Why doesn't ChatGPT mention my agency?

Most agents find out the hard way: a quiet moment, a phone, and a question they half wish they had not asked. When the answer names the agency down the road, it feels like a verdict. It is not. An AI answer is assembled in seconds from whatever evidence the engine can find and trust, and when your name is missing, it means the evidence was missing, not that the machine has judged you and found you wanting. We run this measurement formally across five engines, ChatGPT, Google's AI Mode, Perplexity, Microsoft Copilot and Claude, capturing the answers verbatim so agencies can see exactly where they stand. The same five causes come up again and again.

First, check how you asked

Before any diagnosis, one trap to avoid. If you ask ChatGPT about your agency while signed in to the account you use every day, the answer is contaminated: ChatGPT's memory knows who you are, and it will quietly flavour the response with what it already knows about you. An agent who has ever mentioned their agency to ChatGPT can be shown a warmer answer than any stranger with a house to sell will ever see, which is a comforting result and a useless one. It is why, when we measure, we insist on clean sessions: the answer that matters is the one shown to someone who has never heard of you. Ask from a fresh session, phrase the question the way a seller would phrase it, and ask several variations, because engines that name you for one phrasing can go silent on another. One agency we measure is named in twenty nine of thirty one answers across the five engines, and the gaps were specific phrasings, not a general absence. If your name still does not come up, read on.

Cause one: it cannot read your website

ChatGPT does not browse the way a person does. When it goes looking, it reads the raw, machine-readable layer of your site: the text as served, the structured data that says "this is an estate agent, here is where it operates, here is what it does". Plenty of agency websites, including expensive ones, serve that layer badly or not at all. Content that only appears after scripts run, no schema mark-up, key facts locked inside images. To a visitor the site looks fine. To the engine it is a shop with the shutters down. This is the first thing we audit, because when the machine layer is missing, nothing else you do gets read.

Cause two: nobody else backs you up

Engines are cautious about self-praise. Before naming an agency, they look for corroboration: what do other sources say? Google reviews carry serious weight here, both the rating and the volume, and so do directories, trade press, professional memberships and other websites that mention you. An agency with a thin footprint beyond its own site gives the engine nothing to check its answer against, and engines do not like recommending names they cannot verify. Your rivals being named is usually not a sign they are better agents. It is a sign they are better evidenced.

Cause three: your Google Business Profile is working against you

The humble Google Business Profile turns out to be one of the most quoted sources in AI answers, and it hurts you silently when it is wrong. Wrong or vague category, stale description, unanswered reviews, missing service areas. It is one of the most common faults we find when we audit an agency's footprint, and one of the most invisible: Google fills in categories and descriptions with defaults and guesses, nobody at the agency ever looks at the profile from the outside, and the engines simply take Google's word for it. If your profile says less than it could, or says something slightly wrong, that is the version of you the machines are repeating to every seller who asks.

Cause four: the answer is built from stale information

AI engines mix live lookups with older, indexed and remembered data, and they do not always tell you which is which. In one measurement we watched an engine confidently quote an agency's review score from figures the agency had outgrown months earlier, lifted from a forgotten line of text in the site's own code. If your website, profiles and directory entries carry outdated numbers, old branch details or a previous trading name anywhere, an engine will eventually serve them back to a potential client as fact. Consistency across everything the machines can read is not cosmetic. It is what gets quoted.

Cause five: you are asking for a seat that is already taken

For any given question there are only a few names in an AI answer, usually three to five, and engines lean towards names they have seen recommended before. If established rivals already hold those seats for "best agent in [town]", displacing them takes time and evidence. But the same measurement that shows you the occupied seats also shows you the empty ones: the niches, nearby towns and specific questions where no agency has made a case yet. Being first to properly evidence a seat is far easier than evicting someone from one. That is where measurement turns from a scoreboard into a plan.

What actually fixes it

The fix list follows the causes: a website that serves a proper machine-readable layer, a corroboration trail beyond your own site, a Google Business Profile that says precisely what you do, consistent and current facts everywhere your name appears, and content that makes the case for seats still standing empty. We have written up what the engines actually read, in order, and the six moves that shift it from our captured answers, and a full five-engine measurement of one agency's patch if you want to see what the evidence looks like. None of it is instant. Engines revisit, re-crawl and slowly re-decide, which is exactly why we measure monthly rather than promising overnight miracles.