AI Visibility Checker
Measure what makes a page liftable by an answer engine — extractability, structured data, llms.txt — and which AI crawlers you block.
Paste any public URL. This reads the page and its robots.txt, then reports what an answer engine would find. Nothing is stored.
What this checks, and what it cannot
This tool fetches a URL and its robots.txt, then reports what an answer engine would find. Every number describes your page. None of it predicts whether any engine will cite you — that depends on the prompt, the competing sources, and a ranking process nobody outside those companies can observe.
What it can tell you is why you might be skipped, which is a different and far more actionable question.
Reachability comes first
Before anything about content quality matters, a crawler has to be allowed in. The most common self-defeating configuration is blocking a retrieval crawler while meaning to block a training one:
- Training crawlers —
GPTBot,ClaudeBot,Google-Extended. These permissions concern collection for training; Google-Extended also concerns Gemini grounding and is a permission token rather than a crawler. Review each policy separately. - Retrieval crawlers —
OAI-SearchBot,Claude-SearchBot,PerplexityBot. Blocking these can restrict search crawling and discovery. Allowing them does not promise a visit or a citation.
They are separate agents with similar names, and a rule written for one frequently catches the other.
Extractability: can a passage be lifted?
Answer engines build responses out of passages they can quote without the surrounding context. A page can contain exactly the right answer and never be used, because the answer only makes sense after three paragraphs of throat-clearing.
The strongest structural change most pages can make is an opening paragraph that directly answers the question in the title, short enough to stand alone. Question-shaped headings help for the same reason — they match how people actually phrase prompts.
Attribution and specifics
The published research here is reasonably clear. Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024), tested a range of interventions and found that citing sources, adding quotations and adding statistics produced the largest visibility gains among those tested. This checker counts external domains cited and quotable figures for that reason.
An unattributed claim on an unknown domain is the weakest possible evidence. The same claim with a primary source and a specific number attached is something an engine can repeat with confidence.
Fixing what it finds
Most of what this reports is fixable by editing: shorten the opening, phrase a heading as the question, cite the source you already used, add the number you already know. Crawler rules and llms.txt are configuration rather than writing, and are usually a one-off.
Frequently asked questions
- Can this tell me if ChatGPT will cite my site?
- No, and no tool can. Whether an engine cites a page depends on the prompt, the competing sources and a ranking process none of us can observe. This checker reports crawl directives and content signals you can review: opening length, headings, attribution, structured data and an optional site summary. None is a prerequisite or guarantee of a citation.
- What is the difference between GPTBot and OAI-SearchBot?
- GPTBot collects content for model training. OAI-SearchBot fetches pages to answer live questions in ChatGPT. Blocking GPTBot is a reasonable editorial choice about training. Blocking OAI-SearchBot can restrict search crawling and discovery in ChatGPT. It does not remove every existing reference or user-requested fetch. Review retrieval permissions separately from training permissions.
- What is llms.txt?
- A plain-text file at the root of a site that lists what the site covers, in a form a language model can read cheaply — a title, a short summary and a set of links with descriptions. It is a convention rather than a standard, and no engine is obliged to read it, but it costs almost nothing to publish and gives a model a map of the site that does not require crawling every page.
- Why does the length of my first paragraph matter?
- Answer engines assemble responses from passages they can lift whole. A page that opens with a long scene-setting introduction may contain the answer three paragraphs down and still never be quoted, because there is no self-contained passage to take. A short opening that directly answers the question implied by the title is the single most effective structural change most pages can make.
- Does the score mean anything on its own?
- Treat it as a way to rank your own pages against each other, not as an absolute measurement. It weights the signals with the strongest published support — attribution, quotable statistics and extractability — but the weights are a judgement call, not a published formula.
Want the audit across your WordPress pages? Crawlune reports content weaknesses and crawler policies locally. The audit and controls need no account or license.