How to track your brand's mentions in ChatGPT (and every AI engine)

You typed your brand into ChatGPT, read the answer, felt relieved or worried, and closed the tab. That is the spot-check, and almost everyone has done it. It feels like measurement. It is not.

Tracking your mentions in AI answers is a real discipline, and it looks nothing like a one-time lookup. Here is how to do it properly, across ChatGPT and every other engine your buyers use.

Why the spot-check fails

A single manual check breaks in four ways, all at once.

• Answers are non-deterministic. Ask ChatGPT the same question twice and you can get two different shortlists. One reading tells you nothing about what the engine usually says.
• There is no history. You saw today’s answer. You have no record of last week’s, so you cannot tell whether you are gaining ground or quietly losing it.
• It is one engine. ChatGPT is not Perplexity is not Gemini is not Google AI Mode. Your buyers spread across all of them, and each one names a different set of brands.
• There are no competitors in frame. “Were we mentioned?” is the wrong question. The one that matters is “who got named instead of us, and how often?” A spot-check rarely captures that.

Put together, the manual check gives you a feeling, not a number. And feelings do not go in a board deck.

What to actually measure

If you want a metric you can trust and report, measure five things for every prompt, every time.

1. Named or not. Did the engine mention your brand at all? This is the floor.
2. Position. First brand named, buried at the end, or part of a long list? Order signals how strongly the engine recommends you.
3. Sentiment. Being named as a leader is not the same as being named as the cheap option or the one with caveats.
4. Share of voice. How often you appear versus the competitors the AI names instead. This is the real scoreboard. We go deeper on it in share of voice in AI answers.
5. Cadence. None of the above means anything as a single data point. Measure on a schedule so you get a trend, not one lucky answer.

If you are new to the underlying idea, what is AEO covers why being the answer is now its own channel, separate from ranking on a page.

The step-by-step

Here is the loop, concretely.

1. Pick the prompts your buyers actually ask. Not your product name. The questions a buyer types before they have heard of you: “best X for Y,” “X alternatives,” “how do I solve Z.” Start with ten. These are the prompts where the decision gets made.

2. Run them across every engine, on a schedule. ChatGPT, Claude, Gemini, Perplexity, Copilot, and Google AI Mode. Run each prompt multiple times per engine so non-determinism averages out, and repeat on a fixed cadence — weekly at minimum, daily if the category moves fast.

3. Score every answer the same way. For each captured answer, mark named-or-not, position, sentiment, and which competitors appear. Consistency is the whole point. A score is only comparable across weeks if you judged every answer by the same rule.

4. Watch the competitors, not just yourself. Roll your scores into share of voice. The week a rival jumps from being named in two answers to being named in eight is the week you need to know about, whether or not your own number moved.

5. Alert on movement. A trend you check once a month is a trend you find out about too late. The useful signal is the change: a citation lost, a new competitor entering the shortlist, sentiment turning. Set the threshold and let the movement come to you.

How this differs from a one-time lookup

A lookup answers “what did the AI say just now.” Tracking answers “what does the AI usually say, how is that changing, and who is winning the answers I am losing.”

The difference is the same one that separates checking your Google rank by hand on a Tuesday from running rank tracking. One is a glance. The other is an instrument. The brands that treat AI visibility as an instrument will see the shift coming. The ones still spot-checking will read about it in a competitor’s case study.

This is exactly what Sonarvue automates: it runs your prompts across all six engines on your cadence, scores every answer for naming, position, sentiment, and share of voice, and alerts you the moment a citation or a rival moves. No tag, no install, and the first read finishes in minutes. But the discipline starts the moment you stop closing the tab and start keeping the record.