Start with the thing most agent marketing refuses to accept.
You are not going to win "homes for sale in [your city]". The portals have spent two decades and enormous sums owning that query, they hold the inventory data, and an answer engine looking for listings will go where the listings are. Every hour spent trying to out-rank them on property search is wasted.
The good news is that the property was never what you were selling.
The agent is the product
When someone asks an AI about buying or selling, the questions that actually matter to your business are not about houses:
- how do I choose a real estate agent
- what commission do I pay now
- should I sell before I buy
- what is [neighbourhood] actually like to live in
- is now a good time to sell in [city]
None of those are inventory questions, which means none of them are questions the portals answer well. They are judgement questions, and judgement is what a local agent has that a national database does not.
That reframes the whole content problem. Stop publishing listings that Zillow already has. Start publishing the reasoning a good agent provides.
The commission question changed, and the web has not caught up
The single largest opportunity in this vertical right now is a stale corpus.
Following the National Association of Realtors settlement, effective 17 August 2024, the rules changed materially:
- MLS systems removed buyer agent compensation fields. A listing agent can no longer advertise buyer broker compensation through the MLS.
- Sellers are no longer automatically responsible for paying the buyer's agent.
- Buyers working with an MLS participant generally need a written agreement before touring homes, and that agreement must spell out compensation.
- Compensation is fully negotiable and handled deal by deal. In practice most sellers still offer something, commonly around 2% to 2.5%, down from the 2.5% to 3% that was typical when it was posted in the MLS, because refusing narrows the buyer pool. But buyers should be prepared to pay their agent directly if no concession is negotiated.
Meanwhile an enormous volume of web content still describes the old arrangement as current, and models grounded on that corpus repeat it. Consumers are being told, confidently, that the seller always pays the buyer's agent.
Write the accurate version for your market. Date it explicitly, state what changed and when, explain what it means for a buyer and for a seller separately, and link the primary source. Correcting a widespread and consequential error is one of the most reliable ways to be cited, and this is the biggest such error available in real estate.
Hyperlocal is the one thing a portal cannot synthesise
Portals have data. They do not have knowledge.
They can tell you a neighbourhood's median price and school ratings. They cannot tell you which streets flood, which blocks get noise from the freeway on southerly winds, where the elementary catchment boundary actually cuts, which new development is going to change the commute in two years, or why the corner that looks marginal on a map is the one everybody wants.
That is the content moat, and it is the last one agents have. A genuine neighbourhood guide, written by somebody who has sold there, is:
- Specific in a way no national dataset can be
- Impossible to generate credibly without local experience
- Exactly the kind of source an answer engine reaches for when asked a question about a place
Write one per neighbourhood you actually work. Include the unglamorous specifics: parking, bin collection, which side of the street gets afternoon sun, what a typical HOA covers, how the commute really behaves at 8am. The temptation is to write tourism copy. Resist it. The value is entirely in the detail a brochure would omit.
Own "how to choose an agent"
This query is asked precisely because the asker expects every agent to answer it self-servingly. Which is what makes an honest answer disproportionately valuable.
A page that sets out the questions a seller should ask any agent, including ones that would be awkward for you, earns the trust the question was testing for. Real criteria: transaction volume in that specific neighbourhood rather than citywide, average days on market versus the area median, list-to-sale ratio, whether the person you meet is the person who does the work, what happens if you want to cancel.
Publish your own numbers against those criteria. An agent who states their actual list-to-sale ratio and neighbourhood transaction count is making a checkable claim, which is both more persuasive to a human and far more usable to a model than "top producer" or "your neighbourhood expert".
What to stop doing
- Listing syndication as a content strategy. The portals win that. Republishing IDX feeds adds nothing an engine wants.
- Market reports that are just charts. If the analysis is in an infographic, it is invisible. Put the interpretation in text.
- "Top 1% agent" badges. Unverifiable, self-awarded in many cases, and meaningless to a model. Transaction counts and ratios are not.
- City-wide pages. "Homes for sale in Austin" competes with portals. "What it is like to live in Bouldin Creek" does not.
FAQ
Who pays the buyer's agent commission now?
Since the NAR settlement took effect on 17 August 2024, sellers are no longer automatically responsible for buyer agent compensation, and it can no longer be advertised through the MLS. It is negotiated deal by deal. In practice most sellers still offer something, commonly around 2% to 2.5%, because refusing narrows the buyer pool, but buyers should be prepared to pay their agent directly if no seller concession is agreed.
Do buyers have to sign an agreement before viewing homes?
Generally yes. Buyers working with an MLS participant need a written agreement before touring homes, and that agreement must set out how the agent is compensated. This is one of the most significant practical changes from the settlement and a large share of online content still does not reflect it.
Why can't a real estate agent rank for property searches?
Because portals hold the inventory data and have spent two decades building authority on those queries. An answer engine asked about available listings will go to the source with the listings. Agents compete effectively on judgement questions instead, such as how to choose an agent, what commission applies, and what a specific neighbourhood is actually like.
What content actually works for real estate agents in AI search?
Hyperlocal knowledge and decision guidance. Neighbourhood guides containing detail no national dataset holds, an accurate and dated explanation of current commission rules, and an honest treatment of how to evaluate an agent. All three are things portals cannot produce and national content sites cannot localise.
Should agents publish their own performance numbers?
Yes. Transaction count in a specific neighbourhood, average days on market against the local median, and list-to-sale ratio are checkable claims that a model can use and a client can verify. Self-awarded badges and "top producer" language assert nothing and are ignored.
Is now a good time to write about the commission changes?
It remains a strong opportunity because a large volume of existing web content still describes the pre-settlement arrangement as current, and models grounded on that material repeat it. Publishing a dated, accurate, primary-source-linked explanation corrects a consequential and widespread error, which is one of the most reliable routes to being cited.
The short version
You will not beat the portals at inventory and you were never selling inventory. What an agent has that a database does not is judgement about places and about transactions, and both of those are now question types that answer engines are being asked constantly.
Write the neighbourhood guides only a local can write. Publish the current commission rules, dated, while most of the web is still wrong about them. Answer "how do I choose an agent" honestly enough to include the awkward questions, and back it with numbers somebody could check.
Kaymak tracks whether ChatGPT, Google AI Overviews and Perplexity name you when someone asks for an agent in your area, and who gets named instead. The live demo is free, no signup.
Sources: Opendoor, who pays realtor fees after the 2024 NAR settlement · Adler Pollock & Sheehan, understanding the NAR settlement · ShowSmartly, MLS commission changes one year on
Commission practice varies by market and brokerage and continues to evolve. Confirm current local practice and your own brokerage's policy before publishing specifics.