For twenty years, a homeowner thinking about selling did roughly the same thing: opened Google, typed something like "estate agents in Solihull", and scanned a list of links. The agency's job was to appear on that list and give someone a reason to click. That behaviour is now changing at the top of the journey, and the change is worth understanding precisely — because much of what is written about it is either overstated or imported wholesale from the American market.
This is an attempt to separate what is actually happening from what is being sold as inevitable.
What "AI search" actually means
The term covers two related but distinct things. The first is the AI summary that now sits above traditional results on Google — AI Overviews, and the more conversational AI Mode. The second is the standalone answer engine: ChatGPT, Perplexity, Google's Gemini and others, where a person asks a question in full sentences and receives a composed answer rather than a page of links.
The common thread is that the machine no longer just retrieves — it composes. Ask a traditional search engine which estate agents are well regarded in Harrogate and you get links to judge for yourself. Ask an answer engine the same thing and it may return a synthesised view that names firms, summarises their reputation and cites a handful of sources it considers authoritative.
That is the substantive change: a layer of synthesis now sits between the seller and the agencies competing for their attention.
How consumers are actually starting to use it
The honest answer is: earlier in the journey than agencies think, and less decisively than the headlines suggest.
A Pew Research Center analysis of browsing data from 900 US adults found that users who encountered a Google AI summary clicked a traditional search result in 8% of visits, compared with 15% when no AI summary appeared. A link inside the AI summary itself was clicked in just 1% of visits. That is US evidence, not UK property evidence, but it demonstrates how an answer layer can change what happens before a website visit.
In UK property specifically, the picture is genuinely early. A GetAgent survey of 622 UK estate agents in March 2026 found that 22% of agents reported buyers referencing AI when explaining how they found a property. Only 17% said homeowners were asking whether AI-powered tools were being used to promote their property, and 7% said AI search was already improving listing visibility.
Expectations were much higher than current usage: 84% of the surveyed agents believed AI-powered search would become the primary way buyers search for property in future, while 88% thought property businesses that failed to adapt risked falling behind. Those are respondents' expectations, not a forecast of what will happen.
What is hype
Three things are oversold. The first is urgency framed as emergency. Current UK evidence shows early adoption; it does not give us a reliable measure of instructions being won or lost because of AI search. The commercial effect is therefore still uncertain.
The second is the imported American playbook. Much of the advice circulating on this topic is written for a market with different portals, terminology and consumer behaviour. Principles can transfer; market-specific claims cannot simply be copied across.
The third is the promise of a trick. There is no schema tag or clever phrase that reliably makes an answer engine recommend an agency. These systems change, use multiple sources and remain partly opaque. Certainty about guaranteed AI citations is not supported by the public evidence.
What this changes for UK estate agencies
If some sellers form an impression of an agency inside an AI answer before reaching its website, the public material those answers can draw from matters more. That includes the agency's own site, third-party mentions, directories, reviews and local coverage.
This makes accuracy a useful way to think about the problem. The question is no longer only "do we rank?" but also "if a machine describes our agency to a prospective seller, is the description correct, current and complete?"
What agencies can realistically do
Start with the unglamorous work. Keep the agency name, areas covered, specialisms and contact details clear and consistent. Publish genuinely useful local material that answers the questions sellers in those areas actually ask. Use direct headings and plain answers rather than pages built around marketing filler.
Structured data belongs in the technical hygiene, with an important caveat: it helps search engines understand page content and is required for some Google search features, but current public evidence does not establish that adding schema makes an agency more likely to be named in an AI answer. Do it because the information is easier to parse, not because somebody promised a citation.
Then inspect the outputs. Ask answer engines what they say about the agency and its area, check which sources they cite, and correct factual gaps at their source where possible.
The harder question: discovery is not intent
Everything above concerns discovery — being found, described and shortlisted. It operates before contact. The moment a seller actually gets in touch, the problem changes.
An answer engine may help a homeowner find an agency. It does not tell that agency which of the enquiries now sitting in its inbox represents someone close to instructing and which represents somebody at the beginning of a much longer decision.
Discovery is a retrieval problem. Understanding the person who has just made contact is an inference problem. They are two separate problems, and treating the rise of AI search as if it solves both confuses where the technology is actually acting.