The short answer
Fill in a short fact sheet after the valuation, paste it into a ChatGPT project that already knows your house style and the material information rules, and get a portal-ready description, key features and a social caption back in under a minute. Once that works, a free Google Form and a free Make scenario can email the draft to the negotiator automatically. You still check every fact before it goes live.
- Who it’s for
- UK sales and lettings negotiators, branch managers, independent agents
- Time to set up
- 20 minutes (ChatGPT) + 30 minutes (optional automation)
- Difficulty
- Beginner
- Monthly cost
- £0–£19.99
What does this automation do?
It turns the facts you already collect at a valuation into three pieces of copy: a portal description, a list of key features and a short social caption. The facts come from you, the wording comes from AI, and the checking comes from you again.
You can stop at the first half (a ChatGPT project you paste into, about 20 minutes to set up) and still save most of the time. The second half connects a Google Form to ChatGPT through Make so the draft lands in the negotiator’s inbox without anyone opening ChatGPT.
What do you need before you start?
- A ChatGPT account (free works; ChatGPT Plus is smoother for daily use)
- Three or four of your best past descriptions, to teach it your house style
- For the optional automation: a Google account and a free Make account
How do I set it up?
Step 1: Build a one-page property fact sheet
The quality of the description depends on the facts you give it. Use the same headings every time. The withdrawn Part A, B and C framework still makes a sensible checklist:
ADDRESS (street + town only for the draft):
ASKING PRICE / RENT: PRICE QUALIFIER (e.g. guide price, OIRO):
TENURE: (freehold / leasehold / share of freehold) LEASE YEARS LEFT:
GROUND RENT & SERVICE CHARGE (if leasehold):
COUNCIL TAX BAND: EPC RATING:
PROPERTY TYPE: BEDROOMS: BATHROOMS: RECEPTIONS:
ROOMS & MEASUREMENTS:
HEATING: PARKING: OUTSIDE SPACE:
BROADBAND / MOBILE (from Ofcom checker):
UTILITIES (mains water/gas/electric/drainage):
KNOWN ISSUES (flood risk, restrictions, building safety, rights of way, non-standard construction):
STANDOUT FEATURES (in the vendor's/your words):
NEARBY (schools, stations, shops: facts only, with distances):
TONE NOTES (e.g. "first-time buyer friendly", "premium"):Step 2: Create a ChatGPT project with your house style and rules
In ChatGPT, create a Project called “Listings” (projects let you save instructions so you don’t paste them every time; if your plan doesn’t show Projects, save the instructions in a note and paste them at the top of each chat). Paste the prompt from the next section into the project’s instructions, then upload or paste three of your best past descriptions as style examples.
Step 3: Paste the facts and generate the description
Start a chat inside the project and paste a completed fact sheet. You’ll get the description, key features and caption back. If the tone is off, say so (“shorter, less flowery, more like example 2”) and it will adjust.
Step 4: Check every fact against the fact sheet
Read the draft next to the fact sheet. Look for anything the AI added that you didn’t give it: a “south-facing” garden, “recently renovated”, “walking distance” to a station that is 25 minutes away. Delete it. Check the material information fields in your CRM or portal feed are filled in separately; the description is not a substitute for them.
Step 5 (optional): Turn the fact sheet into a Google Form
In Google Forms, create one question per heading from Step 1. Under Responses, choose Link to Sheets so each submission becomes a row in a Google Sheet. Negotiators can now fill it in on their phone straight after the valuation.
Step 6 (optional): Connect the form to ChatGPT with Make
In Make, create a new scenario with three modules (outline below): watch the sheet for new rows, send the facts to an AI module with your prompt, and email the result. On Make’s Free plan you can use Make’s built-in AI provider, so you don’t need a separate OpenAI API key; it uses extra credits based on the length of the text.
Step 7 (optional): Email the draft to the negotiator for approval
Send the draft to the person who submitted the form, never straight to the portal. Approval stays human. Switch the scenario on and test it with a real (or made-up) property.
The copy-paste prompt
You write property descriptions for [AGENCY NAME], an estate agent in [TOWN], UK.
RULES (follow strictly):
1. Use ONLY facts in the fact sheet. Never invent features, sizes, distances,
orientation, condition, history or "recently renovated" claims.
2. If a material fact is missing (price, tenure, council tax band, EPC, lease
length/charges for leaseholds), do NOT guess. List it under
"MISSING INFORMATION" at the end.
3. Include known issues from the fact sheet neutrally and clearly. Never hide
or soften them.
4. British English. Plain, warm, specific. No clichés: avoid "nestled",
"boasting", "deceptively spacious", "must be viewed".
5. Don't describe who should or shouldn't live there (e.g. "ideal for young
professionals", "no children"). Describe the property, not the buyer.
6. Use the price qualifier exactly as given.
OUTPUT:
A) Portal description: 120-180 words, first sentence = the strongest true
reason to view.
B) Key features: 6-8 bullet points, max 8 words each, facts only.
C) Social caption: 1-2 sentences + 3 relevant hashtags.
D) MISSING INFORMATION: list, or "None".
Match the style of the example descriptions I've provided.Make template outline
You can rebuild this in about 15 minutes. Module names can shift slightly as Make updates its apps; search for the app name and pick the closest match.
| # | Module | Settings |
|---|---|---|
| 1 | Google Sheets › Watch New Rows | The sheet linked to your form; “Table contains headers: Yes”. On the Free plan it checks every 15 minutes. |
| 2 | OpenAI › Generate a response (or Make’s AI text module) | Model: a current small model is enough. Instructions: the prompt above. Input: map each sheet column under its heading. |
| 3 | Gmail › Send an email (or Microsoft 365 Email) | To: the negotiator’s email column. Subject: Draft description: {{Address}}. Body: the AI output plus “Check every fact before publishing.” |
Add a filter between 1 and 2 so rows missing the price or tenure don’t run; instead, route them to an email saying “fact sheet incomplete”.
What UK rules apply?
- Material information is a legal duty. The Digital Markets, Competition and Consumers Act 2024 replaced the Consumer Protection from Unfair Trading Regulations on 6 April 2025. Misleading actions and omissions of material information are prohibited, and the CMA can now fine directly. NTSELAT’s Part A, B and C guidance was withdrawn in May 2025, but the duty remains (Propertymark, CMA207).
- More guidance is coming. The government has said it will publish non-statutory material information guidance and a code of practice for agents in 2026. We will update this guide when it lands.
- Photos and AI images. Don’t use AI to “enhance” photos in ways that change what the property looks like (removing a pylon, adding a lawn). That is a misleading action, whether a person or a tool did it.
- Vendor data. Fact sheets are not usually sensitive, but don’t paste vendor names, phone numbers or personal circumstances into AI tools; they aren’t needed for a description. In ChatGPT settings, you can turn off “Improve the model for everyone” so chats aren’t used for training.
What can go wrong?
- Invented details. The most common failure. The prompt reduces it but doesn’t eliminate it, so Step 4 is not optional.
- Every listing sounds the same. Feed it better “standout features” in the vendor’s own words, and refresh your example descriptions every few months.
- Missing facts slip through. The “MISSING INFORMATION” section only helps if someone reads it. In the Make version, filter out incomplete fact sheets before they reach the AI.
- Running out of free credits. If your branch lists a lot, watch Make’s credit counter for the first month. Make warns you at 75% and 90% of your allowance.
Frequently asked questions
Is it legal to use AI to write property descriptions in the UK?
Yes. There is no rule against using AI, but you are responsible for what is published. Under the Digital Markets, Competition and Consumers Act 2024 you must not mislead buyers or leave out material information, whoever or whatever wrote the words.
Do the NTSELAT Part A, B and C rules still apply?
The NTSELAT guidance was withdrawn in May 2025 when the DMCC Act replaced the old Consumer Protection Regulations. The legal duty to disclose material information did not go away, and many agents still use the Part A, B and C headings as a practical checklist.
Will Rightmove penalise AI-written descriptions?
We have found no Rightmove rule against AI-written text. What matters is that the listing is accurate and the material information fields are filled in. Write for buyers, not for keyword stuffing.
Can ChatGPT see the photos and describe the rooms?
ChatGPT can describe uploaded photos, but it guesses at sizes, materials and condition. Use photos for tone and features only, and take measurements and facts from your own fact sheet.
Is the free version of ChatGPT good enough?
For a few listings a week, usually yes. Projects and higher limits come with paid plans; if you write descriptions daily, ChatGPT Plus (£19.99 a month in the UK app store as of October 2026) saves the copy-pasting of instructions.
Sources
- Propertymark: Material information guidance withdrawn as the DMCC Act takes over
- CMA: Unfair commercial practices guidance (CMA207)
- TLT: Government plans for material information guidance and an agents’ code of practice in 2026
- ChatGPT plans and pricing
- Make pricing
- Make Help: Credits (including AI credit use)
This guide is general information, not legal, tax or financial advice. Tool prices and screens change; we re-check every guide regularly. Spotted something out of date? Tell us and we’ll fix it.