How do I check if AI search mentions my brand?

If you ask ChatGPT a question today and it names you, then ask again tomorrow and it does not, nothing broke. Models sample their answers with some randomness, and responses can be shaped by per-user memory and context, so two runs of the same prompt routinely differ in wording and sometimes in which brands they list. A single answer is one sample from that distribution. Treating it as a verdict - "we're in ChatGPT" or "ChatGPT ignores us" - is the most common mistake people make when they first look. The same trap is why a competitor's screenshot proves nothing about their standing either. For the full picture of why direct control is impossible and only the inputs move, see whether you can control what ChatGPT says about you.

You can run a credible check for free. It costs time and discipline, not money.

• Write the buyer's questions, not yours. List the prompts a prospect would actually type when they have your problem but do not know you yet - "best tools for X," "how do I solve Y," "alternatives to Z." Do not ask the engine about your brand by name; that tests recognition, not whether you surface on your own.
• Ask every engine that matters. Run the same prompts on ChatGPT, Claude, Gemini, and Perplexity, because they read different sources and disagree often. Being named on one is not being named on all.
• Repeat each prompt several times. Three runs per prompt per engine is a sane floor. You are sampling a distribution, so you need more than one pull to see the pattern.
• Log two things: named and linked. For each answer, note whether you were mentioned in the prose and, separately, whether you were attached as a cited source. Tally it. The output is a simple rate - the share of answers that named you, and the share that cited you.

Where you searched with the buyer's intent and still did not appear, you have found an omission, which is a citation-footprint gap rather than a mistake to correct. That gap is the subject of AI search visibility.

Keep these two columns apart, because they tell you different things. A mention is the model naming you in its answer; it shows the engine associates your brand with the topic, even if it did not link anywhere. A citation is the engine attaching one of your pages as a source it drew from, which both signals that your page was liftable and can send you a real click. You can be mentioned without being cited, and occasionally cited as a source without being named in the prose. Tracking only one of them hides half the picture.

A one-time check tells you where you stand today; a repeated one tells you whether anything you change is working. Once you have a share - say you are named in a third of answers for your core query - the useful move is to watch that number across the same prompts and engines over weeks, not to re-judge it from a fresh screenshot each time. That cadence, and the share-of-voice framing that makes it comparable, is the whole subject of measuring AI citations. It is also why the real cost of checking is repetition: the signal lives in the trend, and the trend needs the same test run again and again.

It is worth knowing how low that share can start. In our Invisible 10 study, ten funded vendors with crawlable sites were named in zero of 600 answers across the four largest engines - a reminder that "we checked once and weren't there" is a common, fixable starting point, not a dead end.

The manual method works, and the free 10-minute AI search audit structures it so you are not inventing the prompts or the scoring yourself. When you want a read without doing the runs by hand, the Free Snapshot checks where you stand today. For the repetition that turns a check into a measurement, the SXO Audit runs 25 buyer prompts across six engines with three trials each and repeats over time, so you watch the share of answers that name and cite you move rather than guess from one look.

Try the Free Snapshot   See the SXO Audit

By the Web Cited Editorial Research Team. Last updated 1 June 2026.