AI Search Visibility Checker

What this AI search visibility checker looks at

Paste a URL and three things are fetched server-side: the page itself, its site's robots.txt and its /llms.txt. From those it runs about twenty checks in four groups. Access: can the AI search crawlers, the user-triggered fetchers and Googlebot reach this exact path, and does a noindex, nosnippet or max-snippet:0 directive keep it out of answers? Understanding: a title, a meta description, one H1, real section headings, enough readable body text, a declared language and a canonical. Attribution: JSON-LD and which types it declares, an Organization or Person entity, an author, and published and modified dates. Site-level: an llms.txt file and a sitemap declared in robots.txt. Each check shows the evidence it found and, where something is wrong, the fix.

Why the score is a heuristic, and says so

No answer engine publishes its ranking inputs, and Google has stated that no special markup or file is needed for a page to appear in AI Overviews or AI Mode: the same Search index and the same quality signals apply. So this tool does not pretend to measure how likely you are to be cited. It measures the documented prerequisites (crawler access, snippet controls, indexability) that can rule a page out entirely, plus the structural signals every engine's own guidance names, and adds them up. A blocking failure costs more than a missing nicety, optional extras such as llms.txt are never scored against you, and a score of 100 means nothing is in the way, not that citations are guaranteed.

How to measure AI visibility

Three kinds of AI crawler, and why they are scored differently

Blocking "AI bots" as one group is the mistake this checker is built to catch. AI search crawlers such as OAI-SearchBot, Claude-SearchBot and PerplexityBot build the indexes their assistants cite from; disallowing them removes this page from those answers, so that is a blocking failure. User-triggered fetchers such as ChatGPT-User and Claude-User only open a page when someone pastes its URL into an assistant; blocking them makes the page unusable inside assistants but does not affect search, so that is a warning. Training crawlers such as GPTBot, ClaudeBot and Google-Extended collect data for model building; blocking them opts you out of training without touching AI search or Google Search, so the tool reports the choice and does not score it either way.

AI Crawler Access Checker (per-bot rules) · The full AI crawler user-agent list

Why Googlebot gets its own row

Google's AI Overviews and AI Mode are not fed by a separate AI crawler. They draw on the ordinary Search index, so a page Googlebot cannot crawl is out of them too, whatever the AI-specific tokens say. The row is there because the most expensive robots.txt mistake is a rule written for AI crawlers, or a staging Disallow promoted to production, that catches Google Search as well. Google-Extended, the token that controls training use, is listed with the training crawlers, where it belongs.

The snippet controls that quietly remove a page from AI answers

Google documents nosnippet, max-snippet and data-nosnippet as the controls that decide whether a page's content can appear in AI Overviews and AI Mode, and the same directives are respected by other engines. A site that once added nosnippet to keep a paragraph out of a search result has, since, also kept the whole page out of every AI answer. The checker reads the meta robots tag, the googlebot-specific variant and the X-Robots-Tag header, and flags noindex, nosnippet and max-snippet:0 as blocking. If only one passage is sensitive, the fix is data-nosnippet on that element, not a page-wide rule.

Build and test a robots.txt

Structure an engine can quote from

An answer engine does not cite a page, it cites a passage. It finds that passage by heading, so the structure checks reward what makes a passage liftable: one H1 that states the topic, H2 headings that read like the question a person would type, the answer in the first sentence beneath each, and a body worth quoting served in the HTML rather than assembled by JavaScript the crawler will not run. The tool counts how many of your H2 and H3 headings are phrased as questions because that is the cheapest rewrite with the most direct effect on which passage gets chosen.

Entity, author and date: the attribution signals

Engines prefer sources they can name and date. The checker looks for JSON-LD and lists the types it finds, then asks three narrower questions: is there an Organization, Person or WebSite entity that says who is behind the page; is there an author in the markup or the meta tags; and are there published and modified dates in the JSON-LD, the article meta tags or a time element. A missing author or date is scored as something to improve rather than a blocker, because plenty of pages are cited without them, but a page that carries all three is easier to trust and easier to keep fresh.

Schema Markup Generator · Structured data checker (whole-site)

llms.txt: reported, not required

An llms.txt file is a short markdown list of the pages you most want assistants to read. The checker fetches /llms.txt and reports whether it exists, whether it is really a text file or an HTML page answering 200 for a path that does not exist, or whether it is missing. It is scored as a bonus when present and never as a defect when absent: no major engine has committed to reading it, and Google has said it is not needed. It is cheap to add and harmless, which is a different thing from being required.

Generate an llms.txt · What llms.txt is

One page here, every page there

A page-level check tells you about the page you already suspected. The costly problems are the ones spread across a template: a robots.txt rule that blocks an AI search crawler on every blog post, a theme that emits two H1s on every product page, an author block that never made it into the JSON-LD. Finding those means crawling the whole site. Crawl Cove's AI search visibility audit runs its AI visibility pack on every page in one desktop crawl and groups the findings by cause, alongside its full technical SEO audit. The app's crawler-access check covers GPTBot, ClaudeBot, PerplexityBot and Google-Extended; the per-token breakdown and snippet controls on this page are, for now, wider than the app's.

AI search visibility audit for your whole site · Full site audit