Who Actually Answers a Local Search in India: 48 Queries, One AI Overview

The measurement, and the number that undercuts the premise

Over August 2026, Raveneye Global measured eight local service trades, chartered accountant, dentist, gym, interior designer, digital marketing agency, coaching institute, physiotherapist, and wedding photographer, across six ordinary, non-metro Indian cities: Jaipur, Indore, Lucknow, Surat, Kochi, and Nagpur. The underlying data came from live APIs: Google Search results and AI Overview data, Google Places, Google Ads search volume, Google Knowledge Graph, Google PageSpeed Insights, and DNS-over-HTTPS. This article is one cut of that dataset, the cut that asks a single, specific question: when a buyer types a trade and a city into Google, what actually answers back.

The premise worth testing directly, rather than assumed, is a specific one: that the generated answer has already taken over the local decision the way it is widely described as taking over search in general. Coverage of AI Overviews and AI Mode tends to describe a single, undifferentiated shift, search becomes an answer instead of a list, without separating local commercial queries, the ones that actually decide which chartered accountant or dentist a nearby buyer calls, from the broader mix of informational and explanatory queries most published AI Overview statistics are built on. This study isolates that local commercial slice specifically, for India, and reports what it found there rather than assuming the broader trend applies to it unchanged.

To answer it, Raveneye Global ran 48 live Google searches on 2026-08-10, one query per trade per city, following the fixed pattern best [trade] in [city], for example best dentist in Kochi or best gym in Surat, and recorded exactly what Google returned above the fold: an AI Overview, a local map pack, or neither. Across these 48 local commercial searches, an AI Overview appeared in exactly 1 of the 48 results, 2.1 percent. Forty-seven searches out of forty-eight returned no AI Overview of any kind.

What did answer, some of the time, was Google's local map pack, the block of business listings with a small map, ratings, and hours that sits above the organic list. It appeared in 17 of the 48 searches, 35.4 percent, averaging 3 businesses per pack, the standard Google three-pack. That overall rate hides the more interesting fact underneath it, taken up in the next two sections: the map pack did not spread evenly across the eight trades measured. It was either close to universal or entirely absent, depending on which trade was being searched.

Why this query pattern, and what it deliberately does not test

The fixed pattern, best [trade] in [city], was chosen because it is close to the plainest, most direct way an Indian buyer types a local comparison search, closer to ordinary phrasing than a keyword-tool style query like [trade] near me, which trades a specific city name for a device's own location signal instead. Holding the template constant across all 48 searches, the same three words connecting a trade to a city every time, means what varies between results is the trade and the city being searched, not the wording used to ask for them.

That same fixed design has a real limit worth stating rather than hiding. A single phrasing cannot speak for every way a buyer actually searches. A near me query, a voice request into a phone, or a follow-up question typed into a conversational search product is a different linguistic object than best [trade] in [city], and each has been shown elsewhere to trigger Google's generative features at meaningfully different rates. A comparison introduced later in this piece, from a separate 2026 study that tested three different intent phrasings per trade, found the AI Overview rate swinging from 15 percent to 97 percent depending entirely on which phrasing was used for the same underlying category. This study intentionally tested one phrasing, cleanly, across every trade and city, rather than several phrasings thinly. The numbers that follow should be read as an answer to that one specific, common, comparison-shaped query, not as a claim about how every kind of local search behaves in India.

A third of the time, and only for three of eight trades

Broken out by trade, the 35.4 percent overall map-pack figure resolves into two distinct groups rather than a gradient. Three of the eight trades measured showed a local map pack in nearly every city tested. The other five showed one in none of theirs, across 30 combined searches.

• Chartered accountant: local map pack in 6 of 6 cities (100%).
• Dentist: local map pack in 6 of 6 cities (100%).
• Interior designer: local map pack in 5 of 6 cities (83%).
• Gym: local map pack in 0 of 6 cities (0%).
• Digital marketing agency: local map pack in 0 of 6 cities (0%).
• Coaching institute: local map pack in 0 of 6 cities (0%).
• Physiotherapist: local map pack in 0 of 6 cities (0%).
• Wedding photographer: local map pack in 0 of 6 cities (0%).

Reading the split without overreaching

One AI Overview in forty-eight tries, and why that is not an availability problem

It would be a mistake to read the 2.1 percent AI Overview rate as evidence that the feature has not really arrived in India. It has, at a scale few products reach this quickly. Google placed India in the first group of countries to receive AI Overviews at all, launching it on August 16, 2024 alongside Brazil, Japan, the United Kingdom, Indonesia, and Mexico, built for English and Hindi from the first release rather than added later. Ten weeks after that, on October 28, 2024, Google scaled the feature past 100 countries and projected it would reach more than a billion monthly users. An Indian searcher has had access to AI Overviews, in principle, on a very wide range of queries, for close to two years by the time these 48 searches were run.

What the data shows instead is a selection effect specific to this kind of query. Google is explicit that AI Overviews are not shown on every search: its own Search Help documentation states that the feature appears "when our systems determine that generative AI can be especially helpful, for example, when you want to quickly understand information from a range of sources." A query of the form best [trade] in [city] is not really a request to understand information from a range of sources. It is closer to a request for a short, specific list of nearby options, the exact job the map pack and the organic list already do. That reading fits ordinary local buyer behavior too: someone typing best dentist in Lucknow is usually a step or two from calling or visiting, not from reading an explanation. Three names, ratings, and a phone number, exactly what the map pack already supplies where it appears, resolves the query about as completely as a written summary would, which may be a second, complementary reason Google's systems judged an additional summary unnecessary on 47 of the 48 searches, alongside the selection standard the company states outright.

Independent research outside India points the same direction, at a higher floor. A 2026 case study by Whitespark, a local-search research firm, ran 540 manual queries across three large US cities, Houston, Phoenix, and Denver, and six local-business verticals, varying each query across local, informational, and hybrid intent. Plain local-intent searches, of the same best-in-city shape used in this study, returned an AI Overview just 15 percent of the time, against 92 percent for informational queries and 97 percent for hybrid ones. That 15 percent describes a different market, a different sample size, and a mixed set of intent phrasings rather than the single fixed template used here, so it should be read as directional corroboration, not a replication. But the direction agrees: of every kind of search Google fields, a plain local commercial query is among the least likely to return an AI Overview, in the United States, and at an even lower measured rate, in these six Indian cities.

What actually filled the other spaces

Whether or not a map pack or an AI Overview appeared, every one of the 48 searches also returned Google's classic list of ten organic links. For the five trades that never triggered a map pack, gym, digital marketing agency, coaching institute, physiotherapist, and wedding photographer, that organic list was not a fallback sitting below the real answer. It was the entire answer, the only surface Google offered for 30 of the 48 searches in this study.

Who occupies those ten positions, individual business websites, directories, aggregators, or review platforms, is a separate question from the one this piece measures, and it is the subject of a companion study in this series, The Aggregator Toll on Organic Search. What matters here is narrower and more specific: for most of the trades in this study, the organic list is not one surface among three. It is the surface, full stop, since neither of the other two showed up.

A resolvable entity matters more, not less, once the map pack and the AI Overview are both absent, since the organic list is then carrying the entire weight of the decision on its own. Whether an individual business's own page, as opposed to a directory listing about it, is even resolvable as a distinct, verifiable entity to Google in the first place is the question taken up in a second companion study in this series, on the entity gap many small Indian businesses carry into search, The Entity Void.

A snapshot of a moving surface, not a fixed one

Every number in this piece describes one date. The 48 searches behind it were run on August 10, 2026, once each, in six cities, for eight trades. That is a real, precisely counted measurement, not an estimate, but it is also a single probe of a result page that Google rebuilds constantly. A map-pack rate of 35.4 percent, or an AI Overview rate of 2.1 percent, describes what Google showed on that day, for that fixed query pattern, in these six cities. Treating either number as a permanent constant would be a larger claim than this dataset supports, and this piece does not make it.

Turning this from a single, precisely measured snapshot into an established pattern would take repetition: the same 48 queries, or a larger set built the same way, run again at a fixed interval, monthly or quarterly, so a rate of 2.1 percent or 35.4 percent could be reported with a trend behind it rather than a single date. That repeated version of this study does not yet exist. Until it does, every number here is reported as exactly what it is, one dated count, and graded accordingly through this piece's evidence tiers.

The direction Google itself has moved suggests the more likely expectation is change, not stasis. The same company that took AI Overviews from a six-country launch to more than 100 countries in ten weeks has shown no pattern of leaving a search feature's reach fixed for long. Nothing in the public record says when, or whether, that expansion reaches this specific local-commercial query shape at meaningfully higher rates than the 2.1 percent found here. What the record does say is that the company's own AI-answer products have moved quickly before, in this exact market, which is reason enough to re-run this measurement in time rather than assume today's numbers hold.

The position, as of August 2026, is narrow and specific rather than sweeping: for these eight trades, in these six cities, on this one date, the AI Overview answered once and the map pack answered for three trades. Both facts are stated as exactly that, a dated reading of a surface still in motion, not a forecast of where it settles.

What this means for machine readiness today

The practical reading of this dataset splits cleanly along the same line the map pack itself drew. For a chartered accountant, a dentist, or, with slightly less certainty, an interior designer, the map pack is close to the entire local contest right now: three visible slots, decided by category match, distance, and prominence, exactly the mechanism Google documents. Machine readiness for those trades starts with the unglamorous basics that mechanism rewards, the correct primary category, information that matches across every listing, and a review base substantial enough to read as prominent, not with anything specific to AI answers at all.

For the other five trades measured here, gym, digital marketing agency, coaching institute, physiotherapist, and wedding photographer, neither the map pack nor the AI Overview reliably appeared in this study. The organic list carried the entire decision in every one of those 30 searches. Readiness there is a narrower, more classic problem: being a resolvable, well-represented entity inside a list of ten links, not inside a three-listing pack or a summary that, for these trades, essentially did not show up.

None of this argues that the AI Overview is irrelevant to plan for. The Whitespark comparison above shows a market where the same style of query already clears 15 percent, roughly seven times this study's Indian rate, and Google has repeatedly moved its India products from a narrow launch to broad coverage inside a matter of months. The position is that the AI answer is not yet the local battleground these eight trades compete on in India, not that it never will be. The only way to know which surface actually decides a specific business's own local search, the pack, the list, or eventually a generated answer, is to measure that business directly rather than apply a category average it may or may not resemble.