Back to the blog
· SEOWrench

How SEOWrench measures GEO (technical AI readiness)

Answer engines — ChatGPT, Claude, Perplexity, Google's AI Overviews, Copilot — read your site very differently from a human or the classic Googlebot. GEO (Generative Engine Optimization) is the technical groundwork that decides whether they can use your pages at all.

Most tools that sell "GEO" lean on hype. We built ours on a simple, honest principle: the technical layer is necessary, but not sufficient. We measure — with precision — what is technical. We signal what depends on content and authority. We never promise a citation, because nobody can.

Here's exactly how our GEO score works.

The method: access → read → interpret

Every GEO check maps to one of three steps. Each depends on the one before it — if an AI can't access your page, it doesn't matter how good your schema is.

1 Access Can the AI bot reach your page? (robots.txt) 2 Read Is the content in the raw HTML? (no JS) 3 Interpret Can it understand & trust it? (schema, author)
Our method: each step depends on the one before it.

1. Access — can the AI bot even fetch the page?

AI crawlers are controlled by your robots.txt, and there's a subtle trap: search bots and training bots are different.

  • Search / grounding bots (OAI-SearchBot, PerplexityBot) feed live answers. Block them and you disappear from citations.
  • Training bots (GPTBot, ClaudeBot, CCBot, Google-Extended) feed model training. Blocking them is a legitimate choice and does not affect whether you're cited.

A common mistake: blocking Google-Extended thinking it removes you from AI Overviews. It doesn't — that only controls Gemini training; AI Overviews use the normal Googlebot. Our geo_ai_access check reads your robots.txt and tells you exactly which AI bots are blocked, and whether that costs you citations or not.

2. Read — is your content actually in the HTML?

This is the single most important — and most overlooked — check. AI crawlers do not run JavaScript. Vercel analysed ~500M GPTBot fetches: zero JavaScript execution. ChatGPT, Claude and Perplexity read the raw HTML. If your content only appears after JS runs, they see a blank page.

JS-only (client-rendered) AI sees a blank page SSR / SSG (server-rendered) AI reads your content
Server-render (SSR/SSG) your content. Google still renders JS — the other engines don't.

Our geo_js_content check flags pages whose content only exists after JavaScript. We measure this for free, because our own crawler reads the raw HTML — the exact thing a non-rendering AI bot sees.

What each engine reads today

Engine Runs JavaScript? What it reads
ChatGPT (OAI-SearchBot) No Raw HTML only
Claude No Raw HTML only
Perplexity No Raw HTML only
Google / Gemini (AI Overviews) Yes — via Googlebot Rendered page
Copilot (Bing) Yes Rendered page

So "AI ignores JavaScript" is true for ChatGPT/Claude/Perplexity and false for Google. We show the honest version, per engine.

3. Interpret — can it understand and trust the page?

Once an engine can read your text, two softer signals help it understand and cite you:

  • Structured data (schema). Evidence is mixed — Google says no special schema is needed, some studies find no citation uplift, others find correlation. We treat it as an opportunity, never a guarantee (geo_schema).
  • Freshness & citability. Pages without a clear author or date are harder for an AI to trust as a source. Our geo_freshness check flags content pages missing that provenance.

We also reuse classic signals — headings, title/meta, readable text, status 200 — because a page an AI can't parse is a page it won't quote.

Truth vs myth

The GEO space is full of confident claims. Here's how we separate signal from hype:

Claim Verdict
A JS-only site is invisible to ChatGPT/Claude/Perplexity True — they don't render JS
Blocking Google-Extended removes you from AI Overviews Myth — it only controls Gemini training
Schema markup guarantees AI citations Myth — mixed evidence; it aids understanding
llms.txt is an AI ranking factor Myth — no major engine honours it today
Being cited is a purely technical outcome Myth — it's mostly content and authority

Why we never promise citations

Here's the honest part most tools skip: the strongest correlation with AI visibility isn't technical. It's brand mentions across the web (Reddit, YouTube, Quora), answer-first writing, and being a credible source. Technical readiness gets you eligible to be read. Whether you're chosen depends on content and authority.

We measure the technical readiness precisely, and we're explicit that citation depends on things outside the technical layer. Honesty is the whole point — a tool that promises guaranteed citations is selling you hype.

How the score is weighted

A GEO problem doesn't hurt every engine equally. ChatGPT dominates AI search, so a page it can't read costs you far more than one Claude can't read. Our score weights each engine by its real share of the market, and counts only real problems — never "passed" checks.

ChatGPT 45% Gemini 20% Perplexity 18% Copilot 10% Claude 7%
Approximate share of AI search. Failing the biggest engines drops the score the most.

See where your site stands

We built this because almost nobody measures AI readiness honestly. The result is a GEO score you can trust — precise about the technical layer, and upfront about the rest.

Run a free audit and see, in seconds, whether ChatGPT, Claude, Perplexity, Gemini and Copilot can access, read and interpret your site.

Sources

  • Vercel — The rise of the AI crawler (~500M GPTBot fetches, zero JS execution): vercel.com/blog/the-rise-of-the-ai-crawler
  • OpenAI — bots documentation (GPTBot vs OAI-SearchBot vs ChatGPT-User): developers.openai.com/api/docs/bots
  • Cloudflare — From Googlebot to GPTBot: who's crawling your site in 2025: blog.cloudflare.com
  • Ahrefs — schema markup & AI citations study; What is llms.txt: ahrefs.com/blog
  • Google — AI Overviews / AI Mode structured-data guidance