Original research
We measured what AI actually cites in India
Almost every published study of AI search is global, blended across engines, and run by a tool vendor. We wanted to know what happens on Google, in India, for the queries Indian shoppers really type. Nobody had published that, so we went and did it.
The study
100 queries. Five categories. Three draws each.
In August 2026 we ran 100 high-volume Indian D2C shopping queries through Google's AI Overview, across skincare, supplements, hair care, home and kitchen, and jewellery. Every query was run three separate times, spaced out across the day, and every citation was coded by hand.
The query list was built from real search volume data, frozen before collection started, and published along with a log of every keyword we dropped and why.
We also wrote down three possible headlines before collecting any data, and published the one the data actually supported. One of the three died on contact with the first category.
That is an uncomfortable thing to commit to in public. It is also the only reason the numbers mean anything. If we would only have published an exciting result, it would have been marketing rather than research.
What we found
Three things worth knowing
24%
of citations appeared on all three draws. Ask Google the same question three times and three quarters of what it cites changes.
34%
of citations point at brands' own websites, the largest single bucket. Your own site is still the thing most likely to be quoted.
32%
of citations come from pages ranking below organic position 10. Being quoted is not the same as ranking, and the gap is bigger than most people assume.
Why an agency published this
Because we could not find the answer anywhere
Clients kept asking whether AI search was going to take their traffic. The honest answer was that nobody had measured it for India, and the global numbers being quoted were blended across engines that do not even accept an India location parameter.
So we built the instrument into Dorqa, our own platform, and ran the study ourselves.
It is also a commitment. This is edition one of an index we intend to repeat, so that the second edition can say what actually changed rather than what we think might be changing.
If you spot an error in it, tell us. A study that invites people to check the arithmetic is the point, and the risk.
Want to know where you stand?
We can run the same measurement against your category. Send us your site and we will come back with a free audit: where you are in search and AI, and what we would do first.