Aipedia: Surveyors and valuers

AI and comparable evidence: bias, verification and scepticism

Aipedia, by the Apex broking team · Last reviewed 2026-10-27

AI tools can find and adjust comparables quickly, but the valuer must still verify the evidence and judge whether the selection is sound. Comparable evidence is the backbone of most valuations, so a biased or unverifiable set can undermine the whole figure.

Part of: AI in surveyors and valuers

In short

AI comparable tools search transaction and listing data, rank similar properties and suggest adjustments. The risks are hidden bias in the dataset, adjustments the valuer cannot explain, and evidence that cannot be checked. The RICS AI standard requires baseline knowledge of bias (section 2), supplier due diligence on dataset quality and bias (4.1), and professional judgement and scepticism with a written reliability decision (4.2). Valuation claims based on poor comparables are usually treated as negligence under RICS-compliant PI, subject to the terms.

How valuers use AI for comparables

Comparable selection used to mean searching sales records, phoning agents and building a short schedule by hand. AI tools now return ranked lists of similar properties in seconds, suggest weights for size, condition and location, and in some cases propose an adjusted value range. Commercial valuers use similar tools for rents and yields. The convenience hides a question the valuer must still answer: why these comparables, and are they true?

The duties that apply

Section 2 baseline knowledge: members must understand AI limitations, erroneous output and bias. A valuer who cannot describe how a tool chose its comparables is exposed.
4.1 due diligence: ask the supplier in writing about dataset quality and bias, data permissions and provider liability, record the answers, and log gaps as risks.
4.2 scepticism: apply professional judgement, document the reliability decision under a named surveyor, and tell the client if the output cannot reasonably be used.
4.4 explainability: on request, explain the system type, how it works and its limits.

These sit on top of the Rules of Conduct and, for Red Book work, the Red Book, which RICS says now covers modelling, data management and transparency in technology use. The RICS AI standard is mandatory from 9 March 2026.

Where bias and bad data come from

Coverage bias: datasets may be thin for rural, unusual or high-value property, so the tool fills gaps with weaker matches.
Listing data: asking prices and agent descriptions can creep in alongside completed sales.
Stale or duplicated records: the same sale can appear twice, or a resale after refurbishment can be treated as like-for-like.
Opaque adjustments: a tool may adjust for condition or specification using assumptions the valuer cannot see.
Feedback loops: if past valuations feed the dataset, earlier errors can be reinforced.

What can go wrong

Illustrative: a valuer uses an AI tool to select five comparables for a converted rural property. The tool picks recent sales of modern houses in a nearby town, applies a location adjustment and returns a tidy range. The valuer adopts the mid-point. Two years later the lender sells after default at a much lower figure and alleges the comparables were unsuitable. The valuer cannot say why the tool chose them or what the location adjustment was.

A court is likely to ask whether a reasonably competent valuer would have selected and relied on that evidence; see the standard of care for valuers and our valuation negligence scenario. A file with no verification of the comparables makes that harder to answer.

How PI usually responds

A claim that a valuation was negligent because of poor comparable evidence is usually handled as professional negligence under the RICS Approved Minimum Wording, subject to the policy terms. Using a data tool does not by itself restrict cover. Insurers and their lawyers will look closely at the valuer's own checks, and at whether the reliability decision required by 4.2 exists. Limits should reflect your largest lending exposures, not only the RICS minimum.

Practical controls

Verify each comparable you rely on against a primary record or an agent you can name.
Write down why you kept or rejected each tool suggestion.
Recalculate key adjustments yourself rather than accepting the tool's, and note where your figure differs.
Prefer fewer, verified comparables over a long list you cannot check.
Record coverage gaps and bias questions in your 3.3 risk register.
Check measured areas behind the comparables; see AI measurement and floor areas.
Link the comparable check to your AVM controls.

Apex arranges PI for valuers and can help you check that your limit and wording suit your lending and Red Book work. See the AI in surveying and valuation hub and the compliance checklist.

Sources

Frequently asked

Do I have to verify every comparable an AI tool suggests?

You must verify the ones you rely on. Standard 4.2 requires professional judgement and scepticism and a written reliability decision. In practice that means checking the evidence you use against records you can trust and recording why you accepted or rejected suggestions.

What should I ask a comparable data supplier?

Standard 4.1 sets minimum topics: compliance with data law, permissions to use the data, dataset quality and bias, environmental impact and provider liability. Send the questions in writing, record the answers, and log any gaps as risks in your quarterly reviewed risk register.

Can AI tools introduce bias into valuations?

Yes. Datasets can be thin for unusual property types or areas, mix asking prices with completed sales, or reinforce earlier valuations. Section 2 of the RICS AI standard requires members to understand bias, and section 3.3 requires it to be rated in your risk register and reviewed at least quarterly.

Can a client ask how the AI chose the comparables?

Yes. Under standard 4.4 you must, on request, provide written information on the type of AI system, how it works and its limits, your due diligence, your risk management and your reliability decisions. Keep notes that let you answer that clearly.

Check your PI covers how you use AI

Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.

Start a PI quote Request a callback

Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.