GEO courseModule 8 of 10: Accurate and Above BoardLesson 8.3
Checking and Correcting What AI Says About You
A regular check of what AI tools say about your agency (name, fees, areas, services, reputation), a record of what's wrong and how serious it is, and a routine for fixing it: the source first, then the platform's own feedback route.
Why it matters
AI can state an old fee, a closed branch or a service you don't offer, with total confidence, and nothing tells you it's happening.
The evidence
Evidence: Official guidance
The platforms' own help pages and ICO guidance for the routes. That fixing the source works best is practitioner opinion, but it's the only fix you control.
Do this first
Ask four assistants "How much does [agency] charge?" and "Where is [agency] and what are its opening hours?", and tick every fact.
01What it is
If a parrot keeps repeating the wrong thing about you, shouting at the parrot doesn’t help much. It learned it from somewhere. Fix what it heard, and ask its owner nicely to help too.
Think of it like
A rumour in a village. The fix that lasts is correcting what people are reading and hearing at the source, then asking the loudest repeaters politely to stop.
02Why it matters
To the people searching
People act on what AI says: they ring the old number, expect the old fee, or rule you out over a complaint that was about a different firm. Unlike a bad review, nobody tells you.
buyers
May act on a wrong address, phone number or opening time.
tenants
May expect fees you no longer charge, or that are no longer allowed.
sellers
May rule you out on a wrong fee or service.
landlords
May rule you out on a wrong fee, service or reputation.
To Google, Bing and AI
Evidence: Official guidanceGoogle says AI responses may contain mistakes. Each platform has a different correction route, and none promises to act. Google AI Overviews: the thumbs-down or "Report a problem" option under the answer. ChatGPT: OpenAI's privacy portal can remove personal data about a person from responses, but not a company's reputation, and OpenAI says it can't investigate disputed facts. Claude: Anthropic takes privacy requests about personal data. Copilot: thumbs up or down feedback. Under UK GDPR, a named person (a director or negotiator) can ask for inaccurate personal data to be corrected, normally within a month, but that right covers people, not a limited company. Owners can't correct Google's Ask Maps answers directly: you improve what it reads.
An agent example
At its first monthly check, a lettings branch finds ChatGPT giving its old phone number, Copilot quoting a fee it dropped two years ago, and Google AI Mode listing a branch that closed. It screenshots each, rates them (the fee is high severity), traces the old fee to a PDF still on its site, deletes it, corrects its directory listings, then uses the feedback route on two platforms. A month later it re-runs the same prompts.
03Your turn
Check yours, then fix it
- 1Name: "Tell me about [agency] in [town]."
- 2Fees: "How much does [agency] charge to sell a house?" and "…charge landlords?"
- 3Branch: "Where is [agency] and what are its opening hours?"
- 4Services and areas: "Does [agency] manage rental properties in [area]?"
- 5Reputation: "What do customers say about [agency]?"
- 6Run them on ChatGPT, Google AI Mode, Copilot and Perplexity, logged out, and tick each fact against your fact sheet.
- 1High: could cost someone money or break a rule (wrong fee or rent, "available" when sold, a false membership claim).
- 2Medium: wrong but low harm (old hours, old address).
- 3Low: minor or opinion.
- 4Columns: date, platform, prompt, what it said, correct fact, severity, likely source, action, fixed date. Keep dated screenshots and don't edit them.
- 1Search Google for the wrong text, and search your own site and profiles for it.
- 2Common culprits: an old PDF, an old blog post, a portal profile, a directory, a namesake agency.
Here is an answer an AI gave about my agency, and my agency fact sheet. List every factual claim in the answer and mark it as matches, doesn't match, or not on the fact sheet. [paste both]
- 1Update or remove the page, profile or listing that supplied the error.
- 2Make the correct fact clear, dated and easy to quote on your fact sheet and key pages.
- 3Ask portals, directories and review sites to correct their own entries.
- 1Google AI Overviews: thumbs down or "Report a problem" under the answer, with the correct fact and your source.
- 2ChatGPT and Claude: their privacy request routes, for personal data about a named person only.
- 3Copilot: the feedback buttons on the answer.
- 4For unlawful or defamatory content: Google's legal removal form, and take legal advice.
- 1Date, platform, route used, outcome.
- 2Re-test after a week and a month, logged out. Recrawling can take days to months.
Write a short, polite, factual message for an AI platform's feedback form saying its answer gave the wrong [fee/address/service] for [agency]. The correct fact is [fact] and it's published at [URL]. Under 100 words, no legal claims.
Then: If an error is causing real harm (a wrong fee being quoted to clients, or something defamatory), talk to a solicitor or your data protection adviser before relying on any data protection or defamation route.
04Keep it up
Monthly, with the same prompts each time, and alongside your monthly prompt run.
That's lesson 8.3
Worked through it? Mark it done and we’ll keep track on this device.
← Back to 8.2 Rental Adverts and Guides Under the Renters' Rights Act