How we ran the test
In short. Same 20 questions, same way, three tools, three runs each.
We fixed the list of 20 questions on 29 October 2026 and did not change it. They cover national, city, industry, cost and shortlist questions, and three questions about PromotEdge by name.
Each question was asked three times on ChatGPT, Gemini and Perplexity. ChatGPT and Perplexity were first asked on 2 November and again on 6 November. Gemini was asked on 6 November. We used a logged-out browser, with no account signed in.
For every answer we wrote down three things: was PromotEdge named in the text, was a promotedge.com page linked, and which sites the tool pointed to. That gave 180 recorded answers.
We did not use Copilot, because it asked for a sign-in. Google AI Mode answered only twice in our log. So we make no claim about either.
What the three tools did
In short. Perplexity named us most often. All three often linked a promotedge.com page when they named us.
- ChatGPT named us in 17 of 60 answers and linked a promotedge.com page in 15.
- Gemini named us in 23 of 60 and linked a page in 16.
- Perplexity named us in 30 of 60 and linked a page in 20.
Add the three questions about our name, and every tool named us every time. Take those 27 answers out and we were named in 43 of the other 153 answers.
Where we were named most
In short. The strongest results came on topics where our site has a page written for that exact topic.
On the question about marketing for building materials brands, we were named in all 9 answers. We have an industry page for building materials that speaks to that buyer. On the question about a Kolkata agency that does ATL (TV, print and hoardings), BTL (shop and on-ground work) and digital together, we were named in 7 of 9. See our Kolkata page.
This is what we saw in this test. It is not proof of how the tools choose. But it fits the plain rule that a clear page for an exact question gives a tool something to quote.
Where we were not named at all
In short. Seven questions, 63 answers, no mention.
We were not named on the question about an integrated agency for all of India, or on the one about a full-service agency for B2B (business-to-business) brands. Nor were we named on the industrial B2B question or the D2C (direct-to-consumer) agency question. The same was true of the how-to on selling through Blinkit and Zepto, the monthly cost question, and the shortlist for a yearly retainer.
On these, the tools gave lists of large, well-known national agencies, or general steps. We do not know exactly why our name did not come up. We would rather show the zeros than hide them.
Where the tools got their answers
In short. Directory and list sites appear often, alongside company sites and trade news.
In the answers where the tools showed their sources, directory and list sites were common, next to agencies' own websites, LinkedIn and trade news sites. A promotedge.com page was one of the sources in 51 of the 180 answers. So the tools still send readers to web pages. The old work of a clear page still counts.
What we are doing next
In short. Three plain steps. We will report the next run in December.
- Put a short, standalone answer under each question heading on our pages, so a tool has a clean passage to quote. We began this on 8 November 2026.
- Add one fact to each page that the top results do not have, with its source and date.
- Run the same 20 questions again in December 2026 and show the numbers and the date, zeros included.
What this means for LLM SEO
In short. Treat it as normal SEO with one extra check: ask the tools and write down what they say.
If you want your business named, write a clear page for each question your buyers ask. Give facts and sources. Then ask the tools the same questions every month and note the dates. Our AI search optimisation guide has the steps, and the GEO glossary entry explains GEO, which is generative engine optimisation, and the other words.
A word of care
In short. This is a small test. Read it as a first reading, not a rule.
We asked 20 questions, 180 times in all, on two days. Answers change from day to day, from one tool to another, and from one person to another. We tested only questions about agencies, and only on our own name. Being named is not the same as getting an enquiry. We measure leads from our own enquiry record, not from this test.
We make no claim about how the tools work inside.
