AI Visibility Checker
Find brand mentions and citation links in AI answers you have captured. Compare competitors, review the evidence and measure changes across matching samples.
Start with an answer
A mention and a citation tell you different things. An AI answer can name your business without linking to you. It can also cite your research without naming the business behind it. Keep both in view before deciding what needs work.
Add your brand name, any names customers genuinely use for it, and the domains you want counted as citations. Then paste an answer or import a batch of captured responses. The checker reads the supplied text and citation lists. It does not query ChatGPT, Google or another AI service, and it does not turn a website URL into a guessed visibility score.
Count the right thing
If your brand appears five times in one answer, that is one answer mentioning you. If it appears in two of three completed answers, the sample mention rate is 66.67%. It is not a measure of how often all AI users see your business.
Citation rate has its own denominator. Suppose you captured full citation lists for only two of those three answers, and one of those lists links to your domain. The citation rate is 50% of the two complete lists. The third answer stays incomplete. Treating it as a missing citation would make the result look more certain than the evidence allows.
Failed collections and searches that show no AI answer remain visible in the report, but neither enters the completed-answer denominator. A report with no completed answers has an unavailable mention rate, not a zero score.
Keep the context
Save the prompt exactly as you asked it. Record the product, model if shown, country, language, capture time and collection source. Use Unknown where a detail cannot be established. A product-interface answer and a provider-API response should not silently become the same observation.
For a batch, use the JSON or CSV template beside the import field. Each row represents one capture and needs its own ID. Citation URLs belong in a JSON list in the CSV citations column, such as ["https://example.com/research"]. An empty list is []; mark it complete only when you actually checked the full citation list. Keep private conversations out of uploads.
The report retains each supplied answer, its matched alias and its citation links. Open the answer evidence when a count looks surprising. A common-word brand name can refer to something else. A broad alias can match a different product. Literal matching helps make the result inspectable; it cannot resolve those meanings for you.
Compare like with like
Choose a repeatable set of questions that matters to your customers: shortlisting products, solving a problem, checking alternatives or evaluating a supplier. Keep brand-led questions separate from neutral discovery questions when drawing conclusions. Asking an AI to explain your brand is a different test from asking it to recommend options without naming you.
The platform breakdown keeps product, model, market, language and collection source together. The optional previous-period comparison is stricter: both periods need the same prompts, contexts, outcomes, citation coverage and repeat counts, and the older captures must precede the newer ones. Otherwise, the tool withholds a percentage-point change. You can still inspect both sets of answers.
Even matched captures are only observations. Generated answers vary, and small samples move easily. A change after an article update is a reason to investigate, not proof that the update caused it. This checker does not estimate audience size, search demand or statistical significance.
Turn findings into work
If a competitor appears where you do not, read the answer and its sources first. Is the question relevant to your business? Is your product information clear and current? Does the cited source offer a useful comparison, original research or an explanation that your pages lack? Make a specific editorial decision from that evidence.
When your domain is cited but the answer describes the business incorrectly, check the source page and the wording in the answer separately. A citation alone does not establish that a claim is accurate or supported. Supplied links are matched by hostname, including subdomains; redirects and destination availability are not checked here. Use the Redirect Checker for those links and the Indexability Checker for crawl and directive evidence.
Google’s guidance for AI features keeps the emphasis on established SEO and useful content, not a special score that guarantees inclusion. Ahrefs’ visibility metric definitions also distinguish mentions from citations. Here, every percentage is limited to the captures you supplied.
Common questions
Does this check live AI answers?
No. It analyzes answers you paste or import. Collect responses from the actual AI product or a licensed provider, then keep their source and capture time with the report.
Can I compare competitors?
Yes. Add each competitor’s name, relevant aliases and citation domains. All brands are checked against the same answers. This is a comparison within your sample, not a market-share estimate.
Why is citation data unavailable?
The citation list may be partial or uncaptured, or you may not have entered a domain for that brand. Only complete lists with configured domains enter the citation-rate calculation. A known link in a partial list is still shown as observed evidence.
Are my prompts saved?
This plugin does not store prompts, answers or generated reports. It processes the submission and returns the results. Downloaded JSON includes the full supplied answers, so treat exported reports as private when appropriate.