Does AI agree with Google on who's best in town? The Awkward Media Local Search Index 2026

Columbus, Ohio, on a Monday morning in November. Ask Google Maps for an electrician from downtown and the top three listings have 39, 3 and 2 Google reviews. Ask ChatGPT, Perplexity, Gemini and Google's own AI Mode who the best electrician in Columbus is, and all four give the same first answer: Safe Electric & Plumbing, a firm with 2,967 reviews that Maps puts at number 19.

Same city, same trade, same morning. Two different answers to "who is best".

We wanted to know how often that happens. So we asked the question 1,600 times.

What did we find?

We asked four AI assistants to name the best business in 10 trades across 20 US and Canadian cities: 200 searches, each put to each assistant twice, a day apart. Then we checked every first answer against the Google Maps top 20 for the same trade, searched from the same city centre.

• Only 7% of the businesses an AI assistant named first were in the Google Maps top 3 for the same search: 53 of 792 readable answers on November 2, and 54 of 790 when we asked again on November 3.
• 68% were not in the Maps top 20 at all: 541 of 792 on day one, 536 of 790 on day two.
• The assistants rarely agree with each other. In 102 of the 200 searches, no two of the four named the same business first. All four agreed only 6 times.
• Ask again tomorrow and the "best" often changes. A day later, the first business named was the same in only 42% of cases (330 of 782).

The short version for a business owner: being in the Maps top 3 does not put you first in an AI answer, and being first in an AI answer does not need the Maps top 3. They are two different shop windows, and each needs its own attention.

Why should a business owner care?

A customer who types "plumber near me" into Google sees a map with three businesses on it and a list below. A customer who asks ChatGPT "who's the best plumber in Austin?" gets one name first, and often only a handful in total. There is no page two.

Google Maps is still one of the main places local customers find a business, and the work that gets a business into the map's top three (here's how that works) keeps paying off. AI answers are an extra place where customers now decide who to call. What our numbers show is that a strong position in one does not carry over to the other on its own.

How did we run the study?

In two lines: on November 2, 2026 we asked ChatGPT, Perplexity, Gemini and Google AI Mode "Who is the best [trade] in [city]?" for 10 trades in 20 cities, with the city named in every question and, where the assistant allows it, its location set to that city. We checked the first business each named against the Google Maps top 20 for the same trade, searched from the city centre that morning. We asked every question again on November 3, 24 to 25 hours later, to see whether the answers held, and checked those answers against the same November 2 Maps list.

• Trades (10): plumber, electrician, roofer, HVAC, dentist, personal injury lawyer, auto mechanic, pest control, landscaper, chiropractor.
• Cities (20): Austin, Houston, Dallas, Phoenix, Denver, Chicago, Atlanta, Miami, Tampa, Charlotte, Nashville, Seattle, Las Vegas, San Diego, Philadelphia, Columbus, Toronto, Vancouver, Calgary, Ottawa.
• Answers: 800 a day, 1,600 in total. 792 were readable on day one and 790 on day two. A few answers, mostly from Google AI Mode, were cut off before any business was named; we counted those as unreadable rather than guess.
• "First pick" means the first business the assistant named in its answer. For each answer we recorded its opening (about 600 characters): the first business named, the others named and the sources shown.
• Matching: a strict match on the business's own name. A firm with several Maps listings counts at its highest one. A renamed firm counts only when its own website confirms the new name.
• Maps: the top 20 unpaid listings for "[trade] in [city]", searched from one point, the city centre. Paid "sponsored" listings were left out.
• Limits: Maps results change with where you stand, so a search from the suburbs would differ. What a signed-in person sees in an assistant can differ from an answer to a fresh question with a set location. Two days is a snapshot, not a trend.

Which AI assistant agrees most with Google Maps?

Google's own AI Mode came closest, and even it disagreed most of the time. Its first pick was in the Maps top 20 in 39% of answers (74 of 192), which means that 61% of the time Google's assistant named a business outside Google's own Maps top 20.

ChatGPT was furthest away. Its first pick was in the Maps top 3 in 9 of 200 answers, and in the top 20 in 26% (52 of 200). Perplexity sat at 29% for the top 20 and Gemini at 34%.

Do the AI assistants agree with each other?

Mostly, no. In 102 of the 200 searches, the four assistants named four different businesses first. In 75 searches two of them agreed, in 17 three did, and in only 6 did all four put the same business first.

When they do agree, it can still be someone Maps puts far down the list. In Miami, all four assistants named T&S Roofing Systems first. Maps had it at number 9, with 1,264 reviews; the Maps top three held 5, 122 and 3 reviews.

Agreement is possible. In Ottawa, three of the four named John The Plumber first, a firm Maps placed at number 5. But it is the exception.

Why do AI and Maps pick different businesses?

We can't see inside either system, so this part is what the data suggests, not a proven cause.

Google says its local results weigh three things: relevance (how well a listing matches the search), distance (how far it is from the searcher) and prominence (how well known the business is, built from many signals: links and mentions across the web, its website, reviews, and more). In our searches, the Maps top 3 often went to businesses that were close to the city centre and whose names matched the search word for word:

• 148 of the 600 Maps top-3 spots were held by a business whose name contains both the city and the trade.
• 130 of the 600 were held by a listing with fewer than 20 Google reviews.
• 74 of the 600 had no website listed.

The assistants seemed to lean the other way, towards businesses that are talked about. The typical (median) AI first pick that did appear in the Maps top 20 had 710 Google reviews, against 128 for the typical (median) business in the Maps top 3.

A note on that number. We use review counts here because they are public and countable, as one readable stand-in for how well known a business is. They are not the whole picture, and they do not decide rankings on their own. Collecting reviews alone is not a plan for either Maps or AI answers.

Where do AI assistants get the names they recommend?

Three of the four assistants showed the sources behind their answers, and the same kinds of sites came up again and again: "best of" lists, directories and review sites.

• Perplexity cited prosgrade.com in 61 of 200 answers, consumeraffairs.com in 60, forbes.com in 48, thumbtack.com in 47, expertise.com in 42 and cityvetted.com in 41.
• ChatGPT cited angi.com in 22 answers, bestinhood.com in 21, consumeraffairs.com in 19, contractorsranked.com in 17 and healthgrades.com in 16.
• Google AI Mode mostly linked to Google's own pages (157 of 200 answers), alongside Yelp, Reddit, Angi and Thumbtack.
• Gemini sent its links through a Google redirect, so the sites behind its answers could not be seen.

That is the most useful finding for an owner. The assistants build their answer from what other websites say about you. Being listed, and described well, on the sites an assistant reads for your trade is something you can actually work on.

Does it differ by trade and city?

Yes, a lot. Plumbers were the trade where AI and Maps overlapped most: 49% of AI first picks were somewhere in the Maps top 20. Pest control came next at 43% and electricians at 41%. Dentists and HVAC were lowest, at 18% each.

Our guess, and it is only a guess: dentists and HVAC firms have crowded Maps results in big cities, and the assistants lean heavily on health and home directories for them (Healthgrades, Zocdoc, Angi), which rank businesses their own way.

By city, Philadelphia showed the most overlap (48% of first picks in the Maps top 20), followed by Ottawa and Vancouver (45% each). Houston, Phoenix and Atlanta showed the least, at 13% each.

Is the AI's "best" the same tomorrow?

Often not. We asked every question again 24 to 25 hours later. The first business named was the same in only 330 of 782 readable pairs: 42%.

• Perplexity kept its first pick in 65% of searches (130 of 200).
• Gemini kept it in 41% (82 of 199).
• Google AI Mode kept it in 34% (63 of 184).
• ChatGPT kept it in 28% (55 of 199).

The gap with Maps did not move, though. On day two, 6.8% of first picks were in the Maps top 3 (54 of 790) and 32.2% in the top 20 (254 of 790), against 6.7% and 31.7% the day before. The names change; the distance from Maps stays.

For an owner, that cuts both ways. Asking ChatGPT once about your business tells you very little. But it also means the first spot is not locked: the businesses that show up across many answers, day after day, are the ones the assistants keep finding good things written about.

What can a business owner do about it?

None of this replaces the work that gets you onto the map. It adds a second front. Here is what we would do, in plain steps:

1. Keep the Maps work going. The map is still one of the main places customers find you. Our guide to ranking higher on Google Maps walks through it, and our Google Maps service is built for it.
2. Ask the assistants yourself. Open ChatGPT, Perplexity, Gemini and Google's AI Mode and type the question a customer would: "Who is the best [your trade] in [your city]?" Do it on three different days and write down who comes up. That is your starting point.
3. Look at the sources they show. Three of the four assistants we tested list the websites behind their answer. If the same two or three directories or "best of" lists keep appearing for your trade, make sure your business is on them, with the same name, address and phone number you use everywhere else.
4. Earn mentions on local sites. A write-up in the local paper, a place on a city "best of" list, a link from a trade association: these are the things assistants read. Our link building team works on exactly this.
5. Say plainly on your own website what you do, where, and for whom. A clear page for each service and each area you cover, with proof of your work, gives assistants something specific to quote. Our on-page work and local SEO service cover this.
6. Treat reviews as one signal among many. Keep asking happy customers, and keep going beyond them.

If you'd rather have someone do this for you, our AI search work covers ChatGPT and Google's AI answers alongside the map, and our explainer on Google's AI answers shows how they read the web. We work with plumbers, electricians, roofers, HVAC companies, dentists, law firms, pest control companies, landscapers and chiropractors, in cities including Toronto, Austin, Chicago and Vancouver. You can see the results we've delivered and what it costs. Want to see whether the assistants name you or a competitor? Book a free call and we'll run the same check for your trade and city with you.

Can I use this data?

Yes. Journalists, researchers and anyone writing about local search are welcome to quote the findings and reuse the charts, with a credit and a link to this page: "Source: Awkward Media Local Search Index 2026". For the full tables by trade and city, contact us through our contact page.