How to tell if AI is citing your brand (and what to do about it)

Why AI citations are invisible in your analytics

A citation inside a generated answer usually produces no click, so it leaves no trace in the tools you already watch. Google Analytics records visits, and Search Console records impressions and clicks in the classic results, but neither reports that Perplexity summarized your page or that an AI Overview named you as a source. The visibility is real and the measurement gap is total, which is why brands routinely underestimate how present, or absent, they are in AI answers.

This blind spot is dangerous because AI answers shape opinions upstream of the click. A buyer who reads a comparison in ChatGPT forms a shortlist before they ever reach a website. If you are not measuring citations, you are flying blind on the surface where more and more decisions are actually made.

The manual way to check right now

You can audit your AI visibility today with nothing but the answer engines themselves. Build a fixed list of the questions your buyers actually ask, the ones where you would want to be named, and run each one through ChatGPT with search, Perplexity, and a Google query that triggers an AI Overview. For each result, record one thing: are you cited, is a competitor cited, or is neither named.

Keep the prompts stable and neutral so the test is repeatable. Ask the question a customer would ask, not a leading prompt with your brand name in it, because the honest signal is whether the model reaches for you unprompted. Run the same list again in a few weeks and you have a trend, not just a snapshot.

• Write 10 to 20 real buyer questions in your category, free of your brand name.
• Run each through ChatGPT with search, Perplexity, and Google AI Overviews.
• Log whether you, a competitor, or nobody is cited for each query.
• Note which competing pages get named so you can study what they do right.
• Repeat on a schedule so you measure movement, not a single moment.

Reading the results without fooling yourself

Answer engines are non-deterministic, so a single run is noise, not signal. The same prompt can name different sources on different days, which means one disappointing answer does not mean you are invisible and one flattering answer does not mean you have won. Look at the pattern across a batch of queries and repeated runs, and treat the rate at which you appear as the metric rather than any single response.

Watch who wins when you lose, too. If the same two competitors keep getting cited for your core questions, their pages are the benchmark. The gap between their cited passages and yours, in clarity, structure, and directness, is your specific to-do list, not a vague instruction to make better content.

What to do when a competitor is cited instead

When the model names someone else, the fix is almost always structural rather than a matter of writing more. Open the page you want cited and make the target question resolvable in a single clean passage near the top: state the direct answer first, use the question as a heading, tighten the sentence so it stands alone, and name the specific entities involved so the model can connect it to what it already trusts.

Then reinforce the signals around it. Confirm the page ranks at all, since answer engines lean on pages that already surface, add FAQ and Article schema, and make your brand a clear entity so a citation has a name to attach to. Change the page, rerun your prompt list, and watch whether your appearance rate climbs. That loop, measure, fix, remeasure, is the whole discipline.