Nobody in that Winkworth office woke up and thought, “Let’s mislead a couple today.” That’s what makes the whole thing so awkward - and so relevant to every estate and letting agent in the country.

A franchisee used AI to stage a property. Soft furnishings, lighting, a bit of furniture. Normal stuff. The sort of thing thousands of agents are doing this week because the tools are cheap, quick, and they make a listing look less like a crime scene.

Then the AI removed a chimney breast. And the disclosure in the PDF brochure was, reportedly, white text on a white background. A couple trekked across London, turned up and found a room that didn’t match what they’d been shown. Then The Times shamed the agency and head office had to make a statement.

Nobody set out to mislead them. But they were misled. And crucially: accidental misrepresentation is still misrepresentation. The Consumer Protection rules don’t give you a special exemption for “it was the algorithm” or “the junior did it on Friday afternoon”.

This is the real AI risk in agency right now. Not deepfake vendors, invented floor plans, or phantom houses. It’s the incremental creep: the negotiator using ChatGPT for descriptions at home on Sunday; the photographer “cleaning up” a bathroom shot and going a touch too far; a brochure proofed in a hurry where the disclosure vanishes into the footer.

The tools are already in your business. The only question is whether you’ve put guardrails around them.

So here’s the simplest useful framework I encourage agencies to adopt: three steps and three lines. Not a 40-page policy nobody reads.

Step 1: Find the AI you’re already using

Start with visibility, not enforcement. Ask your team - no blame, no theatrics - what they’re using: ChatGPT, Canva features, virtual staging apps, transcription tools, chatbots, automated follow-up. You’ll almost certainly discover it’s widespread and unofficial.

You can’t manage what you can’t see.

Step 2: Draw three clear lines

Make them short enough to remember, strict enough to matter.

1) Don’t change the fabric. Stage, don’t restructure. If a buyer walks in and feels deceived, you crossed the line.

2) Say it clearly. If an image is AI-modified, disclose it visibly - on the image or right beside it - not buried in a PDF.

3) Human signs off. AI drafts; your team approves. Every description, claim, measurement, and “close to the station” implication still belongs to you.

Print those three lines. Put them on the wall. If your team can’t recite them, they won’t follow them.

Step 3: Bake it into workflow, not policy

Winkworth didn’t get caught out because they’d never heard of disclosure. They got caught out because the check wasn’t sitting at the exact point where the mistake happened.

Add an AI tick-box to listing sign-off. Add a named approver for AI-assisted copy. Brief photographers and virtual staging suppliers as if they’re part of your brand - because they are. Do occasional spot-checks.

Agents who do this well won’t be the ones with the longest policy document. They’ll be the ones who changed three small steps in how listings go live.

AI isn’t going away. The portals will be flooded with increasingly “perfect” imagery, and Trading Standards will get increasingly interested in what “perfect” really means. The smart play isn’t to ban the tools. It’s to use them with your eyes open - and keep your chimney breasts where they belong: in the building, not in the prompt.

Join me on the Rightmove Resistance Tour 3 - The Tipping Point, to learn more about using AI in your business to win market share, cut costs and increase revenue - safely! Go to www.modelprop.ai/summer26