Skip to content

GEO courseModule 7 of 10: What Others Say About YouLesson 7.1

Map Your Town: The Sources AI Uses and the Rivals It Names

Finding out which websites AI tools actually cite when people ask about agents in your town, and which agencies they name instead of you, then turning both into a one-page map that tells you where to spend the rest of this module's effort.

4% of your GEO effort5 min read~3 hours, then quarterly

Why it matters

The sources AI leans on differ by platform, by town and by month. Guess, and you'll spend months on sites your town's AI never reads.

The evidence

Evidence: Practitioner view

A practitioner method resting on vendor studies, which show sources vary by platform and month. They don't prove any one source moves your visibility, which is why you measure your own town.

Do this first

Run six real prompts on three platforms, three times each, and list every website cited and every agency named.

01What it is

Before handing out flyers, you’d watch which noticeboards people on your street actually read. A source map is the list of noticeboards your town’s AI reads, so you don’t waste flyers on the wrong ones.

Think of it like

Watching the fastest runner's training plan. You don't wear their number. You work out what builds your own speed.

02Why it matters

To the people searching

Sellers and landlords shortlist from "best agent" and fee answers before they ring anyone. The sources behind those answers, and the rivals they name, decide who gets the valuation call.

buyers

See the agencies AI names first when they ask who covers an area.

tenants

Get pointed to agents and portals by the sources AI trusts.

sellers

Shortlist from AI answers, so a missing name costs valuation enquiries.

landlords

Compare letting agents through the comparison sites and reviews AI cites.

To Google, Bing and AI

Evidence: Practitioner view

Different tools use different sources. In BrightLocal's study, Google's AI leaned on Business Profiles while ChatGPT leaned on Yelp and Bing, and the two named the same businesses only about a fifth of the time. In a UK snapshot (Whito, August 2026, "who is the best estate agent in [city]?" on ChatGPT, Gemini and Google AI Mode in six cities), the tools drew on comparison sites, Rightmove pages, award sites and review platforms, named 60 agencies across 86 slots, and only 37% of those agencies were named by more than one tool. Perplexity and Claude show their sources, which makes them easy to log. Bing's AI Performance report shows which of your pages Copilot cites.

Agencies named by more than one AI tool, across six UK cities (Whito, Aug 2026)
37%Agencies named by more than one AI tool, across six UK cities (Whito, Aug 2026)
Different agencies named across just 86 recommendation slots
60Different agencies named across just 86 recommendation slots

An agent example

Hallowford is a made-up market town. A small sales and lettings agency runs 6 prompts on 3 platforms, 3 times each: 54 runs. It's named in 7. Rightmove pages appear in 22 runs, a comparison site in 15, a local newspaper's "Best of Hallowford" awards page in 9, and its own website in only 5 (mostly the fees page). The rival named most, in 31 of 54 runs, has five times the Google reviews and publishes a monthly market page. The agency's three gaps: a monthly market page, a steady review routine, and an honest entry on the comparison site. (All numbers invented for teaching.)

03Your turn

Check yours, then fix it

Illustration
Illustration of an AI answer naming three Hallowford agencies, with four numbered sources: a portal, a comparison site, a local newspaper awards page and an agency's fees page. An answer that shows its sources. Each numbered source is a row in your source map: count how many runs it appears in. Example numbers.
  1. 11. Pick 6 to 10 real prompts: "best agent", fees and area prompts, for sales and lettings (lesson 2.2).
  2. 22. Pick the platforms: ChatGPT with search, Google AI Mode, Copilot, Perplexity, Claude with web search.
  3. 33. Run each prompt at least three times per platform, fresh private window each time.
  4. 44. Log every agency named and every source cited or shown.
  5. 55. Count, for each source, how many runs it appeared in: "22 of 54".
  6. 66. Group sources by type: portals, comparison and award sites, review platforms, directories, local press, forums and video, agency websites, official bodies.
  7. 77. Pick your actions from the biggest groups.

  1. 1One row per agency named instead of you: type (independent, chain, online agent, portal), named in X of N runs, the sources that named them.
  2. 2Add a quick look at their public footprint: Google review count and rating (with the date), portal profile, awards, fees page, area guides, market data, press.
  3. 3Final column: what you could honestly learn from them. Never copy their words, photos or reviews.

  1. 1Bing Webmaster Tools, AI Performance: which of your pages Copilot cites.
  2. 2Search Console's Generative AI report: which pages appear in Google's AI features.
  3. 3Pages you expected to see and don't are fix candidates.

Copy this prompt

Here is my AI prompt log (no client data). Group the cited sources into types, count how many runs each appeared in, list every agency named with its count, and tell me which three source types appear most. Don't guess any source I didn't list.

04Keep it up

Build the map once, then repeat it each quarter with the same prompts.

~3 hours, then quarterly