What an appraiser remains responsible for when software helps

Does USPAP still apply when software does part of the work?

Yes. USPAP applies to the appraiser, and it keeps applying when a tool performs a step.

The Appraisal Foundation states that compliance with USPAP is required for state licensed and state certified appraisers performing appraisals for federally related transactions. On April 23, 2026 the Appraisal Standards Board adopted Advisory Opinion 41, Use of Technology in an Appraisal or Appraisal Review Assignment. The Foundation describes it as guidance on appraisers' USPAP responsibilities when using tools such as automated valuation models, regression and statistical software, and generative artificial intelligence.

Advisory opinions are guidance, not new standards. The Foundation says Advisory Opinion 41 will be incorporated into a future edition of the USPAP Guidance and Reference Manual. Read it in full, not through a summary (this one included).

Which decisions cannot be handed to a tool?

The decisions that form the opinion: scope of work, comparable selection, adjustments, reconciliation and the final value.

Fannie Mae's Selling Guide puts it plainly for comparables: the appraiser is responsible for determining which sales are the best and most appropriate for the assignment. Its list of unacceptable practices includes adjustments that do not reflect market reaction and adjustments that are not supported.

A tool can rank candidates or fit a regression. Accepting its output is still a choice, and the appraiser must explain it without pointing at the software.

How well does the appraiser need to understand a tool?

Well enough to judge whether its output is credible for the assignment at hand.

USPAP's Competency Rule asks the appraiser to have the knowledge and experience to complete the assignment competently, and the methods used are part of the assignment. So ask of each tool: what data it draws on, how current that data is, how it treats outliers, whether it shows its inputs, and how it behaves when the subject is unusual. If those questions have no answer, the output is a lead, not evidence.

Who is responsible for text that software drafts?

The signing appraiser is, sentence by sentence.

Drafting tools write neighborhood descriptions and comments fast. Fannie Mae lists subjective terms such as "desirable neighborhood" and any reference to crime or protected classes as unacceptable. A model that learned from older reports can reproduce exactly that language, so generated text needs reading as if the appraiser wrote it, because under the certification they did.

People are a separate question. The Selling Guide requires the signing appraiser to acknowledge significant professional assistance from others, such as an employee doing market research or data verification, and to state the tasks each person performed.

What should the workfile show?

Enough for a reviewer to follow what the tool did and what the appraiser decided.

Usually that is the tool's name, the date it ran, its inputs, its output, and a note on what the appraiser accepted, changed or set aside. The page on checking imported data suggests keeping imported and reported values in separate fields for the same reason.