Material Information in Listings: What Estate Agents Must Capture

An estate agent must capture anything that could change a buyer's decision to view or offer: Part A (price, tenure, council tax band, leasehold costs), Part B (property type, rooms, utilities, broadband, mobile signal, parking) and, where relevant, Part C issues such as flood risk, building safety, rights and planning. AI should read documents and chase gaps, never guess values.

The duty itself comes from the Digital Markets, Competition and Consumers Act 2024, usually shortened to the DMCC Act, which replaced the 2008 unfair trading regulations in April 2025 and treats leaving out material information as an unfair practice. The Part A, B and C lists come from National Trading Standards guidance written under the old regulations, and that guidance was withdrawn in May 2025. The duty did not go with it. A listing that leaves out something the buyer needs can be misleading even if every word in it is true, which is why the work is a checklist and a paper trail, not a copywriting job.

Follow me on Instagram@sagnikteaches

What Parts A, B and C cover now the guidance has gone

Until sector-specific guidance appears, the three parts remain the most complete list an agent can work from, and National Trading Standards developed them with the property portals. The withdrawn sales guidance (November 2023) described Part A as information that is always material for every property. Part B is information to establish for every property, mostly physical features and utilities, where something unusual would change whether someone looks any further. Part C only needs establishing if the property is affected by the issue in question.

Connect on LinkedInSagnik Bhattacharya

Part A is short and precise. You need the council tax band, the asking price as a number, and the tenure. For a leasehold you also need the current ground rent with its review period and next review date, the current service charge and the length of the lease. A shared ownership sale needs the share being sold, the rent payable and any further liabilities. Inescapable costs such as estate rent charges belong here too. Do not hide the price: National Trading Standards published an opinion in May 2022 that "price on application" listings are unlawful, and leaving the price out would be just as hard to defend under the new Act. For a letting, the headline items a tenant needs before viewing are the rent, any deposit and the council tax band.

Subscribe on YouTube@sagnikteaches

Part B in the withdrawn guidance lists property type, construction (a thatched roof or a non-standard build, for example), the number and types of room, electricity supply, water supply, sewerage, heating, broadband, mobile signal or coverage, and parking. Part C lists building safety, restrictions, rights and easements, flood risk, coastal erosion risk, planning permission or proposals for development, accessibility and adaptations, and whether the property is in a coalfield or mining area.

Treat those lists as a floor, not a ceiling. Under the DMCC Act the agent has to judge what the average buyer needs to know, so a known damp problem or an unusual running cost belongs in the file even though no part names it. When the government consulted on material information in late 2025, respondents also suggested categories such as pollution risk, neighbour disputes and renewable energy installations, which shows how wide "material" can run.

One practical point about placement. The withdrawn guidance put Part A on the first page of the listing and said Parts B and C should be no more than one click away online. That is still a sensible standard to hold yourself to. Your portal's data fields will do most of this for you, but only if somebody has filled them in.

Why the information goes missing in practice

Agents rarely leave out material information on purpose. It goes missing in four places. The vendor never mentions the 12-year-old extension without a completion certificate. The lease is in a solicitor's file and nobody asked for a copy. The negotiator listing the property knows the broadband is slow but writes "good transport links" instead. Or the information is known to the firm, but sits in an email thread that the listing never sees.

An AI step fits the first three problems well, because they are reading and chasing problems. It does not fit the judgement calls, such as whether a boundary dispute is worth disclosing. So the workflow below runs in the order a listing actually moves, with the judgement calls left to people.

A collection workflow in five passes

Pass 1: build a property fact sheet from what you already hold

Before the valuation visit, gather whatever exists: the previous listing, the energy performance certificate, any floor plan, and the vendor's own enquiry form. Give those to the assistant and ask it to fill a fact sheet with a source column. A prompt that works:

You are helping an estate agent prepare a listing. Below are documents about one property.
Fill the fact sheet table with: tenure, council tax band, number of bedrooms and bathrooms,
property type, heating type, parking, and any restriction or issue mentioned.
Rules:
1. For each field give the value and the exact quote or document name it came from.
2. If a field is not stated in the documents, write NOT FOUND. Never infer or assume.
3. If two documents disagree, show both values and mark CONFLICT.
Documents:
[paste]

Rule 2 is the important one. Language models tend to fill a blank with a plausible value, and a plausible council tax band in a listing is exactly the kind of error the regime exists to prevent. Telling it to write NOT FOUND, and then testing that it does, is the first safety check.

Pass 2: question the vendor for what only they know

The assistant can turn your NOT FOUND list into a vendor questionnaire written in plain language. A vendor will answer "Has any part of the property ever flooded, even briefly?" more reliably than "Please confirm flood history". Ask for answers in writing so you can show later what you were told and when.

A short before and after shows the difference. A negotiator's note reads "Vendor says drainage is fine." The questionnaire version reads: "Is the property connected to mains drainage or a septic tank or private treatment plant? If private, who owns and maintains it, and when was it last emptied or serviced?" The second wording produces a Part B answer you can publish. The first produces a sentence you cannot defend.

Pass 3: verify what can be checked cheaply

Do not leave a verifiable fact on the vendor's word. The government's June 2026 response on material information says information should be gathered from source, meaning the organisation that owns the data or the homeowner directly. The council tax band can be looked up publicly. Broadband availability and mobile coverage have official checker tools. Tenure can be confirmed against the title register. The assistant can prepare the lookup list and record the result, but a person should open the source. If your firm uses a tool with an automated lookup, treat its output as a source and name it in the tracker.

Pass 4: extract from documents with a human second look

Leases are where AI saves the most time. Suppose a flat is being sold and the lease runs to 70 pages. Paste only the clauses that matter, or the whole lease where your firm's terms allow it, and ask for a structured extraction:

From the lease text below, extract: lease start date, term in years, ground rent now,
ground rent review dates and basis, service charge arrangements, any restriction on
pets, short lets or alterations. Quote the clause number for every item.
If a value is not in the text, say NOT IN EXTRACT. Do not calculate the years remaining.

The last line matters. A model can subtract badly, so ask it for the start date and term and calculate the remaining years yourself, or in a spreadsheet. A realistic output might read: "Term 125 years from 1 January 2005. Ground rent 250 a year, doubling every 25 years. Clause 4.2: no short lets." You then confirm clause 4.2 against the page. That second look takes two minutes and is the whole point of the quoted clause number.

Pass 5: reconcile before publication

The last pass is a gap report. Ask the assistant to compare the fact sheet with your portal's required fields and list what is empty or conflicting. The output is a list, not a decision. Someone senior in the branch signs it off, and the sign-off is recorded in the tracker with a date.

The tracker that makes this auditable

Whatever tool you use for this, a spreadsheet works fine, create one row per item and give each row the same columns: the item, the value, the source, who confirmed it and the date. Add a status of Established, Not applicable, or Unknown. "Not applicable" needs a reason ("no coal records for the area; searched on the checker"), because an empty cell could mean either "not relevant" or "nobody looked".

This table is illustrative. The "Vendor-declared" status is worth keeping, because it tells a later reader that the firm did not independently check the item.

A worked example: a one-bedroom leasehold flat, start to finish

Take a one-bedroom flat in a converted house, to be marketed at 215,000. The negotiator uploads the vendor's enquiry form, an energy certificate and the old agent's brochure. Pass 1 produces a fact sheet that gets nine of fourteen fields from the documents. The tenure comes out as leasehold, but the brochure says "share of freehold", and the form says leasehold, so the assistant marks CONFLICT. That alone justifies the pass: a brochure from a previous agent had been repeated across two listings.

Pass 2 produces eight vendor questions. The vendor answers six the same day. Two come back as "not sure": whether the roof terrace is included in the lease plan, and who maintains the shared driveway. Those become Part C questions about rights and restrictions, and they go to the vendor's solicitor with a seven-day date on the tracker.

Pass 3 checks the council tax band (B) and the broadband (the checker shows full fibre at the address, so the entry is "full fibre available" with the date). Pass 4 extracts the lease: 112 years left, ground rent 100 a year with no review, service charge 1,800 a year as budgeted. Pass 5 flags the two open questions and one missing item, the building's fire risk assessment, which the managing agent has not supplied. The listing is published the day after the solicitor answers. The branch kept a dated record of all of it, which cost about forty minutes of staff time beyond a normal listing, a figure I am giving only as an illustration of the order of magnitude.

Lettings and a second Part C example

The same machinery works for rentals, with a lighter set of headline items: the rent and any deposit take the place of price and tenure. Where the workflow earns its keep in lettings is the Part B and C detail that tenants ask about on the first call. Take a two-bedroom rental above a shop. The landlord's notes say "heating included". Ask the assistant to list every statement in the notes that a tenant could rely on, then turn each into a question for the landlord: is the heating gas, electric or a shared system, who pays for it, and is any of that charged back through the rent? You end up with an answer such as "electric panel heaters, tenant pays through own meter", which is publishable.

For Part C, imagine a ground-floor flat where the vendor mentions that the neighbour has a right to walk across the back yard to reach the bins. That is a right over the property, and it belongs under restrictions and rights. An assistant cannot decide whether it matters, but it can make sure the sentence is not lost: ask it to scan every vendor email for words such as "access", "easement", "covenant", "boundary" and "shared", and list each sentence with the sender and date. A negotiator then decides what goes into the tracker. Reading forty emails for four such sentences is the kind of job a person skips on a busy Friday and a model does not.

What the AI must not infer

This is the part to put on the wall. Some outputs look reasonable and are exactly the ones that produce a misleading listing.

• Council tax band from the address or price. It is a lookup, not an estimate. A model may state a band confidently with nothing behind it.
• Remaining lease years. The arithmetic must be done outside the model, and the lease start date must be read from the document, not assumed from the building's age.
• "No flood risk". A model has no access to the flood map for that address unless you gave it the result. Silence in a document is not evidence.
• Planning and building regulation status of an extension. Photos and descriptions cannot show whether consent was granted.
• Broadband and mobile signal. Speeds vary by street and provider. Quote the checker result and the date.
• Dimensions. Room sizes from a floor plan image are often wrong. Use measured figures from a survey or your own measurements.
• Cladding or building safety status. This needs a document from the building's responsible party. Never let a model say a building is "compliant".

The same list covers marketing copy. If you use AI for the description, as described in how estate agents use AI to write property descriptions, feed it the verified fact sheet, not photos and guesses, and tell it not to add features that are not on the sheet.

A before and after on one weak listing line

Here is a realistic failure. The draft line reads: "A quiet, well-connected home with a generous garden and no known issues." There are three material problems. "No known issues" is a statement that something was checked. "Generous garden" gives no size, and the garden is shared with the ground-floor flat. "Quiet" is an opinion the surveyor may contradict.

The corrected version, written from the fact sheet, reads: "Shared rear garden, approximately 9 by 6 metres, used with the ground-floor flat (see lease plan). Fibre broadband available at this address (checker, 4 November). Flood history: none reported by the vendor. Planning status of the loft room: documents requested." It is less poetic, which is fine, because the poetry goes in the part that does not carry the legal weight. Ask your assistant to do this rewrite and then check every number against the tracker.

Privacy and the documents you paste

Leases, searches and vendor questionnaires contain people's names, addresses and sometimes financial details. Use a business plan whose terms exclude your content from model training, and tell your team which tool is approved. On the Microsoft side, for example, Microsoft Learn states that prompts and responses in Copilot Chat under enterprise data protection are not used to train foundation models, with a separate position for web search queries. Other vendors publish equivalent terms, so read the one for your tool. If your firm has customers in the EU, data-protection law such as the GDPR applies to the personal details in those documents. Where you cannot approve a tool, extract the few values by hand and keep the document out of the prompt.

Guidance and sales packs that will change your form

This area is moving. In its June 2026 response to the home buying and selling consultations, the government said it will publish non-statutory guidance explaining agents' existing duties under the DMCC Act, work with industry on a standardised form for gathering material information, and later legislate, when parliamentary time allows, to require sales packs prepared before listing, including searches and a property condition report. It also said agents should have regard to the guidance as soon as it is published, rather than after an implementation period.

Before you roll this workflow out, check whether that guidance or the standardised form has appeared, read your portal's current field list, and set a quarterly reminder to compare your own form against both. Update the assistant's prompt whenever the form changes, because a prompt built on last year's list will cheerfully report a complete file with the new items missing. If a standard form does arrive, map its fields to your tracker columns once and keep the five passes as they are: the collection work does not change, only the list it collects against.

If you want to see where this document-reading step sits next to valuation prep and vendor updates, the stack of AI tools for estate agents by job maps it, and listings that pass material information checks covers the copy side from another direction.

How to check that the workflow actually works

Run a test before trusting it. Take five recently listed properties and rebuild their fact sheets using the passes above, without looking at the published listings. Then compare. Count the fields where the sheet differs from the listing, and for each, decide which was right. If the assistant filled any field that was blank in the documents, tighten Rule 2 and retest. If the sheet found something the listing missed, such as a lease clause the listing ignored, you have the best evidence for rolling it out to the whole team.

After that, spot check one live file a month. Pick a listing at random and ask someone who was not on the case to trace each Part A, B and C item back to its source in the tracker. A good file lets them do it in ten minutes. A bad file shows you where the process leaks.

Questions agents ask about material information and AI

Can AI fill in the material information fields for me?

AI can read documents you already hold, such as a lease, an energy certificate or a vendor questionnaire, and pull out the values. It must not supply a value from nothing. Every field should trace to a document, an official search or something the vendor confirmed, with a source noted beside it.

Should the vendor's questionnaire answers be taken at face value?

Not for items you can check cheaply. Council tax band, tenure and broadband speeds can be verified from public records or checker tools. Use the questionnaire for things only the owner knows, such as past flooding or shared access, and record that the answer came from the vendor.

Is it safe to paste a lease into a chatbot?

Only on a business plan whose terms say your content is not used to train models, and only with your firm's approval. Leases contain names and addresses. Where that is not possible, type the three or four figures you need into the prompt and keep the document itself out.

What if a Part C issue is unknown when I list?

Write 'not yet established' in your tracker, set a date to resolve it and do not publish a statement that the issue does not exist. Check whether the government's promised material information guidance has been published, because it is expected to say what must be shown in the listing and what can follow later.

Further reads

• How Estate Agents Use AI to Prepare Valuation Appointments — Start the material information file before the valuation visit.
• ChatGPT for Estate Agents: Rules for Vendor Updates and Feedback — Keep vendors informed while you wait for searches and documents.
• AI Receptionist for Estate Agents: Cost, Call Flow and Handover — Route enquiry calls so buyers' questions reach the right file.
• Can an AI Chatbot Book Valuations for Estate Agents Out of Hours? — Capture early property details from out-of-hours enquiries.
• Can Estate Agents Use AI Photo Edits Without Misleading Buyers? — Stop edited listing photos from misleading buyers.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: National Trading Standards material information page (live, and the May 2025 copy recording the withdrawal); the withdrawn NTSELAT Material Information in Property Listings (Sales) guidance, November 2023; Propertymark notice on the withdrawal, May 2025; government home buying and selling reform roadmap and material information consultation outcome, June 2026; Microsoft Learn enterprise data protection page.