Technology may assist
The Appraisal Foundation says AI may work alongside other tools, but technology does not substitute for professional judgment. An appraiser must understand whether an analytical output is credible before using it.
Short answer: use AI like a controlled quality-control clerk. It may organize approved inputs, compare fields, flag discrepancies, draft neutral clarification questions, and check whether a statement has a visible source. The appraiser keeps inspection judgment, final comparable selection, adjustments, condition and quality conclusions, market analysis, reconciliation, value opinion, certification, and signature.
Professional and confidentiality boundaryThis educational guide does not replace USPAP, its current Guidance and Reference Manual, state law, credential requirements, client instructions, lender overlays, assignment conditions, or legal advice. Do not place confidential information or assignment results into an AI service unless the client or firm has authorized the system and the data handling has been reviewed.
What the official sources make clear
Responsibility does not transfer
The appraiser still owns competency, confidentiality, nondiscrimination, recognized methods and techniques, credible results, and compliance with the laws and requirements that apply to the assignment.
Current standards checkThe Appraisal Foundation identifies the 2024 edition as the current USPAP edition. Its FAQ says the edition took effect January 1, 2024 and has no end date. The companion Guidance and Reference Manual is updated separately. Always recheck before relying on a saved workflow.
Build the six-part Appraisal Evidence Desk
Set the assignment boundary
Record intended use, intended user, property type, client instructions, credential/competency check, applicable overlays, approved tools, and prohibited data before opening an AI tool.
Output: assignment boundary card
Name every source
Create an inventory with the exact source name, access date, authorized use, and whether independent verification is required. Avoid broad labels such as “public records” when the specific source is known.
Output: approved-source inventory
Build a factual field ledger
Place subject and candidate-sale fields beside their source IDs. AI may compare exact entries and flag conflicts. It may not decide which source is credible or which sale is comparable.
Output: source-linked data ledger
Route discrepancies, not conclusions
Turn missing or conflicting facts into neutral verification questions. Close each item only with a source or documented professional resolution.
Output: discrepancy queue
Run narrative and bias QA
Check for unsupported adjectives, demographic or protected-class references, proxy language, inconsistent measurements, unexplained search expansion, missing verification, invented citations, and conclusions that are not visibly supported.
Output: report QA sheet
Issue a licensed-review receipt
The appraiser reviews every AI-assisted output, resolves discrepancies, makes all valuation decisions, confirms confidentiality and nondiscrimination checks, and signs only after the assignment is complete.
Output: professional review receipt
Give AI bounded jobs, not appraisal authority
| Job | Approved input | AI output | Human gate |
|---|---|---|---|
| Intake clerk | Blank checklist plus approved assignment metadata | Missing-item list and organized intake card | Appraiser confirms scope, competency, and assignment conditions |
| Source-control clerk | Source names and permitted factual fields | Source inventory and blank traceability table | Appraiser decides reliability and required verification |
| Discrepancy checker | Approved, source-linked field table | Mismatches and neutral follow-up questions | Appraiser resolves every conflict |
| Narrative QA clerk | Redacted or approved report draft | Unsupported-language, consistency, and missing-source flags | Appraiser rewrites, validates, or rejects each flag |
| Policy-update clerk | Current public first-party releases | Dated change summary with direct links | Qualified staff checks applicability before changing practice |
Do not delegate these decisions
AI may prepare
- blank assignment checklists
- approved-source inventories
- field-by-field discrepancy tables
- neutral verification questions
- missing-source and internal-consistency flags
- public policy update summaries with direct links
- redacted administrative templates
Appraiser must control
- assignment acceptance and competency
- inspection observations
- source credibility and final comparable selection
- adjustment method and amount
- condition and quality conclusions
- market-area analysis
- reconciliation and value opinion
- certification, signature, and ROV response
Comparable preparation needs a source trail
For covered Fannie Mae assignments, the Selling Guide says the appraiser is responsible for determining the best and most appropriate comparables. It also requires specific data and verification sources and explains that interested-party information may require verification through a disinterested source. That is why this workflow produces a candidate evidence table, not an AI-selected final comparable set.
Safe distinctionA computer can sort candidate records by explicit fields. Only the appraiser can determine whether those records are credible, competitive, comparable, and appropriate for the assignment, then explain the reasoning and any market-area expansion.
Prepare for UAD 3.6 without guessing
Fannie Mae and Freddie Mac entered broad production for UAD 3.6 in January 2026. Current first-party transition materials say all new appraisal reports submitted to UCDP must use UAD 3.6 beginning November 2, 2026. Build a change packet from the official specification, client or lender instructions, vendor release notes, test results, and staff review. Do not ask a general chatbot to invent a field mapping.
Use this prompt only with approved inputs
Act as a quality-control clerk, not an appraiser. Compare the supplied fields exactly as written. Create: (1) a mismatch table, (2) a missing-source list, (3) neutral clarification questions, and (4) a list of statements that lack a visible source. Do not select comparables, make adjustments, infer condition or quality, analyze protected characteristics or demographic proxies, reconcile value, or propose a value conclusion. Do not invent missing facts. Mark uncertainty clearly. If any input appears confidential, stop and label it for approved-system review.
Final review before any report leaves the desk
First-party source desk
The Appraisal Foundation: practicing appraisers and AI Q&AProfessional judgment, credibility, confidentiality, competency, and nondiscrimination when using AI.
The Appraisal Foundation: current USPAP editionCurrent standards edition and official publication information.
Fannie Mae Selling Guide B4-1.3-07Specific data and verification sources, reliable sources, and prior-sale history.
Fannie Mae Selling Guide B4-1.3-08Appraiser responsibility for comparable selection and explanation.
Fannie Mae: UAD 3.6 transition noticeNovember 2, 2026 mandate for new UCDP submissions.
Appraisal Subcommittee: USPAP compliance and appraisal independenceRecognized methods, independence boundaries, corrections, and additional information.
CFPB: final interagency ROV guidanceErrors, omissions, unsupported analysis, discrimination risk, and independence-aware review.
Appraisal Subcommittee national registriesState-authorized appraiser and AMC registry information.
Build the desk before choosing another AI tool.
Download the worksheet, use fictional or approved nonconfidential examples, and have a qualified appraiser review the workflow against actual state, client, and assignment requirements.