Navigating USPAP Compliance in the Age of Artificial Intelligence: AO-41

More than 18 years ago, I took my first Uniform Standards of Professional Appraisal Practice (USPAP) course on my way to becoming an accredited personal property appraiser. A USPAP-compliant appraiser is distinguished from other valuation professionals and market experts by the procedures, guidelines, and ethical standards for appraisal detailed in the USPAP manual. USPAP is an enforceable set of rules that regulate the appraisal profession and apply to all appraisal disciplines, including personal property, real estate, and business valuation. Professional appraisal associations like the International Society of Appraisers, American Society of Appraisers, and the Appraisers Association of America have the legal authority to enforce USPAP compliance by their members. To remain USPAP compliant, appraisers must complete a 7-hour USPAP update course every two years. This year, a central topic of our course was Advisory Opinion 41, "Use of Technology in an Appraisal or Appraisal Review Assignment," adopted by the Appraisal Standards Board. Unsurprisingly, Artificial Intelligence is a hot topic in the appraisal community, and AO-41 clarifies how appraisers can use this groundbreaking technology while remaining USPAP compliant.

The Core Principle: It is the Appraiser Who Adheres to USPAP - Not the Tool

AO-41 shows how the ETHICS, COMPETENCY, SCOPE OF WORK, and RECORD KEEPING rules apply once a technological tool enters the picture. A tool's output is information — not assignment results — until the appraiser verifies it, applies judgment, and decides whether to rely on it. Whether the technological tool is generative AI, statistical software, or a drafting app, it can inform the analysis. It cannot replace it.

Where Technology Shows Up in a Personal Property Assignment

Personal property appraisers don't rely on real-estate-style pricing models, but technology still touches most of our work:

• Researching — auction and sales databases, and AI-based image recognition for comparables and provenance research

• Analyzing — spreadsheets and statistical tools applied to sales data

• Drafting — generative AI to structure reports or check consistency, and grammar tools like Grammarly

AO-41 treats all of these as aids, never a substitute for the appraiser's own analysis.

A Personal Example: AI-Assisted Provenance Research

I ran into this firsthand while researching a work of Chinese calligraphy from the early-to-mid 20th century. The piece carried a dedication that hinted at its provenance but left out a surname — not enough on its own to establish who the work was made for. My clients were uncertain which grandparent had acquired the calligraphy and other paintings in the collection. One grandfather had practiced medicine and been a high-ranking official in China and later Taiwan during that same period. So, working from that profile, I queried AI about historical figures who matched it, aiming to establish a plausible connection between them. It took a series of queries and several false starts before that research pointed me toward digitized archives. Those archives let me draw a credible link between the dedication and my client's grandfather — the kind of documented evidence a provenance claim actually needs, not just a plausible-sounding AI suggestion.

That's the distinction AO-41 is getting at. The AI helped me search faster and consider connections I might not have found on my own. It didn't establish the provenance — the archival record did, and only after I verified it myself.

Judgment Still Belongs to the Appraiser

AO-41 centers on three elements: technology, the appraiser, and assignment results. Only the appraiser produces assignment results — a tool's output, however sophisticated, is not that. Presenting a tool's output as your own conclusion risks an ETHICS RULE violation; relying on it without understanding its limits risks a COMPETENCY RULE violation.

When a Client Restricts or Requires a Specific Tool

Clients may prohibit certain technologies or require a specific one — AO-41 treats both as assignment conditions. Either way, the scope of work still has to support credible results. If it can't, the appraiser must decline, withdraw, or renegotiate — not proceed anyway.

What Belongs in the Workfile

The level of documentation should correspond to the degree of reliance on a tool's output. If your opinion was based on a tool's output, specify the tool, the prompts employed, and your rationale. If you reviewed the output but didn't depend on it, a brief explanation of why is sufficient. Tools used solely for formatting or grammar do not require documentation, as they did not influence the analysis.

Confidentiality Hasn't Changed — Only the Context

The confidentiality protections of the ETHICS RULE are fully applicable to any technological tool utilized. Prior to inputting information into a generative AI system, ensure client consent through the Letter of Agreement. Confirm that the tool will not retain or disclose the information without permission. If this cannot be verified, do not input the information.

The Takeaway

AO-41's message is that credibility rests with the person, not the process. Technology can expand how we research, analyze, and communicate — but it can't carry the responsibility behind the words "in my opinion." That's still ours.

[Note: This blog post and quick reference guide were created in conjunction with Claude.ai ]