GEO / AI search

Vibe Coding

Q: Who coined the term vibe coding?

Andrej Karpathy in an X post from February 2025: ”There's a new kind of coding I call vibe coding, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists." Karpathy is a former OpenAI and Tesla AI lead and one of the best-known AI researchers.

Q: Is vibe coding sustainable for production code?

Pure vibe coding without reviews is not recommended for production. Risks: technical debt, security gaps through hallucinated dependencies, difficult maintenance. Best practice in 2026: vibe coding for prototypes, spikes and boilerplate — classic reviews and tests for production paths.

Q: Which tools are suited to vibe coding?

The most important vibe coding tools in 2026: Claude Code (CLI-first), Cursor (IDE), GitHub Copilot (plugin), Aider (open source), Lovable + Bolt + v0 (browser-based), Replit Agent (cloud). The choice depends on the use case: browser-based for prototyping, CLI- or IDE-based for production refactoring.

Q: How well do vibe coding and SEO go together?

Poorly out of the box. AI coding agents prioritise functional code, not SEO-compliant code. Schema.org, FAQ markup and semantic hierarchies are rarely produced on their own. Anyone wanting to combine vibe coding and SEO should give the agent structured website context — for example via a Rankmio JSON export — along with concrete SEO/GEO instructions.

Also known as: AI-First Coding, Prompt-First Development

Vibe coding is the term for a development style in which AI coding agents write the majority of the code, while the human steers primarily through prompts, reviews and acceptance. Andrej Karpathy coined the term in early 2025 in a viral X post: ”I just say what I want, accept all suggestions, and the vibes are good." What he describes is less a tool than a changed role for the developer.

Characteristics of vibe coding

Prompt-first — requirements are formulated in natural language.
High-trust acceptance — AI suggestions are usually adopted directly.
Rapid iteration through prototypes instead of lengthy planning phases.
Tool stack: Claude Code, Cursor, Aider, Lovable, Bolt, v0, Replit Agent.
Shift in competence from writing to reviewing and curating.

Criticism and limits

Vibe coding is discussed controversially. Advocates see it as a democratisation of software development and a leap in productivity. Critics warn of technical debt caused by unreviewed AI code, difficulties in onboarding new developers and security gaps through hallucinated libraries. In practice, 2026 sits somewhere in between: vibe coding for prototypes and boilerplate, classic review for production code.

Vibe coding and SEO/GEO

SEO is a blind spot in vibe coding. AI coding agents produce functional code, but rarely search engine optimised code. Schema.org, semantic heading hierarchy and citability structures have to be requested actively. Structured website context from an SEO tool such as Rankmio closes this gap by giving the agent concrete knowledge about where optimisation is needed.

Example from practice

Example: A solo developer builds a newsletter sign-up page. Vibe coding workflow: 30 minutes of prompts to Lovable for the page, 15 minutes in Cursor composer mode for refinement, 10 minutes of Rankmio for an SEO/GEO audit, 20 minutes of Claude Code with the Rankmio export for Schema.org patches and FAQ schemas. Result after 75 minutes: a page capable of ranking and of being cited by AI — with 0 manually written lines of code.

Frequently asked questions

Who coined the term vibe coding?

Andrej Karpathy in an X post from February 2025: ”There's a new kind of coding I call vibe coding, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists." Karpathy is a former OpenAI and Tesla AI lead and one of the best-known AI researchers.

Is vibe coding sustainable for production code?

Pure vibe coding without reviews is not recommended for production. Risks: technical debt, security gaps through hallucinated dependencies, difficult maintenance. Best practice in 2026: vibe coding for prototypes, spikes and boilerplate — classic reviews and tests for production paths.

Which tools are suited to vibe coding?

The most important vibe coding tools in 2026: Claude Code (CLI-first), Cursor (IDE), GitHub Copilot (plugin), Aider (open source), Lovable + Bolt + v0 (browser-based), Replit Agent (cloud). The choice depends on the use case: browser-based for prototyping, CLI- or IDE-based for production refactoring.

How well do vibe coding and SEO go together?

Poorly out of the box. AI coding agents prioritise functional code, not SEO-compliant code. Schema.org, FAQ markup and semantic hierarchies are rarely produced on their own. Anyone wanting to combine vibe coding and SEO should give the agent structured website context — for example via a Rankmio JSON export — along with concrete SEO/GEO instructions.

Used in Rankmio for

Vibe coding with an SEO/GEO layer — the workflow in detail

Go to the feature →

Last updated: 2026-06-30 · Browse all glossary entries