The Map Pack Paradox: Winning Google Maps in India No Longer Means Winning the Answer

Two rankings, one search box

A local business owner checking their standing on Google typically looks at one thing: where do I show up in the map pack, the three pinned listings with a photo, a star rating, and a "Directions" button that appear for searches like "electrician near me" or "best bakery in Indiranagar." For over a decade, that ranking was the whole game. A business that ranked in the three-pack was the one the customer saw first.

That is no longer the only surface competing for the same query. Google has been layering a summary produced by an AI system, AI Overviews, above and sometimes instead of the classic results, and it reads and cites a different set of sources than the ones the map pack ranks. A business can hold position one in the local pack and not appear anywhere in the paragraph the buyer actually reads first. These are not two views of the same ranking. They are two rankings, decided by different mechanics, sharing one search box.

That gap is the map pack paradox: the metric an Indian MSME has spent years optimizing for, a Google Business Profile with reviews and a claimed listing, still matters, but it no longer guarantees a place in the answer a growing share of searchers now read instead of the list beneath it.

What actually rolled out in India, and when

AI Overviews reached India in August 2024, in the same wave as Brazil, Japan, the UK, Indonesia, and Mexico, with local-language support built in from the start. Google added India-specific features alongside the launch: a toggle to switch between Hindi and English results without leaving the page, and a "Listen" button so users could hear the summary read aloud, a feature Google noted Indian users engaged with more than users in other markets.

Adoption then moved fast. By July 2025, Google reported AI Overviews had reached two billion monthly users globally, up from 1.5 billion just two months earlier, while AI Mode, the more conversational, multi-turn version of AI search, had passed 100 million monthly active users across the US and India combined. In October 2025, Google expanded AI Mode to seven more Indian languages, Bengali, Kannada, Malayalam, Marathi, Tamil, Telugu, and Urdu, and launched Search Live, a real-time voice-and-camera search mode, in India first outside the US. Google's own announcement named India as its largest user base worldwide for both voice and visual search, and noted that more people use Google Lens in India every month than in any other country.

None of that is a forecast. It is what already shipped, in the language mix Indian MSMEs actually operate in, on the search engine that already sends most of their organic local traffic.

What the independent data shows about the two surfaces diverging

No public, India-specific study yet measures the overlap between AI Overview citations and local three-pack rankings the way researchers have measured it for the US. That is a real gap in the evidence. What exists is US-anchored research on the same underlying mechanism, Google's generative answer layer competing with its own classic results, and the direction it shows is unlikely to be India-exclusive given the AI Overviews rollout uses the same global ranking system.

Ahrefs, tracking AI Overview citations against the organic top 10 across roughly 863,000 keyword result pages, found that in July 2025 about 76% of AI Overview citations also ranked in the top 10 of classic search. By early 2026 that overlap had fallen to roughly 38%. In other words, a page or listing that ranks well in the traditional results is now cited in the AI answer for that query only a little more than a third of the time, down from three-quarters less than a year earlier. The two indexes, the one Google shows as blue links and the one it draws on to write its summary, are pulling apart.

Click behavior is moving with it. Pew Research Center, using a desktop browsing panel, found that when an AI summary appeared, users clicked through to a traditional search result in about 8% of visits, versus 15% of visits when no AI summary was present, and clicked a link inside the AI summary itself in only about 1% of visits. Ahrefs, analyzing 300,000 keywords and comparing Google Search Console click-through data from December 2023 to December 2025, found the average click-through rate for the position-one result had fallen 58% once an AI Overview was present for that query. Ranking first increasingly means fewer people ever reach the ranked page.

Why this lands harder on India specifically

India's internet base gives this shift more reach than a smaller market would see. The IAMAI-Kantar ICUBE 2025 report put India's active internet user base at 958 million as of early 2026, growing about 8% year-on-year, with rural India now accounting for 57% of that base and growing at nearly four times the urban rate. That is not a niche audience testing a beta feature; it is close to a billion people, an increasing share of them in smaller towns, using the same search engine that is rolling AI Overviews out in their own language.

On the supply side, the Ministry of MSME reports over 7.83 crore enterprises registered on the Udyam Registration Portal and Udyam Assist Platform as of February 28, 2026, up from 6.19 crore a year earlier. That is the population of businesses whose local discovery increasingly runs through the same search box now splitting into two rankings. Most of them built their local-search presence around the metric that used to be sufficient: a claimed, reviewed Google Business Profile that earns a spot in the three-pack. That metric has not stopped mattering. It has stopped being the whole answer.

The mismatch is structural, not a matter of trying harder at the old game. A three-pack ranking is decided largely by proximity, relevance, and review signals tied to a Business Profile. An AI Overview citation is decided by what Google's summarization layer judges worth quoting, drawing, per the independent research above, from a materially different and now less-overlapping set of sources. A business optimized only for the first can be structurally invisible to the second, and neither its owner nor its current reporting will show that gap, because Google Business Profile insights report map-pack performance, not whether the business appears inside the paragraph sitting above it.

What actually earns a citation, as far as the evidence goes

The peer-reviewed literature on what earns a citation inside a generated answer is thin but not empty. Aggarwal and colleagues, in the paper that introduced the term "Generative Engine Optimization" at KDD 2024, tested which content changes moved a source's visibility inside generated answers and found that adding cited statistics, direct quotations, and citations to authoritative sources measurably increased how often a page was drawn on. The lever is corroboration, not keyword density: a page that states a specific number with a traceable source behaves differently, in this line of research, than a page that makes the same claim in more general language.

Applied to local business listings, the plausible extension, and it is an extension rather than a directly tested finding, is that the same logic likely governs whether a local citation earns a place in an AI Overview: specific, verifiable facts (accurate hours, a real service area, review content with concrete detail) probably behave differently than generic profile text. That inference follows the corroboration logic the GEO paper demonstrated, but no published study has yet tested it against India's own local-business surface specifically, and it should be held with that qualification rather than presented as settled.

What to actually do with a paradox this new

No vendor, including one running this diagnostic, currently has an audited, India-specific measurement of AI Overview citation rates for local businesses to optimize against. Two things are defensible now. First, treat the two surfaces as genuinely separate and check both, rather than assuming a strong Business Profile ranking implies AI-answer visibility; the US data above says that assumption is now wrong more often than it is right. Second, strengthen the corroboration signals the GEO research shows move citation behavior in the surfaces that have been tested, specific facts, sourced claims, real review substance, rather than treating profile completion as a finished task.

What genuinely does not exist yet is a validated, India-specific measurement of how often a locally ranked business is cited inside the AI answer for the same query, broken out by language, city tier, and industry. That is an open research question, not a solved one, and any number offered as an answer to it today, in India specifically, should be checked against its actual source before it is trusted.