Google ranks you. AI decides whether to quote you.

Q: Why can AI not quote my site?

The three most common causes are: content that renders only after client-side hydration, since most AI crawlers do not execute JavaScript; a robots.txt that blocks retrieval crawlers such as OAI-SearchBot and PerplexityBot alongside training crawlers; and pages whose sections carry no self-contained answer, so no passage is worth quoting.

Q: Do I need to block AI crawlers?

Blocking training crawlers such as GPTBot, ClaudeBot and CCBot is a legitimate policy choice. Blocking retrieval crawlers is almost never what a site wants, because retrieval crawlers are how answer engines fetch a page in order to cite it. WideAngle reports the two groups separately so the decision stays yours.

Google ranks your site. AI answer engines decide whether to quote it. WideAngle is a site auditing tool that scores both jobs across 126 rules — 82 for SEO, 44 for Generative Engine Optimization. It tells you which signals stop Google ranking you and which stop ChatGPT, Claude and Perplexity quoting you, then applies the fixes that can be applied safely.

Founding users get three months of Pro, free.

wide-angle audit https://example.com

  Overall  █████████████████░░░   85  B+
  SEO      ███████████████████░   93  A
  GEO      ███████████████░░░░░   77  C+

  GEO categories
    AI crawler access          ████░░░░░░░░  32
    Entity clarity             ██████████░░  81
    Machine extractability     ██████████░░  87
    Citation authority         ████████████  98
    Answer readiness           ████████████  99

  × Answer-engine retrieval crawlers can reach the site   geo/retrieval-crawlers-allowed
      site
      PerplexityBot (Perplexity) is disallowed from the site root.
      → Allow OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User,
        PerplexityBot and Perplexity-User in robots.txt. Block training
        crawlers separately if that is the policy you want.

Key takeaways

01SEO and GEO are different jobs. SEO decides whether Google ranks a page. GEO decides whether an answer engine quotes it.
02Retrieval crawlers are not training crawlers. Blocking OAI-SearchBot or PerplexityBot removes a site from AI answers with no upside.
03Most AI crawlers do not run JavaScript. Content that appears only after hydration does not exist to them.
04Retrieval happens per passage, not per page. A section that cannot stand alone never becomes the quoted passage.
05WideAngle checks all 126 signals — 82 SEO rules and 44 GEO rules — and reports the fix beside each finding.

Definition

What is Generative Engine Optimization?#

Generative Engine Optimization (GEO) is the practice of making a website readable, trustworthy and quotable by AI answer engines such as ChatGPT, Claude, Perplexity and Google AI Overviews. It shares a foundation with classic SEO — both want crawlable, well-structured pages — but it asks questions SEO never had to. In building WideAngle we found the split is not marginal: 44 of its 126 rules exist only because an answer engine reads a page differently from a search engine, and 18 distinct AI crawlers have to be checked separately in robots.txt. GEO is measured against the llms.txt convention, per-agent robots.txt policy, and schema.org entity markup.

How the two disciplines differ, by the question each one answers.
Discipline	Decides	Audience	Unit of retrieval	Fails when
SEO	Whether you rank	Googlebot, Bingbot	The page	Slow, unindexable or duplicated pages
GEO	Whether you are quoted	OAI-SearchBot, Claude-SearchBot, PerplexityBot	The passage	Blocked retrieval, JS-only content, unresolvable entity

Use WideAngle to audit one thing. Or everything.

AI crawler access
Machine extractability
Answer readiness
Citation authority
Entity clarity

Join the waitlist

Coverage

What does WideAngle check?#

WideAngle checks 126 rules split across two scored pillars. Every finding states what is wrong, why it matters, and what to change. The report is a single self-contained HTML file you can send to anyone.

SEO

82 RULES

The SEO pillar covers whether Google can crawl, understand and rank the site.

Meta and social88
Content structure76
Links and crawl graph92
Images and media70
Structured data84
Technical95
Internationalisation64
Performance signals80

GEO

44 RULES

The GEO pillar covers whether ChatGPT, Claude, Perplexity and AI Overviews can read, trust and quote it.

AI crawler access32
Machine extractability87
Answer readiness99
Citation authority98
Entity clarity81

Includes per-agent robots.txt policy across 18 AI crawlers, llms.txt validity and coverage, and server HTML versus post-JavaScript content.

The gap

Why do answer engines skip a site?#

Content only exists after hydration

Most AI crawlers do not execute JavaScript. Content that appears only once the client-side bundle runs does not exist to them — they see an empty shell.

geo/empty-html-shell

The wrong bots were blocked

Retrieval crawlers are how answer engines fetch a page in order to cite it. They are a different thing from training crawlers, which a site may legitimately want to block. Blocking retrieval removes the site from AI answers with no upside.

geo/retrieval-crawlers-allowed

Pages do not chunk

Retrieval happens at passage level, not page level. A section that does not stand on its own — no answer-first lead, wrong size, context two headings away — never becomes the quoted passage.

geo/answer-first · geo/chunk-size

Platform

Built for the framework you already ship.#

WideAngle findings point at the file and the line, not just the URL. It reads your routes, your metadata exports and your config the way your framework does.

The fixes that can be applied safely get applied — as a reviewable pull request against your repo, with the diff and the reasoning attached. You merge it or you don't.

Scheduled crawls turn one audit into a trend line. When a deploy drops your GEO score, WideAngle names the rule that regressed and the commit window that did it.

Each report is one self-contained HTML file, with no external requests and no dashboard login. Hand it to a client, a stakeholder, or the developer who has to do the work.

126 rules

Across five GEO categories and eight SEO categories, checked on every page of every crawl.

Next.jsAstroNuxtSvelteKitRemix

FAQ

Questions people actually ask#

Is GEO different from SEO?

GEO and SEO overlap but are not the same job. Both want crawlable, well-structured pages. Only GEO cares whether retrieval crawlers are allowed, whether content survives without JavaScript, whether sections stand alone when quoted out of context, and whether the brand resolves to a single identifiable entity.

Why can AI not quote my site?

The three most common causes are content that renders only after client-side hydration, since most AI crawlers do not execute JavaScript; a robots.txt that blocks retrieval crawlers alongside training crawlers; and pages whose sections carry no self-contained answer, so no passage is worth quoting.

Do I need to block AI crawlers?

Blocking training crawlers such as GPTBot, ClaudeBot and CCBot is a legitimate policy choice. Blocking retrieval crawlers is almost never what a site wants, because retrieval crawlers are how answer engines fetch a page in order to cite it. WideAngle reports the two groups separately, as documented in OpenAI's crawler reference, so the decision stays yours.

Which frameworks does WideAngle support?

WideAngle is framework-aware for Next.js, Astro, Nuxt, SvelteKit and Remix, so findings point at the file and the line rather than only the URL. It audits any reachable site over HTTP regardless of framework.

What does WideAngle cost?

The command-line auditor is free and Apache-2.0 licensed. The hosted platform — scheduled crawls, score history, deploy alerts and pull-request fixes — is in early access; waitlist members get three months of Pro at no cost when it opens.

Why we're building it

Search stopped being a list of links.#

Classic SEO decides whether Google ranks you. It says nothing about whether an answer engine can parse your page, resolve your entity, or find a passage worth quoting. Those are different questions, and almost no tooling asks them.

Every audit tool on the market was designed for a search results page. Meanwhile the answer is being written above it, from sources chosen by rules nobody is checking.

We learned this the hard way. Running early builds against live sites, the pattern repeated: pages scoring above 90 on classic SEO were scoring under 40 on retrieval access, usually because one line in robots.txt blocked all six retrieval crawlers at once. Testing against an Arabic site surfaced a second class of problem — a readability score computed with English syllable heuristics, producing a confident and meaningless number.

Hosted WideAngle is in early access: scheduled crawls, score history, deploy alerts, and a bot that opens the fix as a pull request. The command-line auditor stays free and open source.

No spam — one email when it opens.