Skip to main content
Free tool

AI visibility checker

Paste any address. We check whether ChatGPT, Claude, Perplexity and Google's AI answers can reach the page, whether they can read it, and whether the facts a customer asks for are on it.

The crawlers this checks, and what each one is for

They are not one thing. Some build the index an assistant answers from, some open a single page because a person asked, and some collect text for training. Refusing the first kind hides you; refusing the last kind does not.

Crawler Kind What it does
OAI-SearchBot OpenAI Assistant search Finds pages to show and link in ChatGPT search answers
Claude-SearchBot Anthropic Assistant search Indexes pages for search results in Claude
PerplexityBot Perplexity Assistant search Indexes pages to cite and link in Perplexity answers
Googlebot Google Search engine Google Search, which AI Overviews and AI Mode answer from
Bingbot Microsoft Search engine Bing Search, which Copilot answers from
ChatGPT-User OpenAI On request Opens a page when a ChatGPT user asks about it
Claude-User Anthropic On request Opens a page when a Claude user asks about it
Perplexity-User Perplexity On request Opens a page when a Perplexity user asks about it
GPTBot OpenAI Training Collects text that may be used to train OpenAI models
ClaudeBot Anthropic Training Collects text that may be used to train Anthropic models
Google-Extended Google Training A control, not a crawler: whether Google may use the site to train Gemini

What this cannot see

  • Whether an assistant mentions you. That depends on the question asked and on everything else the assistant has read. No tool outside those companies can measure it, and this one does not guess.
  • A block at your firewall or CDN. robots.txt is a request. A security service can refuse AI crawlers by name or by network before it is ever read, and some do so by default. We fetch the page as ourselves, never under a crawler's name, so that kind of block does not show here.
  • Facts on other pages. One page is checked. Hours that live on your contact page are not counted against this one being silent about them.
  • Anything a script draws. A booking widget, a map, or a menu loaded after the page opens is not in the HTML we read. It is not in the HTML a crawler reads either, which is why it is worth knowing.
  • What others say about you. Reviews, directories and local press are read too, and they are outside your site.

Passing everything here means the part you control is in order. It is not a ranking, and nobody should sell you one.

If something needs fixing

  1. A crawler is refused. Open your robots.txt (it is at yoursite.com/robots.txt) and look for the rule named in the table. If you did not write it, look for an "AI bots" or "block crawlers" setting in your site builder, security plugin or CDN.
  2. Hardly any text arrives. The page is being built in the browser. Ask whoever built it for server-side rendering, or put the essentials - who you are, where, when, how much - in the page itself.
  3. No structured description. The LocalBusiness schema generator writes the markup from a few answers. Paste it into the page and run this again.
  4. A fact was not found. Write it as text on the page a customer lands on: the phone number as digits, the hours by day, at least a starting price.
Want the rest of the picture? The website checker reads the same page for SEO, content, accessibility and page weight.

How this is worked out

Three requests are made: the page, the robots.txt of the host it answered from, and that host's llms.txt. The robots.txt is evaluated for each crawler by name as RFC 9309 specifies: a crawler obeys the group that names it and otherwise the wildcard group, the longest matching rule wins, and Allow wins a tie. A block on a search crawler or a search engine is reported as a fault, a block on an on-request fetcher as a warning, and a block on a training crawler as information only, because each vendor documents that training is separate from appearing in answers. The page is then read without running any script: the words left after scripts and styles are removed are counted, noindex and nosnippet are read from the page and from its X-Robots-Tag header, and the phone, hours, address, prices and a next step are looked for in the text and in schema.org JSON-LD. Those last five are pattern matches and are worded as "could not find". There is no score. The page is fetched as our own checker, never under a crawler's name, so a block applied by a firewall or CDN is not detected.

Standards and documentation it follows

Guides that use this

Quoting this tool

Free to quote and link. Cite it as AI visibility checker, a free tool by Webkio, https://webkio.com/ai-visibility-checker. The method above is the whole method, so a figure you quote can be checked by anyone with the same inputs.

Link to this tool

<a href="https://webkio.com/ai-visibility-checker">AI visibility checker</a> by Webkio

Share your result

A link that opens this tool with your numbers already filled in, so the person you send it to sees the same result.

Questions

Nobody can promise that, and this page does not. It checks the part that is yours to control: that the crawlers are let in, that the text is there to read, and that your facts are written down. Whether an assistant then mentions you depends on the question asked and on everything else it has read. A page that fails these checks cannot be quoted at all, which is why they come first.

Because a number here would read as your odds of being recommended, and nothing outside the assistant companies can measure that. You get each check, what was found, and what to change. A tool that sells you a visibility percentage is guessing.

That is your decision, and it is reported as information for that reason. GPTBot, ClaudeBot and Google-Extended collect text to train models. OpenAI, Anthropic and Google each document that refusing them is separate from appearing in answers, so you can refuse training and still be found. What does hide you is refusing the search crawlers, which a blanket rule often does by accident.

A proposed plain-text file that summarises a site for language models. It is not a standard and no major assistant has said it relies on one, so we report whether you have it and never mark its absence as a fault. It is cheap to add and harmless to skip.

No. This reads your robots.txt, which is a request, and fetches your page as itself. A firewall or CDN can refuse AI crawlers by name or by network before robots.txt is ever consulted, and some do so by default. If the checks here pass and you still suspect a block, look for a bot or AI crawler setting in your CDN or security plugin.

Because one page is what can be fetched and read honestly in a few seconds. Start with your homepage, then check the page that answers the question customers ask most: the menu, the price list, the booking page.

A published Webkio site is rendered on the server, so its text is in the HTML, and filling in your business details writes the structured data for you. Its robots.txt lets these crawlers in, and it serves an llms.txt written from the same details: your phone, address and hours, your services and prices, your menu or catalog. The facts are still yours to state: a site with no prices on it has no prices for an assistant either. Paste your address above and see.

More free tools

LocalBusiness schema generator
Website planner
Website checker
Title and description checker
Google review link generator
Seasonal rate calculator

See all free tools