What AI-Recommended Local Businesses Have in Common: 3,149 Checked (2026)

Data last verified: 16 October 2026. Figures re-checked against their primary sources under the Whito rolling re-verification programme.

Last Updated on October 16, 2026

Collected: AI answers 22 August 2026; Google local results 16 October 2026. 159 searches, eight trades, 20 UK towns, 3,149 businesses. This page reports totals only. It names no business.

Plenty of people have a theory about what gets a business recommended by AI. We had a way to test it: a dated log of every business Google AI Mode named when we asked it for a trustworthy tradesperson or practice in 20 UK towns.

So we put those businesses next to the ones it did not name, using the same Google local results for the same search, and looked at what was different.

Of 3,149 businesses in Google’s local top 20 for 159 trade and town searches on 16 October 2026, those ranked in the top three had been named by Google AI Mode on 22 August in 54% of cases, against 8% for those ranked 11th to 20th.

Review volume separated the two groups too. Star rating barely did. The businesses AI Mode named had a median of 140 Google reviews, against 61 for the rest of the top 20. Both groups had a median rating of 4.9.

What we did

On 22 August 2026 we asked Google AI Mode “Recommend a trustworthy [trade] in [town]” for eight trades in 20 towns and logged every business it named. There is one recorded answer for each trade and town; where Google failed to generate one, we asked again the same day. That log is the Whito AI Visibility Index, Q3 2026.

The raw answer text we hold covers 159 of the Index’s 160 answers. The vets answer for Wolverhampton is missing from it, so that search is left out here. The 159 answers contain 728 named entries, a few of them not single businesses, such as a line naming two individual cleaners on a jobs platform. The published Index counts 734 businesses across all 160 answers.

On 16 October 2026 we ran the plain search “[trade] in [town]” on Google, opened the Places results, and recorded the top 20 businesses for each search: position, star rating, review count, main category, years in business where Google shows it, and whether a website button appears. Adverts were left out. For 97 searches the browser loaded the results page in the background; for the other 62, after Google started limiting those requests, we opened each results page in a browser tab and read it there. That gave 3,149 businesses across 159 searches.

We then matched each business AI Mode named in August to that list by name, within the same town and trade. A business counts as “named” if it appeared in the August answer, and as “the rest” if it sits in the same top 20 but was not named. 620 of the 728 August names in these searches were found in the October top 20 (85%).

Either way, this is the public results page any customer sees. We did not use the Google Places API, and we publish totals only, not any business’s ratings or reviews.

Named and not named, side by side

Position in the local list and review count show large gaps. Website button, category and years in business show small ones. The star rating shows almost none, and the businesses AI Mode named were slightly less likely to hold a perfect 5.0.

Named businesses sat near the top of Google’s own local list

The business in first place had been named by AI Mode in 101 of 159 searches (64%). AI Mode and the local results are both Google products drawing on the same Business Profiles, so a strong overlap is not a surprise. The size of the overlap is the finding. In our data, AI Mode’s picks look much like the top of the local results for the same search.

Review volume mattered. Star rating barely did.

The chance of being named rises steadily with review count. Stars only separated the groups at the bottom: businesses rated under 4.5 were named 9% of the time, against 17% to 27% for each band above that. Higher up, more stars did not help, and a perfect 5.0 did worse than a 4.8 or 4.9. The typical perfect score sat on a small review count: the median 5.0-rated business had 35 reviews, against 129 for everyone else. Within each search, the named businesses had a higher median review count than the rest in 136 of 159 searches. Their median rating was higher in 51, lower in 58 and the same in 50.

Google’s own help page for Business Profiles says “More reviews and positive ratings can help your business’s local ranking”. Our data fits the first half of that sentence better than the second.

When position and reviews disagree

Position and review count usually move together, so we looked at the businesses where they split: high in the list with few reviews for their search, or low in the list with some of the most reviews.

In our data, a high position with few reviews was named more often (34%) than a low position with many reviews (12%). The second figure is still above the 8% for all businesses ranked 11th to 20th, so reviews showed some link on their own. These are small groups, so treat the gap as a direction, not a measurement.

By trade

In every trade the named businesses had a higher median review count than the rest of the top 20. In five trades, electricians, plumbers, roofers, dentists and accountants, they had more than twice as many. The smallest gap was among cleaners.

What this does not prove

• Eight weeks sit between the two readings. AI answers were collected on 22 August 2026 and the local results on 16 October 2026. Rankings and review counts move in that time. The gap could blur the link or, if being named in August brought in customers and reviews, exaggerate it. We cannot tell which.
• This is an association, not a cause. AI Mode and the local results are both Google products, and both lean on the same profiles. We cannot say AI Mode reads the local ranking. We can say its picks look like it.
• One engine. Google AI Mode only. ChatGPT, Gemini and Perplexity were not tested here.
• Top 20 only. 108 August names were not in the October top 20 for their search, or could not be matched by name. They are left out of the side by side comparison, not counted as “the rest”.
• Names were matched by rule, ignoring legal suffixes, the town name and words like “roofing” or “estate agents”. The 97 matches that were not exact were checked by hand, and two missed pairs were added by hand. Chains with several branches in one town may be matched to the wrong branch.
• Photo counts were not measured. The local results page does not show them.
• One search per trade and town, run from one UK location while signed in to a Google account. Google adjusts local results by where the searcher is, even when a town is named.
• A few searches returned fewer than 20 businesses: vets in Luton (4), vets in Blackpool (13), vets in Mansfield (12).

What to do with this

Start. Fix the profile before chasing AI. Check your Google Business Profile category says what you do, and that the website button works. These are small differences in our data, but they are free.

Build. Ask every finished job for a review, every week. In our data, review volume tracked being named more closely than a perfect score did. For example, a 4.8 or 4.9 with 100 or more reviews had been named 33% of the time. A 5.0 with fewer than 20 reviews, 7%.

Scale. Check where you sit in the local results for the towns you serve. In our data, businesses ranked 11th to 20th in October had been named 8% of the time, against 54% in the top three.

Which sources AI Mode cites when it names a business is in our AI citation sources study, and how that differs by trade is on the AI citation map. To see what engines can read about your own business today, run the free Visibility Check.

Common questions

What do businesses recommended by Google AI Mode have in common?

In Whito’s October 2026 comparison, they mostly sat near the top of Google’s own local results and had more reviews. Businesses in the local top three had been named by AI Mode in 54% of cases, against 8% for those ranked 11th to 20th. Median reviews were 140 for named businesses and 61 for the rest.

Does a 5.0 star rating get a business recommended by AI?

Not in our data. Named and not named businesses had the same median rating, 4.9. Businesses rated 4.8 or 4.9 were named more often (27%) than those rated a perfect 5.0 (17%).

How many Google reviews does a business need to be recommended by AI?

We found no clear cut-off. The share named rose with review count: 6% of businesses with fewer than 10 reviews had been named, against 31% of those with 250 or more.

Is this the same as Google’s local ranking?

They are different products, but in our data they overlapped heavily: the first local result had been named in 101 of 159 searches (64%), and 54% of top-three businesses had been named.

Cite this research

Whito Research (2026). What AI-Recommended Local Businesses Have in Common. Whito. https://whito.co.uk/research/what-ai-recommended-businesses-have-in-common/

Key finding: Of 3,149 businesses in Google’s local top 20 for 159 trade and town searches on 16 October 2026, those ranked in the top three had been named by Google AI Mode on 22 August in 54% of cases, against 8% for those ranked 11th to 20th.

Whito

Whito is an independent UK research publisher. We publish research on how UK businesses are found, trusted and priced: what they cost to market, how their sectors work, and who actually gets recommended. Studies are free to read, every figure carries the date it was checked, and nobody pays to appear.

See Full Bio