AI search
AI Citation Checker
See which sources an AI answer cites
Enter a prompt and get the answer an AI engine returned, every domain and page it cited in citation order, whether yours is among them, and the engine and timestamp the sample came from so the check can be repeated.
Get early accessWhat it returns
- The answer text the engine produced for your prompt on this run, in full, so you can read the claim rather than a summary of it.
- Every cited source in the order the answer referenced it, with the domain and the specific page, and a flag on your own domain if it appears.
- Which domains were cited more than once in a single answer, because repeated citation and a single passing link are not the same result.
- The engine, the prompt as sent, and the time the sample was taken, labelled as one sample and not a standing fact.
When this matters
A prospect says they found three of your competitors by asking an assistant which tool to use for a job you do, and you were not one of them. You run the same prompt, you are cited, and the obvious conclusion is that they were wrong. They were not: the answer differs run to run, and both results happened. The useful question is not whether you are ever cited but how often, which is a question about repeated sampling rather than about one check.
How to answer this today
AI visibility is built around the non-determinism this tool can only warn you about. It runs your prompts against several answer engines on a schedule, samples each prompt multiple times per run so a citation rate is measurable instead of a single lucky hit, and keeps every cited domain next to your Google rankings so the two move in the same view. A free one-off check tells you what one run said, which is the right size of answer for a question you only have once.
AI visibilityWhy this is harder than it looks
One run is a sample
AI answers are not deterministic. The same prompt, sent twice a minute apart, can cite different domains, reorder the ones it keeps, and drop a source entirely. A single check is evidence that an answer of this shape is possible, never evidence that it is the answer.
Citations are not links
Engines differ in what counts as a source: some footnote every claim, some cite a handful of pages for a whole answer, and some name a brand in prose with nothing to click. Counting only the clickable footnotes understates the mentions that actually influence a reader.
The prompt is the variable
Rewording a question changes the retrieval behind the answer, so two prompts that mean the same thing to a person can produce two unrelated source lists. There is no single correct prompt for a topic, which means a set of prompts is the unit of measurement.
Each check is a paid query
Running a prompt against a commercial engine costs money per call, so this tool is metered and rate limited per IP. The limit exists to stop one script from draining the budget that keeps the tool free for everyone making a handful of checks.