AI visibility for estate agents.

The battle is for vendors, and it has moved

Buyers live on the portals; the portals are not where an agency wins. The commercially decisive question is the vendor’s - which agent should sell my house - and that research has moved into AI assistants: what agents charge, whether online beats high-street, who is best in this postcode, how long homes here take to sell. The answer that comes back names two or three agencies and says why, drawing on review platforms, comparison services and local coverage. Being on Rightmove settles none of it: portals win listings exposure, not the “which agent” decision. And as across every sector we measure, the agencies named are not reliably the ones topping local Google results - the evidence sits in GEO versus SEO.

AI answerillustration

“Best estate agent in [city]”

For selling a home in [city], the agency most often recommended is Your Agency^[1], noted for a published fee structure, dated local sales data and strong vendor reviews. Two other local agencies are also mentioned.

[1] the citation - this is the position the work below is designed to earn.

What vendors actually ask

“Fees on request” donates the answer to someone else

The fee question starts most vendor journeys, and most agency websites refuse to answer it. The engine does not shrug and move on - it builds the answer from whoever will say something: national comparison services, online-first agencies, consumer sites. Every “fees on request” is a small donation to a competitor’s visibility. The alternative is not publishing a race-to-the-bottom number; it is publishing the honest shape: a typical percentage range, what is included at each level, what moves it, on a dated page a machine can quote. Agencies fear the fee page starts price conversations. In practice it filters them: vendors who arrive having read it are asking about service, because the fee shape was already accepted.

Your local data is the authority nobody else can copy

Every agency sits on genuinely original local information: time-to-sell by property type, asking-versus-achieved patterns, what is moving this quarter and what is not. Almost none of it gets published, which is why the “how long to sell in [area]” answers cite portals and national indices instead of the agents who actually hold the ground truth. A short, dated, honest local market note - real figures, stated sample, no spin - published monthly, is the single most defensible authority asset an agency can build: it cannot be templated by a competitor without their own data, engines treat original data sources preferentially, and it gives local press something to cite, which feeds the recommendation answers in turn. This is the honest version of local content. The dishonest version - near-identical pages generated for every postcode - is now explicitly the kind of scaled sameness engines penalise, and we will tell you plainly if your site carries it.

What a fees page looks like when it is built to be quoted

It opens with the direct answer: the typical percentage range this agency charges to sell a home, stated plainly. A real table follows - standard sole agency, premium marketing, multi-agency - with what each tier includes, because inclusions are what make a range meaningful rather than evasive. A short paragraph explains what moves the fee: property type, price band, marketing level. It carries the date it was last reviewed and a named valuer or director, with their profile connected in structured data. Beneath sit the questions vendors actually ask as visible FAQs, schema generated from the same text so page and markup cannot drift. Nothing on the page is a discount signal; it is the agency answering the first question of every instruction journey before a competitor answers it for them.

fig. - the fees page, built to be quoted. Every block is something an engine lifts.

What we know, with sources

• AI answers name a handful of agencies with reasons rather than listing links - unnamed means uninvited to the valuation.
• Only 38% of pages cited in Google’s AI Overviews also rank in the top ten for the same query (Ahrefs, 2026) - local pack position is not the deciding signal.
• Fee and speed questions dominate vendor prompts, and engines quote published, dated figures - silence concedes both.
• Recommendation answers lean on independent surfaces: review platforms, comparison services, local editorial - not the agency’s own site alone.
• Google Analytics now reports AI-referred visits as their own channel; treat the counts as a floor, since many AI visits arrive without a referrer.

Choosing the questions worth winning

Instruction-winning prompts get the effort; applicant prompts do not. City-level “best agent” questions are recommendation battles won through reviews and independent citations; area-level speed and price questions are won by whoever publishes dated local data; the fee question is won by whoever answers it. Seasonality matters too: vendor research swells ahead of the spring and autumn markets, and the engines’ source material is not rebuilt overnight, so the publishing work lands the quarter before. The audit weights the prompt set to your catchment and calendar, marks the unwinnable informational questions honestly, and then holds the set fixed - because a measurement instrument you keep changing is not an instrument.

Branches are entities, and drift is expensive

Multi-branch agencies usually present as several slightly different businesses: names that vary by a word, addresses formatted three ways, phone numbers that disagree between the site, the Google profiles and the portals. Engines resolve identity before they recommend, and near-matches dilute into nothing. The fix is structural: one exact identity per branch, each branch its own properly described local entity, all connected to the parent brand through organisational schema, with sameAs linking the profiles that corroborate each one. Done properly, branch structure becomes an advantage - local answers favour genuinely local entities - rather than the quiet leak it usually is.

The gaps we keep finding in agency sites

Valuation-form-first websites with almost no answerable content - built to capture, not to be cited. Fees absent everywhere. Local “area guides” that are interchangeable prose with the place name swapped - the templated pattern engines now discount. Branch identity drift across site, profiles and portals. Negotiators and valuers invisible as people, in a business where vendors choose people. Review presence concentrated on one platform and disconnected from the site’s entity. Market data hoarded internally while national indices win the local answers. And the silent CDN-edge block that removes the agency from an entire engine while robots.txt looks clean. All checkable; all fixable.

What the audit checks, and the Build fixes

The audit runs a fixed vendor-weighted prompt set across the agreed engines; checks crawler access including the edge; maps identity across every branch, profile and portal; validates the RealEstateAgent and Organization schema stack with sameAs to the corroborating surfaces; grades the fees page and priority local pages for answer-first quality, freshness and honesty; and inventories the independent citations - review platforms, comparison services, local press - that decide the recommendation answers. The Build then closes the gaps in priority order: access corrected and re-verified, one identity per branch properly structured, schema deployed, the fees page rebuilt to be quotable, the first dated local market note published and templated for monthly repetition, and the citation groundwork laid. The sample audit shows the report format.

Where independent citations come from for agencies

The recommendation answers are decided off-site, on surfaces the engine already trusts: the review platforms vendors actually check, the agent-comparison services that hold performance data, and local press property coverage - which a monthly market note feeds naturally, because journalists cite agents who publish figures. What our measurement keeps confirming across sectors holds here too: the rate of fresh independent mentions predicts non-branded visibility better than any single on-page change. Engines recommend the agencies other sources keep vouching for. So the citation layer runs as a monthly discipline inside the Watch - profiles held identical to the site’s identity and connected via sameAs, review flow maintained and answered, genuine editorial taken when it appears, nothing paid for placement.

Measured honestly, or not at all

The baseline records citation share across the vendor prompt set, split branded versus non-branded - and in this sector the split is the whole story. An established local brand often shows healthy total citations that are almost entirely branded: vendors who already had the name, checking it. That is reputation, not discovery. The non-branded share - how often the agency is named when nobody asked for it - is the number that predicts new valuations. From the baseline, the same set is re-run on a fixed cadence and reported plainly, with the standing caveat that AI-referral counts in analytics are a floor. The full protocol is in the measurement guide.

How practices work with us

Questions we get asked

Can you guarantee ChatGPT will recommend our agency?

No - nobody controls what an AI engine generates. What can be controlled are the foundations engines rely on when they choose which agencies to name: crawl access, one consistent identity per branch, structured data, a quotable fees page, published local data and independent corroboration. That is the work, measured openly against a dated baseline.

We're on Rightmove and Zoopla. Doesn't that cover it?

The portals cover buyers looking at listings. They do not decide the vendor's question - which agent should sell my house - and that is the answer this work targets. Portal profiles do matter as identity surfaces: they should match your site's details exactly and be connected through structured data, because engines cross-check them when resolving who you are.

We don't publish fees. Is that really a problem?

For AI visibility, yes: the fee question starts most vendor journeys, engines quote whoever publishes, and silence hands the answer to comparison services and online-first agencies. Publishing an honest range with inclusions - not a race-to-the-bottom number - is enough to be quotable and tends to improve enquiry quality rather than harm it.

Does this bring us buyers or vendors?

Vendors, by design. The prompt set is weighted to the questions people ask when choosing an agent to sell - fees, speed, reputation, who is best locally - because that is where agencies actually compete. Buyer-side visibility rides along where it overlaps, but the measurement and the fixes target instructions, not applicants.

We have several branches with different names locally. Does that matter?

It matters a great deal. Engines resolve identity before they recommend, and branches presenting as near-matches dilute the brand into nothing. Each branch needs one exact identity, structured as its own local entity and connected to the parent through organisational schema. Done properly, the branch network becomes an advantage in local answers rather than a leak.

Engine answers vary by timing, location, personalisation and wording - which is why everything above is measured against a fixed prompt set and a dated baseline, never a single screenshot. If you want the two-minute version first: tell us your firm and the question your clients ask, and we’ll show you who the AI names today.

Get your free visibility check

Or see the full service detail →

Last updated: July 2026