Artificial intelligence can assist with research, data organization, image analysis, market comparisons, financial modeling, and report drafting. These capabilities may change how an assignment is performed, but they do not change who is responsible for the professional opinion.
The American Society of Appraisers’ Statement on AI, effective January 28, 2026, says appraisers remain responsible for confirming and verifying information necessary for credible results. It also states that opinions and conclusions in an appraisal remain those of the appraiser, rather than the sole result of AI analysis.
A useful way to manage that responsibility is to create an accountable judgment layer between an AI output and the reported conclusion. This layer consists of the appraiser’s testing, verification, professional skepticism, documentation, and decision to accept, modify, or reject the output.
The phrase “judgment layer” comes from an October 2026 RICS insight paper. The paper uses the term to describe the professional reasoning that separates an AI generated output from professional advice.
The RICS paper is an insight document, not a universal requirement for appraisers. It was written for surveyors, firms, educators, and other professional stakeholders. Its central idea has broader value: an appraiser should be able to explain what an AI system produced, how that result was assessed, and why the final conclusion is credible for the assignment.
An AI tool may identify potential comparables, extract information from documents, estimate a condition category, summarize a market, or draft narrative language. None of those functions establishes accuracy, relevance, or compliance with the assignment’s scope. The appraiser must decide how much reliance is appropriate.
AI controls should begin with the assignment, rather than with the software.
A business valuer may use AI to organize financial information or identify potential comparable companies. A machinery appraiser may use image analysis to flag equipment characteristics for further inspection. A personal property or gems and jewelry appraiser may use digital tools to assist with object identification or market research. A real property appraiser may use an automated system to sort comparable data or analyze geographic information.
The important question is whether the AI use affected the appraisal analysis, evidence selected, assumptions applied, or report delivered to the client.
A practical first step is to classify the use as administrative, analytical, or conclusion affecting. Administrative uses may include formatting or scheduling. Analytical uses may include research, data extraction, or pattern identification. Conclusion affecting uses may influence adjustments, forecasts, condition assessments, model outputs, or the final opinion.
The greater the effect on the conclusion, the greater the need for testing, documentation, and careful consideration of disclosure.
An appraiser should test an AI system before relying on it for an assignment.
Consider a hypothetical machinery assignment. An image analysis tool identifies a specialized production line as a newer model with a higher market value. The appraiser compares that result with the serial plate, maintenance records, manufacturer documentation, and a physical inspection. The evidence shows that the equipment is an earlier model with a different capacity.
The appraiser rejects the classification, records the reason, and continues the analysis using verified information. The system helped direct attention to the asset, but it did not establish the asset’s identity or value.
Testing should include representative examples and known limitations. A language model may generate unsupported statements or citations. A predictive model may produce a precise result from incomplete or outdated data. An image tool may misclassify an object when photographs are poor or the asset has unusual characteristics.
The appraiser’s workfile should record the task assigned to the system, the material output, the verification performed, and the decision reached. Rejected outputs may also be worth documenting when they reveal a recurring limitation or material risk.
The judgment process will differ across ASA disciplines.
In Business Valuation, verification may focus on management projections, comparable company data, capitalization or discount rates, and the consistency of model assumptions.
In Machinery & Technical Specialties, the appraiser may need to verify equipment identity, condition, operating status, technical specifications, and the relationship between machinery and real property.
In Real Property, automated comparable selection or market modeling should be checked against property characteristics, location, market conditions, and the intended use of the assignment.
In Personal Property, evidence may involve provenance, authenticity, condition, rarity, market venue, and the reliability of auction or dealer information.
In Gems & Jewelry, technology assisted identification does not eliminate the need to assess treatments, grading information, laboratory reports, workmanship, provenance, and market context.
In Appraisal Review & Management, the focus shifts to whether the workfile explains the technology’s role, whether important inputs can be traced, and whether the reviewer can distinguish verified evidence from unsupported output. Practical review questions include: What information came from the system? What independent evidence supports it? What did the appraiser accept or reject? Is the reasoning documented clearly enough for another qualified professional to follow?
The ASA Statement on AI says that direct use of AI in research or analysis should be addressed in the appropriate section of the report when applicable. It offers sample disclosure language, but expressly identifies that language as recommended rather than required. The statement also says appraisers should use their own judgment about when to disclose AI use and how much effort is appropriate for confirming derived information.
For report writing, the ASA statement says appraisers must diligently proofread AI generated material to ensure it is accurate and appropriate. It also offers a recommended, not required, example of disclosure for AI assisted report writing.
RICS has issued separate materials on responsible AI use. Its October 2026 journal commentary discusses the RICS responsible AI professional standard and describes obligations for RICS members and regulated firms. Those provisions should not be presented as requirements for ASA members or as universal appraisal rules. ASA members should follow the standards, laws, client agreements, and professional guidance applicable to their assignment.
AI may help an appraiser locate evidence, organize information, identify patterns, or draft language. The professional opinion remains the appraiser’s responsibility.
An accountable judgment layer makes that responsibility visible. It connects the assignment question to the data used, the output generated, the verification performed, and the conclusion reported.
The objective is not to accept or reject AI as a category. The objective is to apply professional judgment with enough discipline that an appraiser can explain what the system did, what the appraiser did, and why the final opinion is credible for its intended use.