Note · 2026-10-08

Can a valuer use AI under IVS and EVS?

Yes. Neither IVS 2025 nor EVS 2025 prohibits artificial intelligence; the current IVS do not mention it. Both draw the line not at the technology but at whether the valuer applies professional judgement: under IVS, no model without that judgement can produce an IVS-compliant valuation, and under EVS any model remains a tool. My conclusion is that the question is not whether to use AI, but whether the valuer can explain, check and disclose what it contributed.

Source: IVS 105, introductory statement; EVS 2025, Part II, para. 6.9.

What IVS say

The IVS Glossary defines an automated valuation model (AVM) by the absence of judgement: “a type of model that provides an automated calculation … without the valuer applying professional judgement over the model, including assessing, and selecting inputs or reviewing outputs”. IVS 105 states: “No model without the valuer applying professional judgement, for example an automated valuation model (AVM), can produce an IVS-compliant valuation.”

Source: IVS Glossary, para. 10.02; IVS 105, introductory statement.

The specific requirements are:

• the valuer is responsible for assessing and selecting the data, assumptions and adjustments to be used as inputs, on the basis of professional judgement and professional scepticism;
• the valuer must understand the way the valuation model operates; transparency (all persons preparing and relying on the model understand how it works and its inherent limitations) is one characteristic of an appropriate model, and where a model lacks some characteristics the valuer must assess and conclude that it is appropriate;
• significant limitations or adjustments must be explained, justified and documented; if they cannot be sufficiently justified, the valuation would not comply with IVS;
• a model or data may be obtained from a specialist or service organisation: before using one, the valuer must assess and document its capabilities, and afterwards obtain an understanding of the process and findings to establish a reasonable basis to rely on its work; its involvement must be agreed and disclosed;
• the report must convey the valuation methods or models applied, the sources and selection of significant data and inputs, and the findings of any specialist or service organisation.

Source: IVS 100, paras 30.01⁠–⁠30.03; IVS 104, paras 10.04 and 20.01⁠–⁠20.02; IVS 105, paras 10.03, 20.01⁠–⁠20.02, 30.01(d), 30.02, 40.03 and 40.06⁠–⁠40.07; IVS 106, para. 30.06(k), (l) and (o).

What EVS say

EVS 2025 contain no separate standard or guidance note on AVMs. The basic position is in the methodology text of Part II: the valuer should be aware that any analytical tool is only as reliable as the accuracy and quality of the data fed into it, and the value of a property cannot be calculated by mathematical or statistical techniques alone. All valuation models are simply tools, and the valuer’s estimate of value has to be based on best and sound judgement.

Source: EVS 2025, Part II, para. 6.9.

Information paper EVIP 2 Portfolio Valuation adds that statistical techniques, statistical tools and AVMs are mathematical aids to calculation and do not offer the professional judgement needed to report on value; that is the valuer’s task and responsibility.

Source: EVS 2025, EVIP 2, para. 2.7.

Of particular relevance to Ukraine is guidance note EVGN 1 on applying EVS in wartime circumstances. Where values have been derived using AVMs, asking prices or other valuation reports, care should be taken, when making adjustments to arrive at Market Value, to understand how remote the figures are from real evidence of actual market behaviour. The extent of use of each category of evidence (direct comparables, general market data and indexes, other sources) and the opinion of its reliability should be disclosed in the report.

Source: EVS 2025, EVGN 1, paras 4.14⁠–⁠4.15.

How RICS does it

Under the Red Book 2025, a valuation based wholly or partly on the output of, among others, an AVM or a model or process assisted or produced by AI counts as a written valuation only if a valuer has additionally applied professional judgement in accordance with the standards’ mandatory requirements. The use of open-source or commercially available AI is not prohibited, subject to professional judgement, terms of engagement, investigations, reporting and records appropriately and proportionately considering confidentiality, intellectual property, data and input verification, assessment of process and model outputs, transparency with intended users, and other ethical, technical and legal matters.

Source: RICS Red Book 2025, PS 1, paras 1.3⁠–⁠1.4.

RICS’s separate standard Responsible use of artificial intelligence in surveying practice has been in force since 9 March 2026 and is mandatory for RICS members and regulated firms. It applies to AI outputs that have a material impact on the delivery of a service and turns these principles into procedures. The client is informed in writing, in advance, of when and for what purpose AI is to be used, and this is reflected in the terms of engagement. The decision on the reliability of an output is recorded in writing and prepared by, or under the supervision of, an appropriately qualified and named surveyor who accepts responsibility for its use. Where AI automates outputs or produces them in high volume, scrutinising each output and making a written decision on it is generally not necessary, but randomised dip samples at regular intervals are mandatory.

Source: RICS, Responsible use of artificial intelligence in surveying practice, 1st edition, September 2025, sections 1.2, 1.3, 4.2 and 4.3.

What the IVS 2028 draft changes

The draft IVS, proposed to take effect on 31 January 2028, would:

• require the scope of work to indicate any significant proposed use of AI or other technology-based tools whose logic the valuer cannot readily explain or verify;
• subject data, inputs and models obtained with such tools to quality controls, with the valuer remaining ultimately responsible for IVS compliance;
• require documentation to include any significant use of AI and the related quality controls, and the report to convey the significant use of AI or other technology-based tools.

The draft drops the AVM definition; the principle moves to IVS 105, para. 10.07: “No valuation model can produce an IVS-compliant valuation without the application of the valuer’s professional judgement and professional scepticism.” This is a draft, not a standard in force.

Source: draft IVS 2028, IVS 100, para. 50.01; IVS 101, paras 20.01(k) and 20.02; IVS 104, paras 10.05⁠–⁠10.06; IVS 105, paras 10.06⁠–⁠10.08; IVS 106, paras 20.03 and 30.06(n).

A practical minimum under Ukraine’s NVS 1

Under Ukraine’s National Valuation Standard No. 1 (NVS 1), the valuer independently searches for and analyses sources of information and presents the source data collected in the report with a reference to where they were obtained.

Source: NVS 1, paras 52 and 54.

My conclusion, not a rule: a generative language model’s answer is not itself a source of market data within para. 54. The model produces plausible text rather than reproducing transaction data. If it helped locate data, the report cites the primary source that the valuer opened and checked.

Under NVS 1, disclosure needs no separate rule; it fits into the existing sections of a full-form report. The tool and the limits of its use are described in the analysis of source data, the model’s part in the calculation in the account of valuation procedures and calculations, and the reliability of data obtained this way in the valuer’s written statement on the quality of the source data used.

Source: NVS 1, para. 56.

A working disclosure template (my wording, not a rule) fits into three sentences: name the tool and its version; state what it was used for and what it did not do; confirm that the source data were checked against primary sources and that the conclusion of value was formed by the valuer. For RICS members and regulated firms, where AI has a material impact on the service, a written decision on the reliability of the output and advance written notice to the client are also required.

More on this in my paper at the 36th research and practice conference of the Ukrainian Society of Appraisers (UTO), September 2026, and the accompanying abstract.

For how this plays out in practice, see the contract ‘The Shapeshifter’ in my AI lab: the executing agent checked its own work and reported ‘6 of 6’, while an independent check found descriptions of requirements in place of quotations and two references to IVS 104 and 105 instead of IVS 102 and 103.