How real estate agents can stay ahead of AI: what 12 AI assistants said in 1,582 answers about agents

Summary

Key findings

1. AI rarely agrees on who the good agents are. Across 50 metro areas, the 12 assistants named 3,967 different agents, teams and brokerages. 73% were named by only one assistant. The 9% named by four or more took 34% of all picks.
2. Brands get named more than people. National brokerage brands and their local offices were 29% of picks, teams 19% and local brokerages 18%. Individual agents by name were 15%. Assistants answering from memory named a brand for 40% of their picks; Grok for 77%.
3. Individual agents rarely make the shared list. 5% of the individual agents and teams AI named were named by four or more assistants, against 17% of the national brands.
4. "Affordable" hands the answer to discount services. For "affordable" questions, 46% of picks were discount or online services such as Redfin, Clever, Houzeo and UpNest, and 86% of the answers named at least one. For "best", under 1% of picks were.
5. AI describes agents by what they sell and to whom. 25% of picks came with a niche (luxury, first-time buyers, relocation, investors), 21% with reviews, 18% with commission or fees, 14% with a sales record and 9% with a ranking or award.
6. The agents AI agrees on have a track record in public. For agents and teams named by four or more assistants, picks cited a ranking or award 1.9 times as often, a sales record 1.7 times as often and reviews 1.4 times as often as for those only one assistant named. Commission talk was the same for both.
7. AI sends people to Zillow. Zillow came up in 66% of answers and Realtor.com in 49%; in the answers that named no agent, 86% and 82%. Assistants answering from memory named Zillow in 97% of answers.
8. What AI reads is mostly not agents' own sites. Of 3,849 sources cited, 40% were agent-matching, ranking and discount sites (FastExpert, Clever, HomeLight, EffectiveAgents, RealTrends, AgentPronto) and 20% listing portals. Agents', teams' and local brokerages' own sites were 17%; brokerage brand sites 2%.

What it means. When a buyer or seller asks AI for an agent, a name only travels across assistants when it comes with public proof: a sales record, a ranking such as RealTrends, a stack of reviews on Zillow and other sites. Those are things an agent can build and publish. When the question is about price, the answer is about commission, and discount services fill the list unless an agent states plainly what they charge. The last section turns the findings into a plan.

Scope. The October 2026 edition of AI Syndicate Local: 50 metro areas in 9 states, three fixed questions about real estate agents to each of 12 assistants, asked between 1 and 7 October 2026. This paper reads all 1,582 answers that came back and the 7,832 recommendations in them. Method and limits are at the end.

What did we read, and how?

Each month AI Syndicate Local asks 12 AI assistants three questions about a type of business in a metro area: who are the best, who do you recommend, and a good affordable one nearby. For real estate it reads like "I need a real estate agent in Mesa, AZ. Who do you recommend?" In October 2026 we asked in 50 metro areas across Alabama, Alaska, Arizona, Arkansas, California, Florida, Georgia, New York and Texas.

1,582 answers came back. This paper reads what the assistants said: whether they named anyone, who they named, the reasons they gave, the advice they added and the sites they pointed to. We call each agent, team, brokerage or service an assistant names a pick. We sorted each pick by what it is:

We then took the words describing each pick, from its name to the next one named, and sorted them into kinds of reason:

The twelve fall into three groups by how they answer:

• Web search: ChatGPT, Claude, Gemini and Perplexity answer with web search on. Google AI Overviews is the AI Overview Google shows for the question. Microsoft Copilot is the Copilot answer Bing shows; when Bing shows none, we reconstruct one from the same Bing results.
• Model memory: Grok, Meta AI, DeepSeek and Mistral answer from what the model learned, because no consumer live-search access is available for them.
• Voice: Siri and AAlexa cannot be asked programmatically, so we reconstruct their answers in the short, spoken style each uses. A typical Siri answer about an agent is under 50 words and names one or two.

A reason in this paper is one the assistant wrote, not proof of why it picked the agent. The limits are set out at the end.

Will AI name a real estate agent?

Usually. 13% of answers named no agent, team or brokerage. Whether an agent can be named at all depends mostly on the assistant.

Web-search assistants named no one in 3% of answers, memory assistants in 22% and the voice assistants in 23%. The wording matters: "Who do you recommend?" got no names in 22% of answers, against 8% for "Who are the best?" and 10% for an affordable agent. Like with lawyers, recommending one person is where some assistants stop.

When an answer names no one, it sends the buyer or seller to the portals. Of the 213 answers that named no agent, 86% pointed to Zillow, 82% to Realtor.com, 47% to Google and 40% to the state's licence lookup.

Who does AI name: agents, teams or brands?

Mostly brands and teams. Of the 7,832 picks, 29% were national brokerage brands or their local offices, 19% teams, 18% local brokerages and 14% discount or online services. Individual agents named on their own were 15%.

How an assistant answers decides who it names. Assistants answering from memory named a brand for 40% of their picks and an individual agent for 8%; Grok named a brand for 77%. Web-search assistants spread their picks evenly across brands (22%), local brokerages (22%), teams (20%) and individual agents (19%). The voice assistants named individual agents most (28%), usually one agent with one reason.

Share of each group's picks, October 2026. Local brokerages make up most of the rest: 22% for web search, 13% for memory and 18% for voice.

A brand name is easy for an assistant to say and hard to get wrong, which may be why the memory assistants fall back on it. For an agent, it means the brokerage's name on the door can be named in place of the agent who would do the work.

Do the AI assistants agree on which agents to name?

Very little. Across the 50 metros, the 12 assistants named 3,967 different agents, teams, brokerages and services, a median of 78 per metro. 73% were named by only one assistant. 18% were named by two or three, and 9% by four or more.

The 9% named by four or more assistants took 34% of all picks. Every metro had at least one.

Who makes that list. 17% of the national brands named in a metro were named by four or more assistants, and 12% of the discount and online services. For local brokerages it was 8%; for teams and individual agents, 5%. Redfin was named in 49 of the 50 metros, Clever and Houzeo in 48, Keller Williams in 47 and Coldwell Banker in 44. The local agent a buyer would actually work with is the hardest name for AI to agree on.

What does AI say about the agents it recommends?

Often, but less than for other trades. 62% of picks came with at least one reason; the rest were a name, sometimes with a brokerage or a phone number.

Web-search assistants lean on evidence they can read: reviews (30% of picks against 9% for memory assistants) and sales records (18% against 8%). They also quote numbers: a star rating for 13% of picks and a review count for 14%. Memory assistants describe a niche (29%) and talk about commission (19%).

ChatGPT is the most specific. It cited reviews for 55% of its picks, a sales record for 37%, a review count for 30% and a star rating for 29%, and named Zillow reviews for 16%. Perplexity cited reviews for 39% and a review count for 30%. Gemini describes marketing (17%) and neighborhoods. Alexa, when it names someone, usually says they have strong Zillow reviews (12% of its picks).

What do the agents most assistants agree on have in common?

We compared what the assistants wrote about the 352 agents, teams and brokerages named by four or more assistants with the 2,912 only one assistant named, pick by pick, so that being named more often does not give more chances to collect a reason.

The names several assistants agree on come with a public track record: a ranking or award 1.9 times as often, a sales record 1.7 times, years in business and local knowledge 1.6 times, and reviews 1.4 times as often. Commission and fees come up equally for both: price does not separate them.

79% of the names given by four or more assistants were named by at least one web-search assistant and one memory assistant. Being on the shared list means being easy to find today and already written about widely.

These are stated reasons, not proof of what caused the agreement. But the pattern is the same one we found for lawyers: what separates the names AI agrees on is evidence that lives outside the agent's own marketing, such as RealTrends rankings, closed sales on the portals and reviews on several sites.

What changes when someone asks for an affordable agent?

Everything. The "affordable" question turns the answer into a conversation about commission, and the list into one of discount services.

For "affordable" questions, 46% of picks were discount or online services and 58% came with commission or fees as the reason. Sales records fell from 19% of picks for "best" to 5%, rankings and awards from 14% to 2%, and individual agents from 19% of picks to 7%.

The answers explain how commission works. Of the "affordable" answers, 86% named at least one discount or online service, 69% gave a commission percentage, 60% told the person to ask about or compare commission and 54% said commission is negotiable. 15% mentioned buyer agreements or the 2024 commission rule changes. Redfin came up in 59% of these answers and Clever in 42%.

Every assistant but one did this. Discount or online services were 58% of Gemini's and Microsoft Copilot's "affordable" picks, 57% of Siri's, 54% of Perplexity's and DeepSeek's, and 35% of ChatGPT's and Claude's. Google AI Overviews was the exception, at 9%.

For a traditional agent, this is the question where they are most likely to be replaced by a service. An agent who offers a reduced listing fee, a rebate or a clear menu of services, and says so in plain words, gives the assistant something to say other than "try Redfin".

Which sites does AI send people to, and which does it cite?

Two lists matter: the sites an answer tells people to check, and the sources a web-search answer cites.

Sites named in the answer. Zillow leads by a wide margin.

Memory assistants name the portals almost every time: Zillow in 97% of their answers, Realtor.com in 89% and Google in 75%. Asked for "the best", 29% of answers named RealTrends, the ranking of agents by sales.

Sources cited. Eight of the assistants cite sources: ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews, Microsoft Copilot and our reconstructions of Siri and Alexa. The four memory assistants cite nothing. Across 3,849 citations in 866 answers:

The sources cited most were Zillow, FastExpert, Clever, Realtor.com, HomeLight, EffectiveAgents, Homes.com, RealTrends and AgentPronto. Brokerage brand sites were cited rarely (2%), even though brands were 29% of picks. The assistants differ sharply:

Share of each assistant's citations in answers about real estate agents, October 2026. Google AI Overviews cited reviews and social sites for 31% and ChatGPT cited government sites (licence records) for 10%. Microsoft Copilot cited too few sources here (6) to compare.

An agent's own website is a main source for Gemini and ChatGPT. For Claude and Perplexity, an agent mostly exists through profiles on the matching and ranking sites, and for everyone the Zillow and Realtor.com profiles carry weight.

What advice does AI give people choosing an agent?

Most answers coach the client. 49% told people to interview more than one agent, 28% to check reviews, 23% to ask about or compare commission, 21% said commission is negotiable and 18% to check the agent's licence.

The memory assistants and Microsoft Copilot coach hardest. Grok told people to interview several agents in 96% of its answers and Copilot in 89%. Grok told people to check the licence in 62%. Siri and Google AI Overviews almost never added advice.

For an agent, this is the first meeting. A seller who found you through AI has often been told to interview three agents, compare commission and look you up on Zillow and the state licence site. Agents who make those answers easy to find, with a clear listing presentation, a stated fee and current reviews, match what the client has been told to ask.

How can a real estate agent stay ahead of AI?

These are patterns in what 12 assistants wrote across 50 metro areas, not a guaranteed way to be recommended. With that limit, the findings point to where an agent or team can start. Each step is tied to the finding behind it.

1. Measure every assistant, not one. 73% of names are given by only one assistant, and Grok, Alexa and DeepSeek often name no one. Ask each assistant the questions your clients ask, for your city, and track the answers over time.
2. Make your sales record public and specific. Sales records and rankings separate the agents AI agrees on. Publish homes sold, volume and years in plain text, and make sure your RealTrends and portal records show them.
3. Get your own name next to your brokerage's. Brands are 29% of picks and individual agents 15%, and only 5% of individual agents make the shared list. Use your own name on your site, profiles and listings, alongside the brokerage, so assistants can name the person and not only the logo.
4. Build reviews where AI reads them. Zillow is named in 66% of answers, and ChatGPT and Alexa cite Zillow reviews directly. Keep reviews coming on Zillow, Realtor.com and Google, and keep your star rating and count consistent.
5. Complete the matching and ranking profiles. For Claude and Perplexity, more than 60% of citations are agent-matching, ranking and discount sites such as FastExpert, HomeLight, EffectiveAgents and RealTrends. A complete, accurate profile there is how those assistants know you.
6. Say what you charge. For "affordable" questions, 58% of picks come with a commission reason and 46% are discount services. If you offer a reduced listing fee, a rebate or a flat fee, publish it. If you don't, publish what your fee buys.
7. Name your niche. Niche is the most common reason AI gives (25%) and more common still for the agents AI agrees on (30%). Luxury, first-time buyers, relocation, military, investors, new construction: state it on every profile.
8. Write about your neighborhoods. Local knowledge is cited 1.6 times as often for the agents AI agrees on. Pages about the neighborhoods you sell in, with recent sales, give assistants something local to repeat.
9. Keep your licence record current. The state licence lookup is named in 24% of answers and in 40% of the answers that name no one. Make sure your name, brokerage and status match there.
10. Prepare for a coached client. Half of the answers told people to interview several agents and many to ask about commission. Have your listing presentation, fee and reviews ready before the first call.

How this data is used. This research feeds the AI Syndicate platform, so our clients have the leading edge. Agents and teams on the platform see where they stand in their metro's AI Local ranking and how they have moved, inside their dashboard, and their markets are measured more often than the rest of the index. What we learn across the whole index, such as which reasons the assistants give, which sites they cite and how each assistant answers questions about agents, is built into the twelve-assistant measurement the platform runs for each client's own market and questions.

How was this measured, and what can it not tell you?

Method. The October 2026 edition of AI Syndicate Local asked 12 assistants three fixed questions about real estate agents in each of 50 metro areas, with the city named in the question and no user location or account history. A model extracts the names from each answer. Each pick's description runs from its name to the next name, ending at a new heading. We sorted each name into a kind by word lists for brands, teams, brokerages and services, matched the description against fixed word lists for each kind of reason, and the whole answer against word lists for advice and sites. We checked a random sample of matches for each kind by hand and tightened lists that caught unrelated text.

How the twelve are reached. ChatGPT, Claude, Gemini and Perplexity answer with web search on. Google AI Overviews is the AI Overview Google shows. Microsoft Copilot is the Copilot answer Bing shows, reconstructed from the same Bing results when Bing shows none; in this edition every Copilot answer was a reconstruction. Siri and Alexa are reconstructions in the short spoken style each uses. Grok, Meta AI, DeepSeek and Mistral answer from model memory, because we cannot reach their live search; their consumer apps may answer differently.

What this cannot tell you.

• A reason in this paper is one the assistant wrote. Models do not reliably report why they chose something, so these are stated reasons, not causes.
• Siri, Alexa and Microsoft Copilot are reconstructions. Read them as close approximations, not recordings.
• Sorting names into kinds is done by word lists. A person named with their brokerage ("Jane Smith, Keller Williams") counts as a brand pick when the extracted name includes the brokerage, and a team named after a person counts as a team.
• Names are matched within a metro by normalised text. Two spellings of one name, or an agent and their team, can count as two, which makes agreement look a little lower than it is.
• Word lists miss reasons phrased in unusual ways and occasionally catch text that is not a reason. "Niche" and "size" are the broadest lists.
• We did not check whether the sales figures, ratings, commission rates or licence details the assistants quoted are correct.
• Google AI Overviews (76 of 150 answers) and Microsoft Copilot (75) returned fewer usable answers, with none in some states; Claude (124) and Alexa (120) also had gaps. Figures for those assistants rest on fewer metros.
• Each question is asked once a month, so a single answer can vary. The figures are averages over many answers, not predictions for any one.

What stays private. The public index publishes each market's rankings and the assistants that named each business. The individual answers are not published.

Statistical notes

• Unit of measurement. Shares of picks pool every pick by the 12 assistants across all metros. Shares of answers pool every answer. A market is one metro area.
• Ranges. 95% ranges come from resampling the 50 metros 1,000 times. For example: names given by only one assistant, 73.4% (72.1–74.8%); by four or more, 8.9% (8.1–9.7%); their share of picks, 33.7% (31.2–36.4%); picks that are national brands, 29.1% (27.0–31.5%); individual agents, 14.8% (10.9–18.8%); discount or online services among "affordable" picks, 46.4% (41.4–50.8%); answers naming no one, 13.5% (11.9–15.0%).
• Missing answers. A question that returned no answer is left out, not counted as naming no one.
• Agreement. A name's count is the number of different assistants that gave it in that metro, across the three questions. Names were matched after ignoring case, punctuation, "&" and "and", and the words team, group, realty, real estate, realtors, homes, properties, brokerage and legal suffixes.
• Consensus comparison. Shares are of picks, so each mention counts once.
• Citations. A citation counts as an agent's own site when its domain contains a distinctive word from a name given in the same metro, or a real estate word, and is not a known portal, matching, ranking, review, brokerage brand or government site. Dictionary pages are left out.

How to cite this report

Quote any figure in this report with credit to AI Syndicate Research and a link to this page. 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 published under a CC BY 4.0 licence at aisyndicate.com/ai-local/, with machine-readable copies at /ai-local/data.json and /ai-local/all.md. Journalists, brokerages 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 reuse under a CC BY 4.0 licence. Ranges are 95% intervals from resampling markets.

Sources

1. AI Syndicate Local: the October 2026 edition — AI Syndicate. Read Oct 11, 2026.
2. AI Syndicate Local methodology: how the index is measured — AI Syndicate. Read Oct 11, 2026.
3. AI Syndicate Local, machine-readable rankings with the assistants that named each business (CC BY 4.0) — AI Syndicate. Read Oct 11, 2026.