AI visibility

Buyers ask AI first. Do you know what it says about you?

We put the questions your buyers actually ask to four answer engines, every week, and count which brands come back. Then we show you the sources those engines read, so there is something to do about it.

Check your category

From a real run, 2026-08-24

Loop is recommended in 89% of Perplexity answers — and 0% of ChatGPT.

Same category. Same 18 questions. Same week. 89 percentage points apart.

The thing nobody tells you: the engines disagree

Most brands check one engine, once, and treat the result as the truth. Run the same questions across four and the answers pull apart. A brand can own one engine and be invisible on another, which means a single number is worse than no number: it is confidently wrong.

Share of AI answers, by engine

Brand	Google AI Overview	Perplexity	ChatGPT	Gemini	Spread
Loop	86%	89%	0%	83%	89 pp
Alpine	29%	17%	0%	44%	44 pp
Eargasm	21%	39%	0%	22%	39 pp
Mack's	36%	0%	0%	28%	36 pp
Happy Ears	21%	6%	0%	22%	22 pp
Etymotic	7%	11%	0%	28%	28 pp
EarPeace	7%	22%	0%	11%	22 pp
Flare	7%	11%	0%	17%	17 pp
Calmer	7%	11%	0%	17%	17 pp
Vibes	0%	17%	0%	11%	17 pp

Answers returned per engine: Google AI Overview 14, Perplexity 18, ChatGPT 0, Gemini 18. Percentages are of the answers that engine actually gave. Category: reusable earplugs, a public consumer market used here as the worked example.

It also tells you why

Knowing you are absent is half an answer. The engines cite their sources, so we count them. In this category one domain carries 11% of every citation.

Source the engines cite	Citations	Questions it appears on
reddit.com	65	17
amazon.com	48	14
hearadvisor.com	19	6
hearingtracker.com	18	6
nytimes.com	17	10

This is the actionable half: you cannot argue with an engine, but you can be present on the pages it reads.

How it works

buyer questions, drawn from how people actually ask

engines, asked the same questions

Weekly

so you see movement, not a snapshot

Sources

every citation counted, so you know where to show up

What this does not do

Being named in an answer is presence, not endorsement. A brand can be named as the expensive option, and our sentiment read is a marker-word heuristic, not a reading of the answer.

Answer engines are not deterministic. Ask twice and you can get two phrasings. We track the pattern across many questions and weeks, which is why one run is a sample and a quarter is evidence.

We cannot make an engine recommend you. We can show you which sources it reads, which is the only lever that has ever worked.

What the alternatives cost

The tools in this category price per tracked brand and per prompt, and most bill annually up front. Verify the current numbers yourself before you compare; pricing in a young category moves fast.

See your category first

Questions we get

Which engines do you actually ask?

Google AI Overview, Perplexity, ChatGPT and Gemini, every week, on the same set of buyer questions. We keep them separate because they disagree: on one brand we measured an 11 percentage point spread between their best and worst engine. Averaging that away hides the finding.

Where do the questions come from?

Your catalogue and your buyers. We read the product types your store uses and the phrases your reviewers write, so a query like “best earplugs for sleeping” exists because 47 reviews said it, not because we guessed. You approve the list before anything is measured.

Can you prove the answer said what you claim?

Yes. Every answer is stored verbatim with the sources the engine cited, and the report shows the raw prompt next to the raw answer. If we say a brand was recommended, you can read the sentence.

How they build their ads What their customers complain about Share of AI answers, worked example Where AI gets its answers