Local AI in 2027: nine predictions from four editions of AI Local data
What will AI recommendations for local businesses look like in 2027? We have asked 12 AI assistants about local businesses every month since August 2026, in up to 4,100 US markets. This paper sets out what that record shows and makes nine predictions for November 2027, each with a number we can be marked against.
- ChatGPT
- Claude
- Gemini
- Perplexity
- Google AI Overviews
- Microsoft Copilot
- Grok
- Meta AI
- DeepSeek
- Mistral
- Siri
- Alexa
Across four editions of AI Syndicate Local (August to November 2026), AI's #1 local business changed in 45.3% of 1,400 US markets in one month, and only 17% of the first edition's #1s held the place in every edition. All 4,100 of November's #1s were named by a live-search assistant. For November 2027, AI Syndicate Research predicts the monthly change stays between 35% and 55%, fewer than 10% of today's #1s hold all year, live search spreads to at least two more assistants, and national chains stay #1 in 12% to 20% of markets. These are predictions, not measurements.
Executive summary
Source: AI Syndicate Research, Local AI in 2027, November 2026 (aisyndicate.com)
What this paper is
Most forecasts about AI and local business are opinion. This one starts from measurements. Since August 2026 we have asked 12 AI assistants the same three questions about local businesses every month, and the index has grown from 105 markets in one state to 4,100 markets in all 50. Below is what four editions of that data show, and nine predictions for what the November 2027 edition will show. Each prediction has a number, a confidence level and a test for being wrong.
Measurements and predictions are kept apart. Everything under "Measured" is something we counted. Everything under "Our call" is our judgment about the future, and may be wrong. We will mark every prediction, hit or miss, on this page after the November 2027 edition.
The nine predictions
| Prediction | Measured in 2026 | Our call for November 2027 | Confidence |
|---|---|---|---|
| 1. AI's #1 keeps changing every month | The #1 changed in 45% of 1,400 markets from October to November 2026 | From October to November 2027 the #1 changes in 35% to 55% of markets | High |
| 2. Few of today's #1s last the year | 32% of August's #1s were still #1 three editions later; 17% held every edition | 10% to 30% of November 2026's #1s are #1 in November 2027, and fewer than 10% hold #1 in every edition between | Medium |
| 3. A clear lead and broad naming still decide who holds | An October #1 that led by 10 or more reach points held 70% of the time; one that led by under 2, 29% | The same two groups stay at least 25 points apart in the October to November 2027 comparison | High |
| 4. Live search decides who is #1 | All 4,100 of November's #1s were named by a live-search assistant; 49% were named by no memory-only assistant | At least two of the four memory-only assistants answer our questions with live search, and 99% or more of #1s are still named by a live-search assistant | Medium |
| 5. Wrong phone numbers fall but do not disappear | 49% of the phone numbers in AI's local answers came from memory-only assistants, which matched the business's website 23% of the time; live search matched 87% | Under 30% of phone numbers come from memory-only assistants, and the live-search match rate is between 85% and 93% | Medium |
| 6. National chains do not take over | A national chain was AI's #1 in 15.2% of 4,100 markets | A national chain is #1 in 12% to 20% of the same markets, by the same rule | High |
| 7. There is still no single #1: the question splits the market | 23% of markets had the same #1 for the best, recommended and affordable questions (22% in October) | Under 30% of markets have the same #1 for all three questions | High |
| 8. Markets stay close | A typical #1 had 35.9 of a possible 100 reach points, and led by under 5 points in 32% of markets | The typical #1 stays under 45 reach points, and at least 25% of markets are led by under 5 points | Medium |
| 9. Every assistant changes underneath the rankings | Between October and November 2026 we changed how we reach six of the twelve assistants, because their models or search methods changed | All twelve assistants run a different model version or search method in November 2027 than in November 2026 | High |
What it means for a local business
- Being AI's pick is not a prize you win once. Nearly half of #1s change in a month, and we expect that to continue through 2027. Plan to check monthly.
- The assistants that search the web decide who is #1. They carry most of the audience, and no business reached #1 in November without one. Your own website and your listings are what they read.
- A wide lead is the best protection. #1s named by many assistants, well ahead of #2, mostly hold. Narrow #1s mostly do not.
- Most local businesses compete with other local businesses, not chains, and we expect that to stay true.
Scope. Four editions of AI Syndicate Local: 17 August, 1 September, 1 October and 1 November 2026. The November edition covers 4,100 markets (205 metros in all 50 US states, 20 types of business), 12 assistants and 3 fixed questions each. Four months is a short record, and the method section sets out what that limits.
What data are the predictions based on?
AI Syndicate Local is our public index of which local businesses AI assistants recommend. Each month we ask 12 assistants the same three questions about a type of business in a metro area: the best one, the one to recommend to someone who needs one, and an affordable one nearby. A market is one metro area and one type of business, so 36 answers go into each market.
- Live web search: ChatGPT, Claude, Gemini and Perplexity answer with web search on. Google AI Overviews is the overview Google shows for the question. Microsoft Copilot answers come mostly from Microsoft's Bing-grounded AI service.
- Model memory only: Grok, Meta AI, DeepSeek and Mistral answer from what the model learned, because that is the access available to us for them today.
- Reconstructed voice answers: Siri and Alexa cannot be asked programmatically. We reconstruct their answers from the kinds of sources each draws on.
How #1 is decided. Each market is ranked by AI reach. Every answer counts by how many people use that assistant and by how high in the answer the business was named. A score of 100 would mean being named first by every assistant on every question. No assistant counts for more than 30% of a market. The weights and their sources are on the methodology page. Every edition since August has been re-scored this way, so the four editions here are ranked by one rule.
Show the numbers
| Edition | Markets | States |
|---|---|---|
| August | 105 | 1 |
| September | 320 | 2 |
| October | 1,400 | 14 |
| November | 4,100 | 50 |
Source: AI Syndicate Research, Local AI in 2027, November 2026 (aisyndicate.com)
The index grew quickly, so most comparisons over time rest on the markets that were in both editions: 105 from August to September, 320 from September to October and 1,400 from October to November. Where we follow the same markets throughout, we say so.
How long will a business stay AI's #1 in 2027?
Three of the nine predictions are about how long a business stays AI's first pick.
Prediction 1: AI's #1 keeps changing every month
From October to November 2026, AI's #1 business changed in 45.3% of 1,400 markets (95% range 42.8–48.0%). The two earlier comparisons were higher, but they covered far fewer markets.
Show the numbers
| Editions compared | #1 changed | Markets |
|---|---|---|
| August to September | 57.1% | 105 |
| September to October | 52.8% | 320 |
| October to November | 45.3% | 1,400 |
Source: AI Syndicate Research, Local AI in 2027, November 2026 (aisyndicate.com)
The apparent fall from 57% to 45% is mostly the index growing. In the same 105 Florida markets followed since August, the #1 changed in 57%, then 59%, then 50%. With ranges about ten points either side, that is no clear trend. So we do not predict that rankings settle down. We predict the monthly rate stays roughly where it is.
A note on the figure. The November AI Local Report put the October to November change at 40%. It was written before the rankings were re-scored by AI reach, and it joined franchise locations by name. This paper uses today's published rankings and counts two entries as the same business only when the profile or the website is the same, which reads a few points higher.
Change differs by trade, and we expect the order below to hold broadly.
- Measured today. The #1 changed in 45% of 1,400 markets from October to November 2026.
- Our call for November 2027. From October to November 2027 the #1 changes in 35% to 55% of markets.
- Confidence. High.
- We are wrong if the share is below 35% or above 55%.
Prediction 2: Few of today's #1s last the year
The 105 Florida markets in the first edition are the longest record we have. Of the businesses that were #1 in August, 43% were #1 in September, 38% in October and 32% in November. Only 17% were #1 in all four editions (95% range 9.5–25.7%).
Show the numbers
| Edition | Still #1 |
|---|---|
| August | 100% |
| September | 43% |
| October | 38% |
| November | 32% |
Source: AI Syndicate Research, Local AI in 2027, November 2026 (aisyndicate.com)
The first month does most of the damage, and after that the losses slow. A #1 that had already held once held again 68% of the time; a new #1 held 36% of the time. Carrying those rates forward twelve months gives a small group of durable leaders and a large group that turns over, which is where our range comes from. Losing #1 is rarely a collapse: 82% of October's #1s were still in the top three in November and 93% in the top ten.
- Measured today. 32% of August's #1s were still #1 three editions later; 17% held every edition.
- Our call for November 2027. 10% to 30% of November 2026's #1s are #1 in November 2027, and fewer than 10% hold #1 in every edition between.
- Confidence. Medium.
- We are wrong if more than 30% are still #1, or 10% or more never lose it.
Prediction 3: A clear lead and broad naming still decide who holds
Two things about an October #1 went with holding the place in November: how far ahead of #2 it was, and how many assistants named it.
A #1 that led by 10 points or more held 70% of the time (95% range 66.3–73.1%). One that led by under 2 points held 29% (22.4–35.6%). The same pattern shows in how widely a business is named: a #1 named by 8 or more of the 12 assistants held 67% of the time, and one named by 4 or fewer held 29%.
This is the prediction we are surest of, because it follows from how a ranking built from many answers behaves: a wide lead takes many changed answers to overturn. It is an association, not proof that widening a lead causes a business to hold.
- Measured today. An October #1 that led by 10 or more reach points held 70% of the time; one that led by under 2, 29%.
- Our call for November 2027. The same two groups stay at least 25 points apart in the October to November 2027 comparison.
- Confidence. High.
- We are wrong if the gap is under 25 points.
How will the AI assistants themselves change in 2027?
Three predictions are about the assistants themselves, which changed more in 2026 than the businesses did.
Prediction 4: Live search decides who is #1
In the November edition the six live-search assistants carry 80% of the weight in a market's ranking, because they are the ones most people use. The four memory-only assistants carry 13% and the two voice assistants 6%.
The result is plain in the rankings. Every one of the 4,100 #1 businesses was named by at least one live-search assistant, and 49.4% were named by no memory-only assistant at all. ChatGPT named 95% of the #1s. Grok named 25%.
We expect the split between assistants that search and assistants that remember to narrow. In November 2026, Microsoft Copilot moved in our panel from reconstructed answers to answers grounded in live Bing results. We think at least two of the remaining four follow in 2027, because an assistant that cannot look things up gives worse local answers, as the next prediction shows. This call depends on what those companies release, which we do not control, so our confidence is medium.
- Measured today. All 4,100 of November's #1s were named by a live-search assistant; 49% were named by no memory-only assistant.
- Our call for November 2027. At least two of the four memory-only assistants answer our questions with live search, and 99% or more of #1s are still named by a live-search assistant.
- Confidence. Medium.
- We are wrong if fewer than two of the four have switched, or more than 1% of #1s are named by no live-search assistant.
Prediction 5: Wrong phone numbers fall but do not disappear
Our paper Can you trust the phone number AI gives you? checked the phone numbers assistants gave against the businesses' own websites. Numbers from live-search assistants matched 86.7% of the time. Numbers from memory-only assistants matched 23.3% of the time, and those assistants gave 49.2% of all the numbers. 20.9% of the web addresses they gave had no registered domain behind them.
If memory-only assistants gain live search, the share of unreliable numbers falls with them. But we do not expect live search to reach 100%. A number that is wrong on a directory, or out of date on the business's own site, is repeated faithfully by an assistant that searches. That part of the problem belongs to the business and its listings, and it will still be there in 2027.
- Measured today. 49% of the phone numbers in AI's local answers came from memory-only assistants, which matched the business's website 23% of the time; live search matched 87%.
- Our call for November 2027. Under 30% of phone numbers come from memory-only assistants, and the live-search match rate is between 85% and 93%.
- Confidence. Medium.
- We are wrong if memory-only assistants still give 30% or more of the numbers, or live search is outside 85% to 93%.
Prediction 9: Every assistant changes underneath the rankings
Between the October and November editions we had to change how we reach six of the twelve assistants, because their models or their search methods changed (November AI Local Report). That is one month.
For a business, this means a ranking can move with no change on its own side or its competitors'. For anyone measuring, it means month-to-month comparisons need care. We expect no assistant to get through a year unchanged, and we will keep recording which changed and when.
- Measured today. Between October and November 2026 we changed how we reach six of the twelve assistants, because their models or search methods changed.
- Our call for November 2027. All twelve assistants run a different model version or search method in November 2027 than in November 2026.
- Confidence. High.
- We are wrong if any assistant is unchanged.
What will a local market look like in AI answers in 2027?
The last three predictions are about the shape of a local market in AI's answers.
Prediction 6: National chains do not take over
Our paper Do AI assistants favor national chains? found a national chain at #1 in 15.2% of 4,100 markets, and no chain anywhere in the top 10 in 45.4%. The assistants that answer from memory named a chain first in 36% of their answers; the ones that search, in 14%.
Two forces pull in opposite directions in 2027. If memory-only assistants gain live search, they should name fewer chains. Against that, franchise networks are well placed to keep hundreds of local listings accurate. We expect them roughly to cancel, with chains strong in the same few trades (water damage restoration, window replacement, hotels, garage door repair) and rare elsewhere.
- Measured today. A national chain was AI's #1 in 15.2% of 4,100 markets.
- Our call for November 2027. A national chain is #1 in 12% to 20% of the same markets, by the same rule.
- Confidence. High.
- We are wrong if the share is below 12% or above 20%.
Prediction 7: There is still no single #1: the question splits the market
Ask AI for the best plumber, a plumber to recommend, and an affordable plumber, and you usually get three different first picks. Only 23.0% of markets had the same #1 for all three questions in November, and 22.3% in October (November and October AI Local Reports).
Nothing we see suggests the assistants are converging on one answer per market. A business should think of its place as a set of positions, one for each way a customer asks, not a single rank.
- Measured today. 23% of markets had the same #1 for the best, recommended and affordable questions (22% in October).
- Our call for November 2027. Under 30% of markets have the same #1 for all three questions.
- Confidence. High.
- We are wrong if 30% or more do.
Prediction 8: Markets stay close
A typical #1 in November scored 35.9 reach points of a possible 100. It was named by 7 of the 12 assistants, and led #2 by a median 8.5 points. The lead was under 5 points in 32.2% of markets and under 2 points in 14.5%.
No market is close to winner-take-all: the #1 holds about a third of what is possible. New names keep arriving too. 27.1% of the businesses in November's top tens were not ranked at all in the same market in October. We expect markets to stay this open, because twelve assistants reading different sources rarely agree on one business.
- Measured today. A typical #1 had 35.9 of a possible 100 reach points, and led by under 5 points in 32% of markets.
- Our call for November 2027. The typical #1 stays under 45 reach points, and at least 25% of markets are led by under 5 points.
- Confidence. Medium.
- We are wrong if the typical #1 reaches 45 or more, or fewer than 25% of markets are that close.
What can this data not predict?
Some of the biggest questions about 2027 are ones our data cannot answer. We would rather say so than guess.
- Paid placement in AI answers. We measure unpaid recommendations. If assistants begin to sell positions in local answers, the rankings we publish would describe only part of what a customer sees. We will report it when we see it.
- Assistants that book the job themselves. An assistant that calls, books or pays on the customer's behalf changes what being named is worth. Our index records who is named, not what happens next.
- Answers that differ for each person. We ask with the city in the question and no account history or device location. As assistants use more of both, two neighbours may get different picks. Our figures describe the answer a new user gets.
- Voice assistants. Our Siri and Alexa answers are reconstructions. If either becomes directly measurable in 2027, its figures will change for that reason alone.
- How many people ask AI at all. We weight assistants by survey data on who uses them. We do not measure how many local purchases start with an AI answer.
What should a local business do to be ready for 2027?
- Check your position monthly, not once. With close to half of #1s changing in a month, a single check tells you little. Look at the trend.
- Win the assistants that search. ChatGPT, Google AI Overviews, Gemini, Copilot, Claude and Perplexity carry about four fifths of the weight and decided every #1 in November. They read your website, your listings, review sites and local coverage.
- Make your phone number and web address easy to confirm. Live-search assistants got the number right 87% of the time. Much of the rest traces back to out-of-date listings. Put the number in plain text on your own site and keep every listing the same.
- Build a lead, not just a rank. A #1 ten points clear held 70% of the time; a #1 barely ahead held 29%. Being named by more assistants is how a lead is built.
- Cover each way customers ask. Best, recommended and affordable are three contests with three different winners in most markets.
- Do not wait for the memory-only assistants. They name what they learned in training. If they gain live search in 2027, as we expect, the same work that wins the search assistants will carry over.
These are patterns in what the assistants said, not guarantees.
How this data is used. This research feeds the AI Syndicate platform. Clients see where they stand in their market's AI Local ranking, which assistants name them and which name their competitors, inside their dashboard, and what we learn across the whole index is built into the twelve-assistant measurement the platform runs for each client's own business.
How will these predictions be checked?
A prediction that is never checked is worth nothing, so here is how these will be marked.
- When. After the November 2027 edition of AI Syndicate Local is published.
- How. Each prediction is marked hit or miss against the test written beside it, using the same definitions as this paper. The analysis behind this paper is saved so the same counts can be re-run.
- On what. The 4,100 markets of the November 2026 edition, wherever they are still measured. Markets added later are left out of the marking.
- Where. On this page, under each prediction, with the measured figure. Misses stay on the page.
- If the measurement changes. If we change how we rank or reach an assistant in a way that makes a test unfair in either direction, we will say so and mark that prediction as not comparable, not as a hit.
How was this measured, and what are its limits?
Data. AI Syndicate Local, four editions: 17 August 2026 (105 markets, Florida), 1 September (320 markets, 2 states), 1 October (1,400 markets, 14 states) and 1 November (4,100 markets, 50 states). 12 assistants, 3 fixed questions each, with the city named in the question and no user location or account history. A business is ranked when at least two independent answers name it, and each market's published ranking lists up to 25. The November edition also has early markets from Canada, the United Kingdom and six new trades; they are left out here.
Ranking. Every edition is ranked by AI reach (see the methodology page). Editions before November were first published under an earlier rule, by number of answers with every assistant equal, and have since been re-scored. Figures here use the re-scored rankings, so they differ from figures in reports written before the change.
Same business. Two entries in different editions count as the same business when they have the same profile on the index or the same website. Franchise locations listed under different names are not joined, so change is slightly overstated.
Weights. The share of a ranking carried by each type of assistant uses the index's US audience weights with the 30% cap, as of 11 November 2026.
Figures from other papers. Phone number accuracy, chain shares and question wording are quoted from the papers linked beside them, with their own methods and limits.
What this cannot tell you.
- Four editions over three and a half months is a short record. The longest series rests on 105 Florida markets.
- The predictions are judgment. They extend patterns in the data and are not the output of a statistical forecasting model.
- We changed how we reach several assistants during the period, so part of the measured change is ours and not the market's.
- The index measures what assistants say, not what customers do.
- Siri and Alexa answers are reconstructions.
- Links between a wide lead and holding #1 are associations and do not show cause.
What stays private. The public index publishes each market's rankings and sources. Per-assistant answer sets and per-business movement are not published.
Statistical notes
- Unit of measurement. Shares of markets count each market once.
- Ranges. 95% ranges come from resampling markets 2,000 times. They describe uncertainty across markets, not uncertainty about the future.
- Prediction ranges are not statistical intervals. They are the band inside which we commit to being marked right.
- Reach points run from 0 to 100, where 100 is being named first by every assistant on every question.
- Small groups. Trade figures rest on 70 markets each and are directional.
How can I cite or reuse these numbers?
Quote any figure in this report with credit to AI Syndicate Research and a link to this page. When you quote a prediction, please call it a prediction. You may quote up to 50 words of text, or share a single chart with its source line intact. Reproducing the report or large parts of it needs written permission. A suggested citation:
A PDF copy and a plain-text Markdown version are available. The rankings behind the report are at aisyndicate.com/ai-local/. Journalists, businesses and researchers who need a cut of the data that is not here can write to support@aisyndicate.com.
Key figures
Every key figure in this report, in one table. The same figures are in findings.json and findings.csv, free to quote with credit. Ranges are 95% intervals from resampling markets.
| Finding | Value | 95% range | Sample |
|---|---|---|---|
| Share of markets where AI assistants' #1 local business changed from October to November 2026, ranked by AI reach | 45.3% | 42.8–48% | 1,400 markets |
| Share of markets where AI assistants' #1 local business changed from September to October 2026, ranked by AI reach | 52.8% | 47.5–58.1% | 320 markets |
| Share of August 2026 #1 businesses that were #1 in November 2026 | 32.4% | 23.8–41% | 105 Florida markets |
| Share of markets with the same #1 in every edition from August to November 2026, ranked by AI reach | 17.1% | 9.5–25.7% | 105 Florida markets |
| Share of October #1s leading #2 by 10 or more reach points that were still #1 in November | 69.9% | 66.3–73.1% | 707 markets |
| Share of October #1s leading #2 by under 2 reach points that were still #1 in November | 28.7% | 22.4–35.6% | 174 markets |
| Share of October #1s named by 8 or more of 12 assistants that were still #1 in November | 67.4% | 63.4–71.7% | 470 markets |
| Share of October #1s still in the top three in November, ranked by AI reach | 81.5% | 79.5–83.6% | 1,400 markets |
| Share of November 2026 #1 businesses named by at least one live-search AI assistant | 100% | 4,100 markets | |
| Share of November 2026 #1 businesses named by no memory-only AI assistant | 49.4% | 47.9–50.9% | 4,100 markets |
| Share of a market's AI Local ranking carried by the six live-search assistants | 80.5% | US audience weights, 11 November 2026 | |
| Median AI reach of the #1 business in a market, out of 100 | 35.9 points | 4,100 markets (median) | |
| Share of markets where AI's #1 leads #2 by under 5 reach points | 32.2% | 30.8–33.7% | 4,100 markets |
| Share of November top-10 businesses that were not ranked in the same market in October | 27.1% | 13,996 placings, 1,400 markets |
Sources
- AI Syndicate Local: the August to November 2026 editions, AI Syndicate. Read Nov 17, 2026.
- AI Syndicate Local methodology: how the index is measured and how much each assistant counts, AI Syndicate. Read Nov 17, 2026.
- Can you trust the phone number AI gives you? (November 2026), AI Syndicate Research. Read Nov 17, 2026.
- Do AI assistants favor national chains? (November 2026), AI Syndicate Research. Read Nov 17, 2026.
- AI Local Report, November 2026, AI Syndicate Research. Read Nov 17, 2026.
Questions about local AI in 2027
What will AI recommendations for local businesses look like in 2027?
AI Syndicate Research predicts they stay unsettled and open. In 2026, AI's #1 local business changed in 45% of 1,400 US markets in one month, and we expect 35% to 55% a month in 2027. We also expect more assistants to answer with live web search, national chains to stay #1 in 12% to 20% of markets, and no single #1 across different ways of asking. These are predictions based on four months of data.
How often does AI's top local recommendation change?
Often. From October to November 2026 it changed in 45.3% of 1,400 US markets. Of the businesses that were #1 in August 2026, 32% were #1 in November and 17% were #1 in every edition between.
Which AI assistants matter most for a local business?
The ones that search the web: ChatGPT, Google AI Overviews, Gemini, Microsoft Copilot, Claude and Perplexity. They carry about 80% of the weight in AI Local rankings, and every one of the 4,100 #1 businesses in November 2026 was named by at least one of them.
What helps a business stay AI's #1 pick?
A wide lead and being named by many assistants. An October 2026 #1 that led by 10 or more reach points was still #1 in November 70% of the time, against 29% for one that led by under 2 points. This is an association, not proof of cause.
Will these predictions be checked?
Yes. Each has a number and a test for being wrong. AI Syndicate Research will mark all nine, hit or miss, on the paper's page after the November 2027 edition of AI Syndicate Local.
Related pages.
See where your business stands before 2027.
The AI Local index is free and public: pick your state, metro and trade to see every ranked business and which of the 12 assistants named it. Or run the free check to see where your own business stands.