Does ChatGPT Recommend Only 1.2% of Local Businesses, According to SOCi?

What did SOCi actually measure?

SOCi's published methodology describes the setup:

The same study scored each brand on more than 120 metrics, including Google 3-Pack presence, reviews and social activity. The AI results sit alongside those.

What were the headline results?

SOCi's benchmark table for the three platforms, as published on its statistics page:

SOCi also reported that brand locations appeared in AI recommendations 6.5% of the time across platforms, and that locations appeared in Google's 3-Pack 35.9% of the time. It described AI visibility as nearly 30 times more selective than traditional local search, which is the ratio of 35.9% to 1.2%.

Why does the 1.2% figure get misquoted?

Headlines turned “1.2% of multi-location brand locations” into “1.2% of local businesses”. Four details change the meaning:

1. The sample was chain locations. Think of a coffee chain's branch, a bank branch or a tyre chain outlet, not the independent plumber or café. SOCi says language models often favour single-location businesses, whose reviews and data are more focused.
2. One prompt per location. A broad “recommend businesses of this type in this market” question. Real people ask more specific questions, such as a narrow-feet running shoe shop nearby, which can surface different businesses. SOCi argues that a brand missing from the broad question is unlikely to appear in narrower ones; that is its assumption rather than something the study tested.
3. Only the top five counted. A market with 60 coffee shops cannot have more than five in any single answer. Some of the low rate is arithmetic: an answer that names five places leaves most locations out by design.
4. The 3-Pack comparison is not like for like. The 35.9% comes from a different measurement, so “30 times harder” is an illustration, not a controlled comparison.

None of this makes the study wrong. It is one of the largest public tests of AI local recommendations. It measures something narrower than the headline.

Why was ChatGPT so much lower than Gemini?

SOCi's explanation is where the platforms get their local data. Gemini is grounded in Google Maps, which SOCi used as the reference for accuracy, so Gemini's profiles matched perfectly and its recommendations drew on the same listings as local search. ChatGPT and Perplexity have no single central source for places. They combine information from several first-party and third-party sources, which SOCi says produced missing or incorrect details at scale: about 68% of profiles were accurate on each.

Two practical points follow. Your Google Business Profile matters most for Gemini. For ChatGPT and Perplexity, consistency across the other sources that describe you, such as Yelp, Facebook, directories and your own location pages, matters more. How ChatGPT picks places is covered in how ChatGPT chooses which local businesses to recommend, and the role of inconsistent listings in whether inconsistent NAP data can stop ChatGPT recommending you.

What did the study say about ratings?

Locations that ChatGPT recommended averaged 4.3 stars, and Perplexity's averaged 4.2. Gemini's averaged 3.9, so ratings appear to weigh less in its choices, which may reflect its reliance on Maps data. SOCi's conclusion is that reviews weigh heavily in ChatGPT's choice. The figure describes the businesses that were recommended; it does not show a star threshold. The practical step is the usual one: ask every customer for an honest review in their own words, and reply to reviews.

What does this mean for a single-location business?

The study does not measure you directly. Treat it as three signals:

• AI answers name very few businesses, so being one of a handful matters more than ranking somewhere in a long list.
• Data accuracy is a real problem on ChatGPT and Perplexity. Check what they say about your hours, address and services, as covered in why ChatGPT recommends businesses that have closed.
• Your own rate needs your own test. Ask the assistants the questions your customers ask, in your area, several times over several weeks. Tools that do this at scale are assessed in the Local Falcon review.

Whether those recommendations turn into customers is a separate question, covered in whether businesses get leads from ChatGPT recommendations.

What does it mean for a multi-location brand?

SOCi's argument is that assistants judge a brand's whole footprint, so a minority of locations with wrong hours, poor reviews or thin pages can hold back every location. The study's stated behaviours of leading brands were treating accuracy as a strategic asset, differentiating rather than merely being present, winning on reputation, and running traditional and AI channels, and search and social, as one system. Those are reasonable, but they are SOCi's interpretation. The study shows correlations across brands, not the effect of any one change.

We have not rerun SOCi's test, and our own checks of local answers are far smaller. The figures here come from SOCi's published methodology and its statistics page.

What might this look like in practice?

Consider an illustrative 40-location dental group, Zellnell Dental, whose marketing lead saw the 1.2% headline and panicked. Instead of relying on it, the team asked ChatGPT, Gemini and Perplexity “dentist open Saturday in [city]” and “emergency dentist near [neighbourhood]” for ten locations, three times each over two weeks. Gemini named Zellnell in about a third of answers; ChatGPT rarely did, and twice gave a closed branch's old address. The team corrected listings on Yelp, Apple Business and three directories, rebuilt thin location pages with services, hours and insurance details, and asked every patient for an honest review. They repeated the same test monthly and tracked both mentions and accuracy.

What do people ask about the SOCi study?

Does ChatGPT recommend only 1.2% of all local businesses?

No. The 1.2% applies to locations of large multi-location brands in the US, tested with one broad prompt and counting the top five answers.

Which AI assistant recommended the most locations?

Gemini, at 11.0%, followed by Perplexity at 7.4% and ChatGPT at 1.2%.

Are small independent businesses recommended more often?

SOCi says language models often favour single-location businesses, but its 2026 index did not measure them.

How accurate was ChatGPT's business information?

About 68% of ChatGPT's business profiles were accurate against Google Maps data, similar to Perplexity. Gemini's matched fully because it uses Google Maps.