Generative engine optimization India: rank is not enough

Sixteen of thirty-six. That is how many of the sources Google's own AI Overview named, across five Indian buyer queries, had not earned a place in the top 10 organic results for the query that produced them. The overlap runs the other way too, and it is worse: nearly half the sites Google ranked on page 1 were not mentioned in the answer sitting directly above them.

This matters because the entire case for ignoring generative engine optimization rests on one assumption. Rank well and the AI will find you. In our sample that assumption held only partly, and in one industry it barely held at all.

Do AI Overviews even appear on Indian commercial queries?

On these five, yes. Every one of the five queries we ran returned an AI Overview block on Google India.

That is worth flagging against our own earlier measurement. In a 160-query study we ran on 7 August, commercial queries carrying a city name returned no AI Overviews at all, and non-city commercial queries returned them in 22 of 50 cases. Five out of five is a much higher rate than that 44%.

There are two honest readings, five queries cannot separate them, and the finding below does not depend on which is true.

• These five national, cost-and-comparison queries are unusually prone to triggering an overview
• Or coverage has moved since 7 August

What we measured, and how

We locked the method in writing before running anything. Five commercial buyer queries, India, one per industry, none carrying a city qualifier.

• CRM software
• Solar panel installation cost
• Online MBA
• IVF treatment cost
• Term insurance

For each query we then did three things.

• 1

  Pulled the domains cited inside the AI Overview block

• 2

  Pulled the domains holding an organic position in the top 10

• 3

  Compared the two lists

Directories, marketplaces and aggregators were deliberately not excluded, because the question is who Google ranks, not who we think deserves to rank.

The one change we made to the method

The original design compared organic rank against citation totals across seven engines. When the SERP tool turned out to return the AI Overview's own citation list for the same query, we amended the primary comparison to use that more direct measure. The amendment was written down before any statistic was computed, and the five queries never changed.

How often do rank and citation actually disagree?

Three numbers come out of that.

• 16 of 36 cited sources, or 44.4%, did not hold a top-10 organic slot
• 18 of 38 top-10 organic results, or 47.4%, went uncited
• Extending the organic list past position 10 where we captured it moves the first number only to 15 of 36

Google says there is nothing extra to do, and that is not a contradiction

The obvious reading of the table is that Google's AI is doing something separate from ranking. Google's documentation says otherwise, and both things can be true.

Google's guide to optimizing for generative AI features states its position plainly.

"From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO."

Its AI features documentation is equally direct about requirements.

"There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary."

The mechanism that reconciles those statements with our table is query fan-out. Google describes it as a set of concurrent, related queries the model generates to fetch additional results beyond the one you typed. The AI features page then says the point of this is to display a wider and more diverse set of helpful links than a classic web search would.

Read that alongside the data and the picture is coherent. The AI Overview is not ranking your query. It is answering a bundle of sub-questions, and it pulls sources that rank for those instead.

"On the evidence of this sample, you are not competing for ten slots any more. You are competing for a set of slots you cannot see and did not choose."

The clearest case was CRM software

Two of the largest CRM brands in the world sat in the organic top 10 for the query crm software india, and neither appeared in the AI Overview above them.

Five of the eight sources the overview did cite held no top-10 organic position at all. Those five included a logistics software blog, a manufacturing consultancy based in Canada, and a software review site.

Does the AI Overview agree with the other assistants?

No, and this is where the case for treating AI visibility as one thing falls apart. We took every domain involved in that CRM query, both the ones ranking and the ones cited, and pulled citation counts across seven AI engines.

Three things fall out of that table.

• 5 of the 13 domains have no ChatGPT citations at all
• Only 6 of 13 are cited by all seven engines
• The AI-Overview-only group looked worse on ChatGPT than the organic group did, at 3 of 5 against 1 of 5

Being picked by Google's AI Overview did not predict being picked by ChatGPT. At three domains against five we are describing a pattern, not measuring a rate. It does line up with the wider result we got from 28 marketing-services domains on 14 August, where the seven engines disagreed sharply with each other and not only with Google's own overview.

Why the two routes differ

The routes differ because the plumbing differs. OpenAI's crawler documentation names a specific agent for its search feature.

"OAI-SearchBot is used to surface websites in search results in ChatGPT's search features."

That is a separate pipe from Googlebot, with a separate opt-out. We have written before about why absence from an AI answer is two different failures, and this is the supply-side one.

What we would actually do about it

• 1

  Check that your pages meet Google's stated bar first, which is being indexed and eligible to show with a snippet. Nothing below matters until that is true

• 2

  Stop treating your head keyword as the target. Write for the sub-questions a buyer asks around it, because those are what fan-out retrieves

• 3

  Confirm your robots.txt does not disallow OAI-SearchBot, since that alone removes you from ChatGPT search answers

• 4

  Measure per engine and never pool them. A single AI visibility score hides exactly the disagreement we found

• 5

  Skip the tactics Google explicitly says it ignores. Its guide names llms.txt files, content chunking and inauthentic mentions as things you can disregard for Google Search

That last one deserves emphasis, because a lot of GEO advice sold in India right now consists precisely of those tactics. Google's position that its AI surfaces are still SEO is not the whole story across every engine, but for Google specifically it is published, and it is free to read.

What this cannot tell you

• Five queries, one country, one day. This is a probe, not a rate. It is enough to show the assumption fails somewhere, and nowhere near enough to say how often
• The citation counts come from Ahrefs' index rather than from us prompting each engine. Google's own guide warns that no third-party tool has access to its internal ranking or AI systems, and that warning applies to the tool we used
• One domain we saw cited live in the AI Overview showed zero AI Overview citations in that index, so the index and the live SERP do not always agree
• A zero may mean an engine does not cite a domain, or that index coverage for it is thin. We have not ruled out the second
• Citations are not customers. Nothing here measures whether an AI mention produced a single enquiry