How to check whatAI assistants sayabout your brand.

1. Write ten unbranded buyer questions. Write the questions a buyer types before they know any brand names: which is best for a company my size, which is cheapest, which handles a specific need, X versus Y, is it legit, where should I buy. Do not put your brand name in any of them yet.
2. Run each question on each engine, from India, in a fresh chat. Use the consumer apps logged in as an ordinary user, or the search-grounded APIs if you have them. One run per question per engine, a new chat every time so earlier answers do not colour later ones. Save the full answer text and any source links.
3. Log who is named, in what order, and which sources were read. For each answer, list every brand it names, note which one comes first, and record the domains it cites. Count a mention only when the name appears as a whole word, not as part of a URL or another word.
4. Count your share. Your mention rate is the number of answers that name you divided by the number of answers. Thirty answers is enough to see the shape; a single run is not, so treat the number as a range and re-run before drawing conclusions.
5. Separate branded from unbranded questions. Now add five questions that include your name, such as is it good, what does it cost, X versus a competitor. Score them separately. A brand can be named in every branded answer and almost no unbranded one, and the headline number would hide that.
6. Read the sources. List the domains cited across all the answers and count how often each recurs. Those recurring pages are what the engines are reading to decide. Check whether you appear on them, and whether what they say about you is right.
7. Repeat monthly, per engine. Answers vary between runs and the engines change their sources. Re-run the same questions every month, keep the results in one sheet, and read the trend per engine rather than the total.

Zimyo, an HRMS vendor, was named in 10 of 30 answers in our sweep. Every one of those came from the questions that already contained the word Zimyo. On the questions that did not, the ones a buyer asks before choosing, it was named in 2 of 22. The headline number overstated its real discovery presence more than threefold.

Score the two sets separately. The unbranded number is the one that tells you whether new buyers can find you at all.

In August 2026 we ran fifteen buyer questions for each of 20 Indian brands across 18 categories on ChatGPT and Perplexity, 600 answers in all. The median brand was named in 55% of its own category's answers. 9 of the 20 were named in fewer than half, and 7 in one answer in three or fewer.

The full table, brand by brand and engine by engine, is in the research report linked below. Use it to see where a brand of your size in a category like yours tends to land.

• A category with no winner. In Man Matters's category 17 of 30 answers named no tracked brand at all; in Earthful's, 17 of 30. An empty shortlist is a gap someone will fill.
• Your site cited, your name absent. The engine read you and recommended someone else. Check what the cited page actually says.
• A competitor's blog as the category's source. In HRMS, one vendor's blog was cited in nearly half of all answers to its competitors' buyer questions.
• Trust questions answered from a tiny review profile. Note which review site the engine quotes and how many reviews it has.
• A price you did not publish. If the engine quotes a dollar figure from a third-party listing, that is now your price.
• A refusal. Perplexity declined to name any seller for two premium pram brands because it saw no reliable India-specific results, while an Indian retailer that stocks both went unnamed.