressions in AI Overviews and AI Mode become visible — by page, country, device and day. Google does not hand these numbers out through its interface, hence the upload.
- A note comes with it: an impression means the page was one of the sources — not that anyone saw or clicked it.
- New view “Real mentions” in the GEO hub: Bing Copilot and Google AI as a trend line with their combined total, kept apart from the percentage figures. Putting both on one axis would render the smaller numbers unreadable.
“Am I measuring the right things?” (new)
- The new cross-check lists pages that show up in Google’s AI answers but are not watched by the citation tracker. Where the two drift apart, the phrases under observation miss what actually happens.
- The chat agent can read both sources and always states which measurement a number comes from.
August 9, 2026
Research v2 · AI labelling
A tidied research area, sturdier crawls and labelling under the EU AI Act
The research area now shows one source at a time instead of three side by side, with a shared history across all of them. Plus: crawls survive a closed browser, Bing data feeds the traffic chart and the visibility score, and the chat agent is permanently labelled as a machine under Article 50 of the EU AI Act.
- One source instead of three: pick search term, your own domain or a competitor at the top — the other tools no longer sit below each other, waiting to be scrolled past. Your choice survives a page reload, so a paid run does not land you on the wrong tool afterwards. Results you already paid for stay in place when you switch back and forth.
- One shared history: “Recent runs” lists every research run across all three sources, filterable by type. Previously each tool kept its own list — to find an old run you first had to remember which tool produced it.
- The same SERP signals everywhere: AI Overview, Featured Snippet and People-also-ask are now shown in all three result tables. Until now the badges lived only in the keyword finder, so exactly where you look at your own rankings there was no hint that Google has since put an AI answer on top.
- More control in keyword research: match types, bulk seeds and an exclusion list, so unwanted terms stop coming back with every run.
- Six series in the traffic chart: clicks and impressions, each split into Google Search Console, Bing and a combined total. The legend chips are buttons — click one to hide or show that line.
- Bing counts towards visibility: clicks and impressions from the Bing CSV are added to the visibility score, so the number reflects visibility everywhere rather than at Google alone. Average position and the top-10/20/50 distribution deliberately stay Search-Console-based — the Bing export carries no per-query positions.
- My Keywords stay current: every DataForSEO pull for your own domain updates the SERP position on keywords you already track. Nothing is added automatically — only existing rows are refreshed.
- Crawls now finish on the server. Sitemap crawl, GEO audit and the competitor deep-dive keep running when you close the browser. If a run fails for good, the credits come back automatically.
AI labelling — EU AI Act, Article 50
- The chat agent is permanently marked as an AI assistant. The header carries the label at all times, and the opening message states in plain words that you are talking to a machine and not to a person. Article 50 has applied since 2 August 2026.
- Terms and privacy policy updated accordingly, in both languages.
Three new articles — written by the Content Studio
- Finding SEO competitors — with real figures from our own Search Console: ten competitor comparison pages, around 11,000 impressions, zero clicks. Why market rivals and SERP rivals are two different lists, and what we changed.
- (not provided) — what Google hid in 2011, why no filter brings it back and what the Search Console returns instead. Deliberately written in the manner of Ernest Hemingway: short sentences, few adjectives.
- Clickbait — a style experiment in the manner of Charles Bukowski. We then asked ChatGPT to criticise the article rather than improve it. The verdict is printed in full at the end, weaknesses included: 7/10 for credibility, and one objection that made us rewrite a sentence.
- All three were produced entirely by the Rankmio Content Studio without human editing, and are labelled as such — author, byline and structured data name the system, not a person.
July 29, 2026
Data Study · Visibility v2
Reality Check 2026, AI visibility scored honestly and Citation Tracker update
Four updates that together sharpen how meaningful our numbers are: a public data study on the state of the web, a fully recomputed AI visibility score with clear data-source attribution, three improvements to the Citation Tracker, and three foundational bug fixes in the GEO audit and JSON export.
- Citation score reworked: Bing Copilot numbers (real user queries from the CSV upload) take priority when fresh. Rankmio watchlist scans and citation discovery scans now count together. Brand phrases carry a discount factor so they no longer dominate the score.
- Sigmoid scale replaces the log cap — 100 is no longer arbitrarily reached. Example Wikipedia: from a hard 100 down to a realistic 96. The value now has room to grow with real citation volume, fair for small and giant sites alike.
- Data-source badge below the citation bar on
/visibility: shows whether the score comes from fresh Bing data, aging Bing data or from the Rankmio tracker alone — plus a warning when Bing CSV data ages out.
- AIO parser fix: DataForSEO restructured the AI Overview response (citations are now in
references[] instead of items[]). The old parser found zero citations even when the domain was clearly cited. Now correct — old cite/mention labels flip on the next scan.
- Bing Copilot Queries as a new discovery source: real user queries from the CSV upload are now offered as phrase suggestions in the Discovery tab — highest-priority source when data is available (weight 0.6 in the distribution).
- Delete individual phrases from Discovery with a click, plus deduplication on re-suggest. Deleted phrases stay gone even after a fresh suggest run.
- IndexNow submission history persistent per URL: the Bing Coverage view now shows „Submitted DD.MM.YYYY HH:MM" under each URL — no longer lost on reload. Buttons switch from „Submit to Bing" to „Resubmit" once a URL has been sent.
- GEO audit skipped the crawl on certain sitemap imports — root cause of the 100 %-miss classifications some projects saw. The bulk sitemap-import copied cached fields that made the crawler treat all pages as already-analyzed. Fixed.
- JSON export: seven aggregate/ghost bugs in the snapshot fixed — including sitemap stats that returned inflated numbers and summary hallucinations when a block had zero data. The API contract is now aligned with the browser download.
- Team auto-todos: no more duplicates when several team members share a project. The auto-created todo list is now isolated per project rather than per user — three team members no longer see the same item three times.
July 19, 2026
JSON-API v1 · Devs
JSON API v1 for devs — Rankmio data from Claude Code, Cursor and CI
The JSON export now has an HTTP twin. Three endpoints under <code>/api/v1/</code> give AI coding agents (Claude Code, Cursor, Aider) and CI pipelines access to the same SEO and GEO data as the browser download — with a bearer token, no CSRF, no cookie. Discovery endpoints are free, the export costs the same 2 credits as in the browser. Managed in the profile.
Three endpoints, one bearer
GET /api/v1/projects — lists all projects the key can see (id, domain, name, created_at). Discovery for agents that do not know your project layout yet. 0 credits.GET /api/v1/projects/{id} — meta and available blocks per project: which of the 12 blocks have data, count of records and estimated bytes per block. Lets an agent decide the export scope before it costs anything. 0 credits.GET /api/v1/projects/{id}/export?blocks=… — the actual snapshot in the exact same shape as the browser JSON download. 2 credits per call, price is dynamic from CreditService.
Auth, limits and abuse protection
- API keys under
Profile → API keys: create, rotate, revoke. The plaintext key (rmk_…) is shown once and stored only as a SHA-256 hash — industry standard (GitHub, Stripe, OpenAI). Prefix plus last four characters remain visible for identification.
- Rate limit 60 requests per minute + 1000 per day per key. On top a daily charge cap (default 200 credits per key per day) — a leaked key cannot drain your balance. Response headers
X-RateLimit-Remaining-* and X-Charge-Used-Cr tell every pipeline where it stands.
- Error responses follow HTTP conventions:
401 unauthorized, 402 insufficient credits, 403 forbidden (IP/project scope), 404 project not accessible, 429 rate or charge limit hit. Every call is logged in api_key_usage — including failures, for auditing.
- Full workflow in the insights piece „Why Claude Code only reaches its full potential with SEO and GEO context" — including curl examples and a Claude Code slash command.
- Typical use cases: CI pipeline pulls a snapshot before deploy and asserts no GEO regression; Claude Code slash command loads the export and opens PRs per finding; cron job diffs weekly snapshots for a Slack digest.
- MCP server (Model Context Protocol) is on the roadmap but not shipped yet — the JSON API covers 90 % of the agent use cases already.
July 14, 2026
GEO · Actions · Bulk-Workflow
AI Champions, Traffic Recovery, Actions master list and a new Bing Copilot tracker mode
Four new panels that together show where Rankmio has impact: <strong>/geo/ki-champions</strong> finds pages AI systems cite while Google does not rank them — your most profitable backlink targets. <strong>/traffic-recovery</strong> splits GSC keywords with impressions but no clicks into three action buckets by position range. <strong>/actions</strong> becomes the consolidated todo master list across all data sources (GEO, GSC, Bing, Backlinks, Competitors, AI Champions) — with URL filter in the Dashboard style. Plus, the citation tracker gets a third mode that mirrors real Bing Copilot queries into the watchlist.
AI Champions — a new Rankmio USP
- New route
/geo/ki-champions — cross-source detector: URLs on your domain that Bing Copilot has cited at least 3 times in real user queries, while Google either does not rank them or ranks them poorly. Combines bwt_ai_performance with gsc_keyword_data.
- Three categories: Hidden (Bing cites, Google has no impressions), Page 5+ (Pos 30+ — highest ROI for backlink outreach), Page 2-3 (Pos 20-29 — often 1-2 backlinks away from Top 10). Champion score combines citation strength (log-scaled) with position penalty.
- Why this matters: content the AI has already validated has proven quality signals. Investing in classic SEO (backlinks, internal linking, snippet) on these exact pages pays back fastest — much better than promoting random pages without validation.
Traffic Recovery — 3 buckets by position
- New route
/traffic-recovery replaces the vague „CTR emergency" concept. Splits GSC keywords with impressions but few clicks into three actionable buckets based on Google position.
- Position 1-10 — snippet tuning: keywords already in the Top-10 but with CTR below the position baseline (Advanced Web Ranking 2024). Shows baseline vs. actual CTR and the potential click gain if the baseline were reached.
- Position 11-30 — rankings push: close to Top-10, hitting the ranking sweet spot. Action: backlinks, internal linking, content refresh. Position 31+ — rewrite or deprioritize: too far to fix with snippet or link work. Either content rewrite or drop the keyword.
Actions — the consolidated todo master list
- New route
/actions merges the ~400 on-the-fly Dashboard-Todos (competitor, battle, GEO audit, sitemap, broken links, site audit) with the auto-generated feeder todos (Bing coverage, GEO, GSC traffic, backlinks, AI champions, cannibalization, sweet spot). One list, one place — the Dashboard timeline and Site-Audit chart remain untouched.
- Filters: status (open/completed/dismissed/all), priority (high/medium/low), type chips (GEO, GSC, Bing, Backlinks, Sitemap, Site-Audit, Competitors, AI Plan, AI Champion, Local), URL filter with search-as-you-type (identical to the Dashboard picker), full-text search across title and detail. All filters are bookmarkable via query params.
- Complete + Dismiss: „Complete" writes into
todo_completions — the Dashboard timeline picks it up automatically. „Dismiss" hides a feeder todo from the Open list; the next sync respects the user intent and does not recreate it.
- Two new detectors: Cannibalization flags URLs with ≥20 keywords stuck on Pos 30+ (Google cannot decide which keyword to prioritize — fix via targeted internal linking). Sweet Spot Top-10 flags URLs whose primary keyword ranks Pos 4-10 — a single backlink or meta refresh could push them into Top-3.
Citation Tracker: Bing Copilot mode
- Third source-mode next to „GSC Top-N" and „Manual watchlist": Bing Copilot Queries. Automatically mirrors the real user queries from the Bing Copilot report (via CSV upload) into the 5-engine probe watchlist.
- Why: the classic GSC-Top-N watchlist probes queries you rank for on Google — not necessarily the questions AI systems actually ground with. The Bing-Copilot mode probes the exact queries with a proven citation history. Cite rate is significantly higher than the GSC-based watchlist.
- Setup: In the Citation Tracker, pick the third mode tab („Real Bing Copilot queries") → Configuration → Save. Requires at least one CSV upload in
/geo/bing-ai.
Visibility index: adaptive Bing blend
- The citation sub-score now adapts to the data situation. Default blend stays at 90 % Rankmio tracker / 10 % Bing Copilot — but if the Rankmio cite rate is under 5 % AND Bing Copilot shows strong signals, the weight shifts to 40/60 in favor of Bing. Reason: an under-reporting Rankmio watchlist should not water down real Copilot citations.
- Diagnostic transparency: the visibility export shows
weights_used, adaptive_weight_applied and adaptive_reason — no hidden math.
- Monday morning digest email per project: „What did you complete last week", top-5 high-prio todos, Bing Copilot 7-day delta with arrow, top-3 AI Champions, snippet wins with click potential. Deep links to
/actions, /geo/ki-champions, /traffic-recovery. Beta phase: only for the account owner.
July 11, 2026
Bing WMT · Deep Integration
Bing coverage, IndexNow quota, 30-day chart and 10 % Bing share in the visibility index
Bing gets its own place in the entire Rankmio workflow. Instead of a URL input field for IndexNow, the Sitemap Hub now delivers a full coverage report: for every URL from your latest sitemap crawl you see the Bing status („indexed / known-but-not-indexed / unknown / check pending") — plus one-click submit and bulk selection. The live daily and monthly quota from Bing’s <code>GetUrlSubmissionQuota</code> is prominently shown and disables the submit buttons the moment the quota is exhausted. On top: a 30-day crawl-stats chart, six new Bing tools for the chat agent, and a 10 % Bing share in three sub-scores of the visibility index.
- New coverage view under
/bing-search/{pid}/coverage: table of up to 500 sitemap URLs × Bing indexation status. Five status filters, five KPI tiles with mouseover hints on how to act, bulk-select „all unknowns", per-row submit button.
- Status heuristic: DocumentSize > 0 and LastCrawledDate present → „indexed", any Bing signal (AnchorCount > 0 or LastCrawledDate) but DocumentSize = 0 → „known, not indexed", nothing → „Bing unknown". Cache TTL 24 h per URL — button „Bing-Status aktualisieren" runs up to 40
GetUrlInfo calls per click.
- Sitemap Hub: the old URL input field is gone. In its place a compact CTA card „Open Bing coverage" that jumps straight to the new report.
Live IndexNow quota + auto-lock
- Live quota card at the top of the coverage view: red when daily quota is 0, amber below 20, blue otherwise. Shows daily + monthly balance from Bing’s
GetUrlSubmissionQuota plus how many URLs Rankmio has submitted today.
- Buttons lock automatically when the daily quota hits zero. The bulk submit and every row-level „Submit" turn into a grey „Kontingent leer" chip with a mouseover explaining that Bing resets around midnight UTC. This prevents a stream of HTTP-400 „InvalidApiKey" responses (Bing’s misleading error text for quota-exhausted).
- Solo fallback when the batch endpoint rejects a bulk submit: Rankmio retries each URL individually (Bing sometimes rejects batches with HTTP 400 even though the same URLs go through as solo calls). Response indicates
mode: per_url_fallback or partial so users see what happened.
- Toast feedback — the earlier scroll-dependent banner is replaced by a fixed toast top-right (green success, red error, six-second auto-hide) so results stay visible no matter how far the user has scrolled.
- Line chart in the Sitemap Hub Bing card with three series: Bing crawls (blue, filled), In-Index (green), inbound links (amber). Toggle „Errors" switches to a bar chart of
code_4xx + code_5xx + blocked_by_robots per day. Data grows daily through the health-sync cron.
Visibility index: 10 % Bing share
- Traffic sub-score: 90 % Google clicks (GSC) + 10 % Bing clicks (Bing Search Performance CSV, normalised to a 30-day monthly proxy). Log scale, so 10 Bing clicks don’t swing the total.
- SERP sub-score: Google top-10 ratio blended 90/10 with the impression-weighted Bing top-10 ratio from the Bing top-queries CSV. Count-score (breadth) stays Google-only because the Bing keyword sample is usually much smaller.
- Citations sub-score: 90 % Rankmio citation tracker (ChatGPT/Claude/Gemini/Perplexity/Google AIO) + 10 % Bing Copilot first-party citations (BwtAiPerformance, log-scaled to 500 monthly citations = 100 points). URLs that were literally cited by Copilot users finally count.
- Fallback semantics: no Bing site linked → 100 % Google (external domains unchanged). Only Bing AI Performance available but no Rankmio citation tracker → score reflects only the Copilot view, with a diagnostic note.
Chat agent gets 6 Bing tools
- get_bing_search_performance / get_bing_top_queries / get_bing_top_pages — direct answers to „how is my Bing visibility?", „which queries drive Bing traffic?", „which of my URLs perform best on Bing?"
- get_bing_coverage — Bing indexation status across the current sitemap crawl, with the top-20 unknown URLs when requested.
- get_bing_indexnow_quota — live daily and monthly balance, submissions today, exhausted flag with reset hint.
- get_bing_backlinks — top referring domains and anchors from the Bing WMT backlink CSV, ready for cross-check with the DataForSEO view.
- Auto-preload: as soon as the user’s question contains „bing", „microsoft-suche", „copilot-citation" or „indexnow", the top three tools fire in parallel and land in the LLM context. Answer includes real numbers on the first shot — no extra tool round.
- bing_url_index_sync (04:00) — auto-fills the coverage cache: for every project with a Bing site and a done sitemap crawl, up to 40 URLs are checked via
GetUrlInfo. Cache TTL 24 h, oldest/unchecked URLs first. Result: the coverage view fills up over days without a single user click.
- bing_backlinks_sync (03:30) — the deep-sweep Bing-backlinks script existed but was not scheduled. Now registered daily. Pulls
GetLinkCounts plus GetUrlLinks for the top-20 pages, feeds bing_backlink_snapshots. Skipped when the last sync is younger than 20 h.
- Cron registry updated: the admin cron overview finally lists all three Bing jobs (health-sync, backlinks-sync, url-index-sync) — two were missing.
Bing CSV parser: German headers + DD.MM.YYYY dates
- Bing Webmaster Tools export column headers in the property’s language. The Search Performance CSV parser now recognises German headers (Datum, Klicks, Aufrufe, Durchschn. CTR, Suchanfrage, Seite, Anker …) alongside the English ones and normalises DD.MM.YYYY dates explicitly (strtotime interprets „.” unreliably by locale).
July 8, 2026
AI Visibility · Cross-Engine
Bing Copilot as 6th engine in the Citation Tracker — with multi-engine stars
The Citation Tracker now unifies two data streams in one view: Rankmio probes against ChatGPT/Claude/Gemini/Perplexity/Google AIO (actively measured) and Microsoft first-party citations from Bing Webmaster Tools (passively observed). The new load-bearing signal: URLs cited in <strong>both</strong> worlds get a yellow star — and the Bing card shows „X / Y multi-engine URLs" as its own KPI. No forced aggregate score because the metrics are not directly comparable — instead an honest side-by-side view with clear provenance badges.
What is new in the Overview tab
- New blue Bing-Copilot card between the probe hero and the phrase performance: „🎯 Observed · First-Party" badge, 4 KPIs (30-day citations with delta, cited URLs, grounding queries, multi-engine count) and a deep-link to the full Bing AI dashboard.
- Multi-Engine star on URLs that appear in both data sources — yellow star icon and yellow bar in the URL performance list. That is the trustworthy signal: a URL cited in real Copilot answers AND probed engines is no random hit.
- Context note below the Bing card that clearly separates the two data logics: „probes = we ask engines" vs. „first-party = Microsoft reports what real Copilot users saw". Prevents naive addition of the numbers.
- Empty state (no Bing CSV yet) links directly to the upload — the Citation Tracker becomes the natural entry point for both engines.
- Rankmio probes measure „in X of Y sample queries the domain was cited" (percentage of a controlled sample). Bing first-party measures „N real Copilot citations across all real users" (absolute count of live events). Dividing or averaging them would suggest a precision that is not there.
- Instead: side by side with clear badges („🎯 Observed" vs. „🔍 Probed") and the multi-engine star as the reliable cross-source signal. This stays honest and still delivers the „one place for AI visibility" experience.
July 8, 2026
Backlinks · Bing WMT
Bing WMT cross-check: CSV upload flow now smoother and clearer
The Backlinks Hub gets a UX polish around the Bing Webmaster Tools CSV uploads. Uploading a referring-domains or anchor CSV now keeps you on the active tab, shows a success banner with row and backlink counts, and the blue info bar tells you at a glance when the last upload happened. The Bing view also stands on its own — no DataForSEO snapshot required.
- Tab stays put. After uploading an anchor CSV in the Anchors tab, you land back in the Anchors tab — same for Referring Domains. Before the fix the redirect fell back to „Overview".
- Success banner with counts. Green banner above the tab bar reports „Bing-CSV imported: file.csv: Anchors (6 rows, 248 backlinks)" — so you know immediately what landed in the snapshot.
- Upload date in the blue bar. When Bing data exists, the blue banner above the upload button shows the last upload timestamp plus counts („Letzter Anchors-Upload: 08.07.2026, 15:31 · 6 Anker · 248 Backlinks — neue CSV überschreibt die Werte."). Empty state keeps the familiar call-to-action.
Bing-only view without DataForSEO
- The cross-check merge is now built the moment a Bing site is linked — no DataForSEO snapshot required first. Rows that only exist in the Bing CSV are marked
source=bing and rendered like any other row.
- Filter chips (All / Both / Only DFS / Only Bing) work independently of DataForSEO. Uploading Bing anchors alone gives you a complete Bing-only anchor table right away.
July 2, 2026
GEO · Bing AI
Bing AI Performance — import first-party Copilot citations
Bing Webmaster Tools has been shipping an AI Performance report since February 2026 — Copilot citations with grounding queries, intents and topics. Microsoft is the first search platform with first-party AI citation reporting. Since there is no official API yet, Rankmio imports the CSV export from BWT and renders the data in the new dashboard at <code>/geo/bing-ai</code>. The parser handles overview + detail reports and German + English headers.
- Trend chart — Citations and cited pages per day over the last 90 days.
- Top URLs & queries — which of your pages get cited, and which grounding queries Copilot generates to find them.
- Intents & topics — search-intent grouping and topic clusters from the June 2026 BWT update.
- KPI card in GEO Hub — quick view of total citations, 7-day delta and URL count with deep-link to the full dashboard.
- Open Bing Webmaster Tools and switch to the „AI Performance" section.
- Export the report as CSV (both overview and cited-queries formats are supported).
- Upload the CSV under
/geo/bing-ai/upload — parser auto-detects delimiter (comma / semicolon / tab).
- Repeat monthly (or whenever you want fresh data). Old uploads stay accessible in the upload history.
June 30, 2026
UX · Sidebar
GEO as a dedicated main section in the sidebar
GEO is now its own top bucket in the sidebar — placed right after Analyze. All GEO tools live in one spot, ordered as diagnosis → answer simulation → tracking → competition. Previously GEO was hidden as a sub-subgroup inside „Analyze", the Citation Tracker sat under „Monitor", and Battle was filed under „Competitors". URLs are unchanged — only the bucket position in the sidebar has been consolidated.
New GEO bucket — 5 tools, one place
- 1. GEO Audit (
/geo) — full-site GEO scan with topic coverage, AI readiness, schema check.
- 2. Citability Check (
/seiten-analyse) — single-page deep dive on citation readiness (was hidden under „Own Site" before).
- 3. Answer Simulation (
/answer-simulation) — preview how AI engines answer real user questions about your topic.
- 4. Citation Tracker (
/citation-tracker) — continuous citation monitoring across ChatGPT, Gemini, Perplexity, Claude, Google AIO (moved out of „Monitor").
- 5. GEO Battle (
/competitors?mode=battle) — AI-citation showdown against pinned competitors (moved out of „Competitors", URL kept for bookmarks).
- USP visible from the sidebar. Rankmio is no longer „yet another SEO tool" — the GEO + Citation differentiator is now a top-level bucket on first login.
- Workflow clarity. Anyone doing GEO no longer clicks across three different buckets. All five tools sit in one place, in workflow order.
- „Monitor" stays „Monitor". Position Tracker and Sitemap Monitor are the only items left — clean focus on passive monitoring instead of mixing in active GEO tools.
June 30, 2026
Feature · Dev
JSON data export for AI coding agents (Claude Code, Cursor, Aider)
The full SEO+GEO state of a project can now be downloaded as a structured JSON export — 11 data blocks (audit, PageSpeed, visibility, GEO, citation, backlinks, competitors, keywords, sitemap, actions, project meta), directly consumable by AI coding agents. Plus: new insights article "Why Claude Code only reaches its full potential with SEO & GEO context" with slash-command example and MCP outlook. Plus: 6 new glossary entries (Claude Code, Cursor, GitHub Copilot, Aider, Lovable, Vibe Coding) for the developer audience.
JSON export: 11 data blocks, selectable per download
- New "JSON-Export" button in the project header (dashboard and analysis pages). Modal lets you pick which blocks to include — see record counts and estimated kilobytes per block, plus running total.
- 11 blocks available: project meta, on-page audit + crawler, PageSpeed mobile + desktop, visibility index, full GEO audit (pages + scores + schema + link suggestions), citation tracker (scans + results across 5 LLM engines), backlinks (summary + top + anchors + ref-domains + gaps), competitors (tracked + battles), keywords (GSC + opportunity scores + project keywords), sitemap (crawl + URL list), priorized actions.
- Cost: 2 credits per download. Eigene Permission json_export (Pro/Admin/Credits enabled, Visitor/Basis demo-only). CSRF-protected POST, no caching — always returns latest data.
Insights article + 6 new glossary entries
- New insights piece /insights/seo-fuer-ai-agenten with Claude-Code slash-command example, vibe-coding workflow walkthrough, SEO-report vs. Website-Context comparison table, 6 dev FAQs and MCP outlook. Hero image showing the rankmio + Claude Code symbiosis.
- Six new glossary entries for the developer audience: Claude Code, Cursor, GitHub Copilot, Aider, Lovable, Vibe Coding — each with definition, sections, example, and FAQ block.
- Landing page gets a dedicated "for-devs" section with the hero image and CTA to the new article.
June 28, 2026
Feature · Fix
New "Antithesis" brief slot + library self-domain fix
A new "Antithesis" template-element type with three hard prompt rules forces the LLM into a real counter-thesis instead of middle-of-the-road platitudes. Plus: rankmio.de-owned URLs can now be loaded as library sources into the Content Studio (self-domain block lifted in three layers).
Content Studio: new "Antithesis" brief slot
- New template element of category "Hook & Start" with three hard prompt rules: no hedging words (no "often", "can", "tends to"), explicit naming of the opposing camp (e.g. Yoast users, Ahrefs buyers, old-school SEOs), no "but of course" retreat. Renders as an orange highlight box at the top of the article.
- Auto-attached to the standard "Argumentation essay" template (position 2, after the main thesis). Existing briefs using this template inherit the slot without further action.
- New element appears automatically under "Hook & Start" in the template editor — no UI changes required, the catalog is loaded dynamically.
Content Studio Library: rankmio.de URLs now allowed
- Loading your own rankmio.de pages (glossary, insights, features) as library RAG sources is now possible. The SSRF self-domain block had to be lifted in three layers (uploadUrl, importUrl, DocumentProcessorService::fetchWithSsrfGuard).
June 15, 2026
v1.0 · Official launch
Rankmio v1.0 — all essential SEO + GEO features delivered
With today's release Rankmio officially reaches v1.0. All pillars are delivered and hardened: SEO analysis, Keyword Hub, Competitor Data Package, Backlinks Hub including Gap analysis, Bulk-First Position Tracker, Citation Tracker for ChatGPT/Gemini/Perplexity/AIO/Claude, Content Studio with voice dictation, Chat Agent with live access to all trackers. Shipped today: the Backlink Gap tab with anti-spam filter and score formula, 12-month trend chart, 9 new chat-agent tools and a sidebar cleanup.
- Find domains that link to your pinned competitors but not to you — your outreach pipeline at a glance.
- Spam and off-market domains are filtered out automatically so the list stays usable. One click brings them back if you want to see the full set.
Backlinks: 12-month trend chart
- New mini chart shows how backlinks and referring domains develop over the year — at a glance, next to the existing table.
Chat agent: 9 new tracker tools
- Position Tracker: overview (SoS, distribution, quick-wins), combined movers (Position-Tracker + GSC with source tag), competitor gap.
- Citation Tracker: rates per engine, 0%-phrases (highest-value action triggers), top competitor domains in AI answers.
- Backlinks: health score, top gap domains (spam filter on by default), anchor distribution with brand/exact/generic estimate.
- „AI Visibility" sidebar link removed (was a 301 redirect to the Citation Tracker — pure duplicate).
- „Rank Alerts" direct link removed — the tab inside the Keyword Hub still works, GSC-based email alerts unchanged.
June 11, 2026
Major Release
SERP, competitor depth & Backlinks Hub
Six new modules on top: full SERP data in the SEO analysis, competitor data packages with up to 1,000 keywords per domain, a dedicated Backlinks Hub, bulk-first Position Tracker, AI Visibility Monitor and competitor keywords straight inside the Content Studio brief.
- Dedicated hub for your backlink profile with health score, anchor distribution, referring domains and toxic check.
- Six tabs at a glance — overview, backlinks, anchors, referring domains, toxic and gap.
New: Competitor data package
- One click pulls up to 1,000 ranking keywords per competitor — with position, search volume and URL.
- In the detail view your own position and GSC data appear side by side with the competitor ranking.
- Multi-sortable columns, checkbox selection, direct transfer into your project keywords.
New: Position Tracker (bulk-first)
- Tracks your keyword rankings over time. Bulk-first architecture: up to 1,000 keywords per domain in a single call instead of one-by-one — faster and significantly cheaper.
New: AI Visibility Monitor
- Watchlist for your visibility in ChatGPT, Gemini, Perplexity and Google AIO.
- Per phrase: engine status, competitor mentions, answer excerpt — connected to a Brief action.
New: Rankings tab in the SEO analysis
- Fresh SERP data and GSC data combined in one table with multi-sort.
New: Project competitor register
- Central competitor list per project — auto-populated from your Easy and Deep SEO runs.
- Pin your top 1–3 rivals — Briefs, the chat agent and data packages all consume them automatically.
Content Studio: competitor keywords in the brief
- When creating a brief you can pull keywords directly from your pinned competitors into the target keywords — no detour via the Keyword Hub.
Chat agent reads competitor data packages
- Three new tools: pinned competitors, their ranking keywords, overlap/gap analysis.
- Model upgrade to GPT-4.1-mini for faster, more concrete answers.
- 4-bucket sidebar: Analyze · Optimize · Content Studio · Monitor — plus Tech and Manage. Same content, clearer workflow.