We check how AI platforms answer your buyers' prompts and give you clear steps to fix gaps and grow.
Named
The share of answers where an assistant says your name. One credit per answer — a brand named four times in one answer is one.
Cited, not named
Answers that used your site as a source and recommended somebody else. Counted separately, because the two fail apart and need different fixes.
Position
Where you land when named — first pick, further down the list, or added after the recommendation was already made.
Per engine
Never blended. On one real run a brand sat at 31% on Google AI Overviews and 0% on ChatGPT, same questions, same day.
Every panel ends in an answer you can read. Figures below are from one real run of 58 answers.
The same questions run against every assistant and scored apart. Blend them and a brand that is invisible on ChatGPT but fine on AI Overviews reads as an average describing neither — and those two need different fixes. On this run: AI Overviews 4 of 13, Perplexity 3 of 15, Gemini 2 of 15, ChatGPT 0 of 15.
Being mentioned and being picked are different outcomes and one score merges them. Position separates the answers that put you first from the ones that add you after the pick was made. On this run: three first, four further down, two after the recommendation.
Click any point and you get the answer that produced it — the question, the full text, which engine, which run, and who was recommended instead. Around 4,400 characters, nothing summarised away.
Share of voice against the rivals the answers actually named, not a list typed at setup. Here Shiprocket took 32 of 58 answers and Delhivery 30, against the domain's own 9 — which tells you what you are displacing rather than only how you scored.
One domain, four assistants, five locked questions asked three times each — sixty answers, all readable, and a score per engine rather than one blended number.
Check AI VisibilityWhat teams ask before they trust a visibility score.
Ask the question your buyer would type, and ask it more than once. Then count how many answers include your name, not how many times your name appears. Two runs of the same prompt give the same brand list less than 1 time in 100, so a single try tells you very little.
There is no standard benchmark, so be wary of anyone who quotes one. Scores also differ between tools: one writer checked her site on four tools in the same week and got anything from 0% to 41%. The useful comparison is your own score over time, on the same set of questions.
Each AI platform finds and picks its sources in its own way. In one real run, a brand appeared in 4 of 13 Google AI Overview answers but 0 of 15 ChatGPT answers, for the same questions on the same day. Averaging the two would have hidden exactly the gap you need to fix.
Yes, more than most tools admit. In a study of nearly 3,000 repeated prompts, two runs of the same question rarely returned the same brands, and a 2026 analysis found that changes under about 5–7 points are just noise. That is why we ask each question three times and keep every run's result.
No. A keyword leads to a ranked list of pages, while a question to AI gets one answer that names a few brands and cites a few sources. You can rank first on Google and still be missing from the AI answer built on top of it, so we count "cited" and "named" separately.
On the last run we tested 49 questions and 26 passed our checks. We lock the 5 most important and ask each one three times on four AI platforms, which gives 60 answers per reading. We chose depth over a long list on purpose.
Yes, every one. Each answer is saved in full and opens next to the number it produced, and you can share it with anyone by link. The number is the summary; the answer is the proof.
We test buyer queries across top AI engines to measure and track your company's presence.
Prompt Discovery
