Free AI visibility checker

Ask ChatGPT and Claude the questions your buyers ask, and see whether they name you, cite you quietly, or name somebody else. Live, with web search on, against real sources.

An AI visibility check asks an assistant a real buying question and records whether your brand survives into the answer. This one runs ChatGPT and Claude live with web search, takes up to five prompts, and shows the sources each answer actually drew on, so an absence comes with the evidence of what was read instead of you.

Reading your result

There are four things this check can come back with. They have different causes and different fixes, and the difference between the middle two is where most of the money in this category gets wasted.

Named in the answer

The assistant wrote your brand into the text of its reply, not just the source list. This is the outcome the whole category is selling, and if you already have it on your buying prompts, you have less of a problem than a vendor will tell you.

What to do: Widen the prompt set before spending anything. Being named on the two questions you thought of is a smaller result than being named on the twenty a buyer actually asks. Add the comparison and alternative phrasings, which is where competitors usually appear instead of you.

In the sources, not in the answer

The engine fetched and used one of your pages, then wrote its summary without naming you. Retrieval worked. What failed is the passage: nothing on the page was quotable enough to survive the summarising step.

What to do: This is an editing problem and the highest return one on this list, because you are already being read. Find the page that was cited and make each section answer one question in a self contained passage, with the figure next to the thing it describes.

Absent, and competitors are named

The engine had something to say on the question and drew it from somebody else. The sources list under each result tells you where it went instead, and that list is the most useful thing this check produces.

What to do: Read the sources. If they are third party pages, comparison articles, directories, review sites, then this is a corroboration problem and no work on your own site fixes it. If they are competitors' own pages, it is a retrieval or content problem on yours.

Absent, and so is everybody

The answer was generic and cited little. Usually the prompt is the problem rather than your site: a question broad enough that the model answers from what it already knows does not retrieve anything from anybody.

What to do: Rewrite the prompt as a buyer would type it, with the qualifier that makes it specific. "Analytics tool" retrieves nothing. "Analytics tool for B2B SaaS without cookies" retrieves the pages that answer it.

What this check does not tell you

Stated here rather than in a footnote, because every limit below is one somebody else in this category quietly leaves out.

One run is one sample

Generative answers vary between asks. The same prompt can name you at ten in the morning and not at noon, with nothing having changed on your site. A single result is evidence, not a measurement, and anybody presenting a screenshot as a score is showing you one roll of a die.

Two engines, not four

This runs ChatGPT and Claude. Gemini draws on Google's index and tends to track your classic rankings, and Perplexity leans hardest on third party sources. A brand can be fine in two of the four and absent from the others, which is exactly the finding a two engine check cannot give you.

No history, so no direction

A result today cannot tell you whether you are getting better or worse, and direction is the thing worth knowing. That is what monitoring is for, and it is the honest reason paid tools in this category exist.

It measures the outcome, not the cause

Being absent is a symptom. Whether the cause is crawler access, a page that renders empty, a passage nothing can lift, or an absence of third party corroboration is a separate question, and the checks below answer the mechanical part of it for free.

If you are absent, check these before changing any copy

The mechanical causes are the common ones, they are free to check, and they invalidate everything downstream while they are unfixed.

When we crawled 1048 of the most visited websites in August 2026, 28.3% blocked at least one AI crawler while 1.5% blocked Googlebot. The bots that decide whether you can be cited at all are the search and citation ones, PerplexityBot, Claude-SearchBot, OAI-SearchBot, and they are blocked far more often than anybody intends, usually by a robots rule copied from somewhere years ago.

The second common cause is rendering. Of the sites in that crawl whose pages depend on client side rendering, 27.8% served almost no text to a fetch that ran no JavaScript. An assistant reading that page received an empty shell, which is a total failure that looks like a content problem from the outside.

AI crawler checkerEach bot listed separately, allowed or refused Render gap checkerWhat a crawler that runs no JavaScript receives Indexability checkerWhether anything quietly says skip this page llms.txt checkerReads the body rather than the status code Structured data checkerWhether your markup matches the visible page Full site auditAll 36 checks, scored, across your site

Questions people ask about checking AI visibility

What is an AI visibility checker?

A tool that puts a question to an AI assistant the way a buyer would, then reports whether your brand appears in the answer, whether it appears only in the sources, and which competitors were named instead. This one asks ChatGPT and Claude live with web search enabled, so the result reflects what those assistants return today rather than a cached score.

Is this AI visibility checker free?

Yes, and there is no account, card or trial on this page. You get up to five prompts per run across both assistants. What is not free is monitoring over time, which is what paid tools in this category actually sell, and you should establish that you have a problem before paying anybody for that.

How do I check if ChatGPT mentions my brand?

Enter your brand, your domain and the questions a buyer would actually ask, then read the per platform result. Ask it as a customer would, not as you would: nobody types your company name into ChatGPT and then wonders whether they will see it. The useful prompts are the ones where you are one of several possible answers.

Why does my brand appear in one assistant and not the other?

Because they reach the web differently. ChatGPT searches through Bing, so a site that ranks respectably on Google and thinly on Bing can be strong in classic search and close to invisible in the most used assistant there is. Claude fetches live and cites explicitly. A split result is a real finding and it points at a specific cause rather than a general content problem.

The same prompt gives me a different answer each time. Is the check broken?

No, that is the feature behaving normally, and it is the single most important thing to understand about this whole category. Generative answers vary between runs. One check is one sample, which is why a screenshot proves nothing and why serious measurement is a rate across a fixed prompt set rerun on a cadence.

My brand is not mentioned anywhere. What do I do first?

Check whether the engines can read you at all, before changing a word of copy. In our August 2026 crawl of 1048 of the most visited sites, 28.3% blocked at least one AI crawler, mostly by accident. The AI crawler checker and the render gap checker cover that in about a minute and they are free.

Does this check Gemini and Perplexity too?

Not on this page. This runs ChatGPT and Claude, which is enough to tell you whether you have a problem and enough to show the split between two engines that reach the web differently. Tracking four engines on a schedule with history is the paid product, and it is worth buying only once a free check like this one has shown you something.

An absence has a cause, and it is usually mechanical

The free audit runs 36 checks on your live pages, including per bot crawler access and what a crawler without JavaScript receives. It is the half of this that can be measured rather than sampled.

Run the free audit How this is measured properly