To be cited by AI search in 2026, a page must state a checkable fact early, in plain language, near a heading that matches the question. Generative engines extract sentences, not pages. A paragraph reading “Estate agent fees in Manchester typically run 0.9% to 1.5% plus VAT” can be lifted into an answer; “We pride ourselves on competitive, transparent fees” cannot.
The other half is entity consistency: identical business name, address and phone number everywhere, an Organization or RealEstateAgent schema block, and enough third-party mentions that the model can confirm you exist.
A growing share of the people who would once have typed “best estate agent in Leeds” into Google now ask an assistant instead, and get back three named agencies with a sentence each. If you are not one of the three, the search happened and you never saw it, it produces no impression in Search Console and no click in your analytics.
This guide covers what actually determines those citations, based on testing our own and client pages against ChatGPT, Google AI Overviews, Perplexity and Claude through 2026. Some of it is familiar SEO. A meaningful part of it is not.
1. What actually changed about search
The mechanics of discovery have shifted in a way that does not show up cleanly in your analytics.
Someone considering selling used to search, scan ten blue links, and click two or three. Now a growing share ask an assistant, and receive a paragraph naming two or three agencies with a sentence about each. If you are not named, that consideration happened and left no trace on your side: no impression in Search Console, no session in analytics, no way to know it occurred.
This is the uncomfortable part of AI search. Traditional SEO gives you a feedback loop. Generative search largely does not, which means you have to go and look.
Generative engines extract sentences, not pages. A page can rank well and still never be cited, because nothing on it is phrased as a quotable, checkable fact.
2. What gets cited, and what gets skipped
Across our testing of UK property queries through 2026, the pages that get cited share a pattern that is easy to describe and slightly awkward for marketers.
| Written as | Example | Outcome |
|---|---|---|
| Specific, checkable fact | Estate agent fees in Manchester typically run 0.9% to 1.5% plus VAT. | Frequently extracted and cited |
| Named entity plus attribute | LineUp Agency works with estate agents across the UK, Poland, France, Spain and Italy. | Used to establish who you are |
| Direct answer under a matching heading | H2: How long does it take to sell a house in Leeds? Answer in the first sentence. | Strong extraction candidate |
| Marketing claim | We pride ourselves on competitive, transparent fees. | Ignored, nothing to extract |
| Buried answer | Answer arrives in paragraph seven after context-setting. | Usually missed |
The practical rule: put the answer in the first two sentences under the heading that asks the question, in plain declarative language, with a number or a named specific in it wherever you honestly can.
This is also better for human readers, which is convenient. The writing that AI systems extract well is the writing that respects someone’s time.
3. Becoming a nameable entity
The second half of the work is not about pages at all. A model will only name a business it is confident exists, and confidence comes from corroboration across sources.
| Signal | Effort | Observed effect on being named |
|---|---|---|
| Identical name, address and phone everywhere | Half a day, once | Largest single improvement |
| Third-party mentions (press, directories, bodies) | Ongoing | Strong; corroboration is the mechanism |
| Factual about page (founded, by whom, where) | An hour | Consulted often; frequently thin |
| Organization / RealEstateAgent schema | An hour | Removes ambiguity; not a switch on its own |
| Publishing volume | High | Weak on its own without the above |
Consistency, everywhere, exactly
Business name, address and phone number identical across your website, Google Business Profile, Companies House, portal profiles, professional body listings and local directories. Not similar. Identical, down to whether you write “Ltd” or “Limited”. Inconsistency here is the most common reason a perfectly legitimate agency is treated as ambiguous.
Structured data that says who you are
An Organization or RealEstateAgent block with name, URL, logo, address, phone, area served and sameAs links to your verified profiles. No AI provider states that schema triggers a citation, and we have not observed it acting as a switch. What it does reliably is remove ambiguity, which makes a model more willing to name you.
Third-party mentions
Local press, professional body directories, chamber of commerce listings, sponsorship pages, supplier case studies. These matter more for GEO than for classic SEO because corroboration is the mechanism. A business that only appears on its own website is a business the model cannot verify.
An about page that states facts
Founded when, by whom, operating where, how many people, what you specialise in. Most agency about pages are a paragraph of sentiment and a stock photograph. That page is one of the most-consulted sources when a model is deciding whether it knows who you are.
4. Page structure for extraction
Six structural habits account for most of the difference in our testing.
- Phrase headings as questions people ask. “How much do estate agents charge in Bristol?” not “Our Fee Structure”.
- Answer in the first 40 words under that heading, then expand. Context first is a habit worth unlearning.
- Use comparison tables. They are cited disproportionately often, because the structure is unambiguous and the data is separable.
- Date your facts. “As of August 2026” alongside any figure. Models weight currency, and undated numbers are a liability once they age.
- Keep one idea per paragraph. Extraction works at paragraph level; a paragraph covering four points is hard to lift cleanly.
- Add a genuine FAQ block with questions phrased the way people actually type them, and answers that stand alone without the surrounding page.
None of this requires writing badly. It requires putting conclusions before reasoning, which is a discipline rather than a compromise.
5. How to measure it, honestly
There is no Search Console for AI citations. Anyone selling you a precise share-of-voice figure for generative search is selling a model of a thing rather than the thing.
What works is manual and takes about twenty minutes a month.
- Write down eight to twelve prompts a real client would use: best estate agents in [your town], estate agent fees in [your area], how to choose an estate agent locally, who sells flats in [development].
- Run them in ChatGPT, Google’s AI Overviews, Perplexity and Claude, logged out, in a fresh session.
- Record which businesses are named and which sources are linked.
- Repeat monthly and watch the direction of travel.
Results vary by user, location and session, so treat any single run as one sample. The trend across three months is meaningful; a single query result is not.
The traffic expectation to set internally
AI citations produce less direct traffic than the equivalent ranking would, because many people read the answer and stop. The value shows up later as branded searches and direct enquiries that you cannot attribute. Set that expectation before starting the work, or the programme will be judged against a metric it was never going to move.
6. Should you block AI crawlers?
The question comes up because publishers have been vocal about it, and their situation is not yours.
A publisher’s content is the product; when a model reproduces it, the sale is lost. An estate agency’s content is marketing material whose entire purpose is to be distributed. Blocking removes you from the answers your competitors appear in and protects material you were giving away deliberately.
For essentially every UK agency, allow access. The exception is genuinely proprietary data, a subscriber-only market report you sell, for instance, which should sit behind authentication rather than a crawler directive anyway.
If you want to be selective, the major crawlers can be controlled independently in robots.txt: OpenAI distinguishes between its search crawler and its training crawler, and Google’s AI features are governed partly by Google-Extended. Blocking training while allowing search is a defensible middle position, though it is worth knowing that it reduces the corroboration that makes a model confident enough to name you in the first place.
7. Frequently Asked Questions
What is the difference between GEO, AEO and SEO?
SEO aims to rank a page in a list of links. AEO (answer engine optimisation) aims to have a specific answer extracted into a featured snippet or voice result. GEO (generative engine optimisation) aims to be cited as a source inside an AI-generated answer. They overlap heavily, all three reward clear structure and verifiable facts, but GEO puts far more weight on being a nameable, corroborated entity.
Does schema markup help you get cited by AI search?
It helps indirectly. No AI provider states that schema drives citation, and we have not observed it acting as a switch. What it does reliably is remove ambiguity about who you are, where you operate and what you offer, which makes a model more willing to name you. Treat it as necessary groundwork, not a lever.
How do you check whether AI search mentions your agency?
Ask the assistants directly, in the phrasing a real client would use, from a logged-out session: best estate agents in your town, estate agent fees in your area, how to choose an agent locally. Repeat monthly and record which competitors appear. It is manual and slightly tedious, and it is currently the only honest way to measure it.
Do AI search citations bring actual traffic?
Less than the equivalent ranking would, and that is the uncomfortable part. Many users read the answer and stop. The value is in being named as a credible option at the moment of consideration, which shows up later as branded searches and direct enquiries rather than as referral traffic you can attribute.
Should estate agents block AI crawlers?
For most agencies, no. Blocking removes you from the answers your competitors appear in, and the content being protected is marketing material you want distributed. The calculation differs for publishers whose paid product is the content itself, which is not the case for an estate agency.
8. Sources & References
Reference material and further reading used while writing this guide. Figures described as our own are from LineUp campaign data, as set out in the methodology note above.
- Google Search Central, SEO documentation. Official guidance on crawling, indexing, structured data and helpful content. developers.google.com ↗
- Schema.org vocabulary. Structured data types referenced throughout, including RealEstateAgent and FAQPage. schema.org ↗
- Google Search Central, AI features and your website. Google’s own documentation on how AI Overviews source and link to pages. developers.google.com ↗
- OpenAI, ChatGPT search and crawler documentation. How OAI-SearchBot and GPTBot access sites, and how to control them. platform.openai.com ↗