- Official section
- National III: Valuation
- Broker weight
- 8% of the national broker portion
- Expected scored items
- Valuation accounts for about 8 of 100 items
Illinois exam glossary
Automated valuation model
An AVM can price thousands of properties before a person finishes one inspection. That speed is useful, but it can hide a basic truth: the model knows only the data it receives and the relationships it was built to recognize. A strong student reads the estimate, the confidence, the use case, and the missing facts together.
Last updated: August 1, 2026
What does this exam area cover?
Short answer: An AVM is a computerized model that estimates property value from data and mathematical relationships. It is fast, scalable, and useful for screening, monitoring, and some covered credit or program decisions, but it does not observe hidden condition, explain every transaction, or become an appraisal merely by producing a precise number. Users must evaluate data quality, model fit, uncertainty, validation, bias, intended use, and applicable appraisal or evaluation rules.
This guide uses the Illinois appraiser-license statute, the interagency AVM rule effective October 1, 2025, current Regulations B and Z, April 2026 USPAP technology guidance, and Fannie Mae Selling Guide updates through June 3, 2026, all checked through August 1, 2026. The federal rule's five quality-control factors apply to defined mortgage-originator and secondary-market uses. They are also a useful study framework outside that coverage, but this page does not represent them as a universal legal mandate for every model or estimate.
What is on the official outline?
- Topic
- Define the AVM use case
- What to know
- Consumer estimate, listing lead, broker support, mortgage credit decision, securitization, portfolio monitoring, servicing, home-equity review, tax assessment, fraud screen, collateral surveillance, prequalification, appraisal waiver, and prohibited substitution
- Best exam move
- The same model output can be acceptable for screening and inadequate for a decision that requires an appraisal.
- Topic
- Identify the subject correctly
- What to know
- Address normalization, parcel, unit number, legal description, condominium, cooperative, manufactured home, lot, multi-parcel property, building, accessory unit, ownership, geocode, and duplicate record
- Best exam move
- A model attached to the wrong parcel or unit can be confidently wrong before analysis begins.
- Topic
- Understand input data
- What to know
- Recorded sale, deed, assessor, tax, parcel, MLS, listing, permit, building characteristic, geospatial feature, census geography, price index, rental data, imagery, prior valuation, user input, and proprietary database
- Best exam move
- Ask where each material input came from, how current it is, and whether the model can verify it.
- Topic
- Separate price from property facts
- What to know
- Sale price, concession, financing, personal property, related party, distress, transfer type, contract date, closing date, arms-length status, renovation timing, condition, and data correction
- Best exam move
- A recorded price can train the model even when the transaction is not normal market evidence unless data controls identify it.
- Topic
- Recognize model methods
- What to know
- Hedonic regression, repeat-sales index, comparable algorithm, decision tree, gradient boosting, neural network, ensemble, spatial model, time adjustment, feature engineering, outlier handling, and calibration
- Best exam move
- You do not need to code the model, but you must know that method assumptions influence which patterns it learns.
- Topic
- Read the output
- What to know
- Point estimate, range, low and high bound, confidence score, forecast standard deviation, error band, effective date, model version, comparable list, input summary, warning, no-hit, and insufficient-confidence result
- Best exam move
- Never separate the estimate from its date, uncertainty, and vendor-specific score definition.
- Topic
- Measure accuracy
- What to know
- Prediction error, absolute error, percentage error, median error, mean error, root mean square error, hit rate, coverage, overvaluation, undervaluation, tail error, calibration, benchmark, and actual sale
- Best exam move
- A low average error can still conceal large failures for unusual properties or underserved locations.
- Topic
- Test geographic coverage
- What to know
- Urban, suburban, rural, county-record quality, non-disclosure state, sale density, new subdivision, distressed area, rapidly appreciating market, declining market, boundary effect, micro-location, and spatial drift
- Best exam move
- Validate performance where the subject is located instead of relying only on a national accuracy statistic.
- Topic
- Test property-type coverage
- What to know
- Typical detached home, condominium, cooperative, manufactured home, two-to-four unit, luxury, acreage, waterfront, historic, mixed use, unique design, new construction, renovation, accessory unit, and special rights
- Best exam move
- A model trained on ordinary homes can be unreliable at the edge of its property population.
- Topic
- Monitor data freshness
- What to know
- Last sale, current listing, permit, remodel, addition, demolition, fire, flood, zoning change, new road, assessment update, market shift, interest rate, model refresh, data latency, and effective date
- Best exam move
- Recent market data cannot fix stale subject facts, and fresh subject facts cannot fix an outdated market model.
- Topic
- Control manipulation
- What to know
- Input override, user-supplied area, cherry-picked comparables, duplicate sale, fabricated renovation, data poisoning, vendor access, change log, permission, approval, anomaly detection, cybersecurity, and audit trail
- Best exam move
- The current federal rule requires controls designed to protect covered AVMs from manipulation of data.
- Topic
- Avoid conflicts of interest
- What to know
- Vendor incentive, loan approval, volume target, transaction compensation, model selection, override authority, property interest, affiliated provider, adverse selection, governance, independent validation, and escalation
- Best exam move
- Do not choose or tune the model merely because it produces the result needed to close a transaction.
- Topic
- Use random sample testing
- What to know
- Production sample, representative draw, manual review, appraisal comparison, sale comparison, back-testing, exception population, override review, frequency, documentation, corrective action, and vendor monitoring
- Best exam move
- Covered users need policies and controls that require random-sample testing and review, not only review of obvious failures.
- Topic
- Comply with nondiscrimination law
- What to know
- Protected class, proxy variable, redlining history, geospatial feature, coverage disparity, subgroup error, disparate outcome, representative data, fairness metric, reason code, human review, governance, and corrective action
- Best exam move
- Accuracy for the full portfolio does not prove equal accuracy across communities or protected groups.
- Topic
- Set human-review triggers
- What to know
- Low confidence, no hit, wide range, property mismatch, unique design, sparse market, rapid change, condition uncertainty, renovation, inconsistent area, outlier price, consumer dispute, fraud flag, and policy threshold
- Best exam move
- Escalate when the model is outside its strong coverage instead of forcing every property through automation.
- Topic
- Distinguish appraisal and evaluation
- What to know
- Appraisal requirement, federal exception, institution evaluation, state credential, appraiser analysis, AVM support, collateral policy, transaction value, residential threshold, commercial threshold, documentation, and review
- Best exam move
- An AVM can support an evaluation or decision only when the governing rule and policy permit that use.
- Topic
- Understand Illinois procurement
- What to know
- Section 5-5(h), procurement, model output, no appraisal license for procurement, no appraisal title, no appraiser certification, broker use, client disclosure, advertising accuracy, and separate regulated service
- Best exam move
- The Illinois exemption covers procurement of an AVM, not misrepresentation of the output as a certified appraisal.
- Topic
- Understand consumer copy rights
- What to know
- Regulation B, first lien, dwelling, written valuation, AVM report, promptly upon completion, three business days, timing waiver, electronic delivery, denied application, withdrawn application, and latest version
- Best exam move
- An AVM can be a valuation for copy purposes even though it is not an appraisal.
- Topic
- Separate value acceptance
- What to know
- Desktop Underwriter, eligibility offer, appraisal not required, prior appraisal data, one-unit property, occupancy, transaction type, $1 million exclusion, lender representation, four-month offer, property data option, and appraisal fallback
- Best exam move
- Value acceptance is a specific Fannie Mae loan-casefile option, not another name for every AVM.
- Topic
- Use technology in appraisal practice
- What to know
- Appraiser, AVM input, statistical tool, generative AI, AO 41, competency, understanding, reliability, bias, confidentiality, workfile, disclosure, reconciliation, and professional responsibility
- Best exam move
- Technology can assist the appraiser, but it does not own the assignment conclusion or professional responsibility.
Which distinctions produce the most mistakes?
- Terms
- AVM vs. appraisal
- Difference
- An AVM is a computerized value estimate. An appraisal is an appraiser's defined assignment, analysis, opinion, report, and professional responsibility.
- Question cue
- Model output versus professional appraisal assignment.
- Terms
- AVM vs. CMA
- Difference
- An AVM derives an estimate through a model and database. A CMA applies an Illinois broker's market analysis and expertise for a permitted brokerage purpose.
- Question cue
- Automated model versus broker analysis.
- Terms
- AVM vs. BPO
- Difference
- An AVM is automated. A BPO is a broker's written estimate or analysis of probable selling price with Illinois disclosures.
- Question cue
- Computer estimate versus licensed broker opinion.
- Terms
- AVM vs. evaluation
- Difference
- An AVM is a tool or output. An evaluation is an institution's documented collateral-value conclusion for a transaction where an appraisal is not required and may use an AVM as support.
- Question cue
- Input tool versus regulated decision product.
- Terms
- AVM vs. value acceptance
- Difference
- An AVM estimates value. Fannie Mae value acceptance is a DU eligibility offer allowing an eligible loan casefile to proceed without an appraisal under program conditions.
- Question cue
- Estimate versus loan-delivery option.
- Terms
- Point estimate vs. confidence score
- Difference
- The point estimate is the model's value result. The confidence score expresses model-specific reliability or uncertainty, not another property value.
- Question cue
- Price result versus reliability signal.
- Terms
- Confidence score vs. accuracy rate
- Difference
- A confidence score usually describes one estimate's model fit. An accuracy statistic measures performance across a validation sample under a defined metric.
- Question cue
- Property-level signal versus population performance.
- Terms
- Coverage vs. accuracy
- Difference
- Coverage is the share of properties for which the model returns an estimate. Accuracy is how close estimates are to a defined benchmark for the tested population.
- Question cue
- How often it answers versus how well it answers.
- Terms
- Back-test vs. live monitoring
- Difference
- Back-testing compares past predictions with later known outcomes. Live monitoring watches current data, drift, overrides, complaints, and exceptions as the model operates.
- Question cue
- Historical validation versus production control.
- Terms
- Model bias vs. individual error
- Difference
- An individual error is one incorrect estimate. Model bias is a systematic pattern of error or unequal performance associated with data, design, or deployment.
- Question cue
- One miss versus repeated directional pattern.
- Terms
- Data error vs. model error
- Difference
- A data error gives the model a wrong fact, such as area or parcel. A model error arises when the mathematical relationship fails even with correct inputs.
- Question cue
- Wrong input versus wrong inference.
- Terms
- Regulation B valuation vs. Regulation Z valuation
- Difference
- Regulation B's copy rule includes AVM reports as written valuations. Regulation Z's valuation-independence definition excludes an estimate produced solely by an automated model, while its newer AVM quality provisions apply separately to covered use.
- Question cue
- Read each regulation's own definition and purpose.
The M-O-D-E-L AVM audit
- Mission and mandate: identify intended use, user, property, effective date, transaction, program, regulatory coverage, appraisal or evaluation duty, decision threshold, and whether automation is permitted.
- Origin of data: trace parcel identity, sale terms, property features, condition proxies, permits, listing history, market indexes, geospatial inputs, update dates, vendor sources, missing fields, and user overrides.
- Design and domain: understand method, training and validation populations, property and geographic coverage, feature logic, outlier treatment, version, confidence definition, error metrics, and known limitations.
- Equity and errors: test manipulation controls, conflicts, random samples, subgroup and geographic performance, overvaluation and undervaluation, drift, consumer disputes, exception populations, and corrective action.
- Launch with limits: set human-review triggers, document selection and use, provide required copies, preserve audit trails, monitor production, retest after change, and never label the estimate as an appraisal or guarantee.
- Control
- High confidence
- Question
- Does the model produce credible estimates for this use?
- Evidence
- Validation, calibration, coverage, error, monitoring
- Control
- Manipulation protection
- Question
- Can data or output be improperly changed?
- Evidence
- Access, logs, anomaly checks, approvals, security
- Control
- Conflict avoidance
- Question
- Could incentives distort model choice or use?
- Evidence
- Governance, independence, escalation, vendor controls
- Control
- Random testing
- Question
- Are ordinary production cases reviewed too?
- Evidence
- Representative samples, review results, corrective action
- Control
- Nondiscrimination
- Question
- Does use comply with fair lending law?
- Evidence
- Subgroup testing, proxy review, outcomes, remediation
How do the rules work in scenarios?
The model uses the wrong condominium unit
Scenario: An AVM matches Unit 12B to the larger penthouse parcel 12P because both share the same street address. It returns $790,000 with high confidence, while verified 12B sales cluster near $430,000.
- The parcel and unit identity are wrong before the model applies valuation relationships.
- A confidence score cannot cure a subject-matching error outside the model's expected input.
- Identity validation and exception review should stop reliance on the output.
Answer: Reject the estimate and correct the property match before any further valuation use.
A major renovation is invisible to the model
Scenario: A home was fully renovated six months ago, but permits are incomplete and the assessor still shows the prior condition. The AVM estimates $510,000 while renovated competitive sales range from $620,000 to $650,000.
- The model may accurately process the data it has while missing a material current characteristic.
- Recent photographs, permits, invoices, and property observation can change the subject file.
- Human valuation review is appropriate when current condition is outside recorded data.
Answer: Do not treat the stale-data AVM as conclusive evidence of current condition or value.
A national accuracy rate hides rural errors
Scenario: A vendor advertises a 5 percent median error nationwide. In the subject's rural county, the model covers only 42 percent of homes and median error is 14 percent.
- National performance blends very different market and record environments.
- Local coverage and error are more relevant to the subject use.
- Sparse sales and diverse acreage can increase uncertainty and tail risk.
Answer: Use local validation and escalate to a more suitable valuation process.
Two scores cannot be compared without definitions
Scenario: Vendor A returns confidence 88 on a 0-to-100 scale. Vendor B returns confidence 0.92 based on a different algorithm and validation metric.
- The numbers use different scales and may measure different concepts.
- A larger-looking number is not automatically more reliable.
- The user needs each vendor's score definition, calibration, validation population, and threshold policy.
Answer: Compare documented performance, not the raw confidence numbers.
Random testing catches ordinary overvaluation
Scenario: A lender reviews only AVMs challenged by borrowers. Random production sampling reveals a repeated 7 percent overvaluation pattern in a fast-declining market where the index updates slowly.
- Complaint-only review misses errors that consumers do not identify.
- Random sampling tests ordinary production outputs.
- The lender should adjust controls, model use, monitoring frequency, and escalation for the affected market.
Answer: Use the sampling result for documented corrective action and revalidation.
Value acceptance is an eligibility offer
Scenario: DU returns a value acceptance offer for an eligible one-unit purchase below Fannie Mae's stated $1 million exclusion, and the lender obtains no appraisal.
- The option arises from the loan casefile and Fannie Mae program requirements.
- The lender must follow the current conditions, final DU message, four-month offer period, law, and its own risk information.
- The result is not proof that every public AVM or every future transaction can avoid an appraisal.
Answer: Treat value acceptance as a specific program option, not a universal AVM waiver.
An appraiser uses an AVM but owns the analysis
Scenario: An appraiser uses an AVM as one market check. The AVM is $40,000 above the appraiser's reconciled sales evidence because it selected remodeled properties unlike the subject.
- The appraiser evaluates the model's inputs, comparables, fit, and limitations.
- AO 41 addresses professional responsibility when technology is used in an appraisal assignment.
- The appraiser explains why verified sales evidence is more relevant instead of adopting the automated result.
Answer: Use the AVM as a tool, document the analysis, and retain responsibility for the appraisal conclusion.
What are the common exam traps?
- Trap
- Calling an AVM an appraisal
- Correction
- An AVM is an automated estimate. Illinois procurement exemption does not turn it into a credentialed appraisal.
- Trap
- Trusting decimal precision
- Correction
- A precise output can still have broad uncertainty, stale inputs, poor fit, or a wrong property match.
- Trap
- Ignoring the effective date
- Correction
- The estimate applies to a model run and data time; markets and subject facts can change afterward.
- Trap
- Comparing vendor scores directly
- Correction
- Read each score's definition, scale, calibration, error measure, validation population, and threshold.
- Trap
- Using national accuracy for every property
- Correction
- Validate the subject's geography, type, price tier, condition, and data environment.
- Trap
- Treating missing data as neutral
- Correction
- Missing renovations, damage, rights, concessions, units, or parcel identity can bias the estimate materially.
- Trap
- Assuming more data is always better
- Correction
- Duplicate, manipulated, nonmarket, stale, or mislabeled data can degrade a model unless governed and cleaned.
- Trap
- Reviewing only complaints
- Correction
- The current covered federal rule calls for random-sample testing and reviews in addition to exception controls.
- Trap
- Testing average accuracy only
- Correction
- Examine overvaluation, undervaluation, tails, coverage, geography, property segments, and protected-group outcomes.
- Trap
- Selecting the model that closes the loan
- Correction
- Model choice and override must follow policy, validation, independence, and conflict controls rather than desired value.
- Trap
- Treating value acceptance as a consumer estimate
- Correction
- It is a Fannie Mae DU offer with eligibility, representations, prior-data, timing, and appraisal-fallback requirements.
- Trap
- Assuming appraisal waiver means no property risk
- Correction
- The lender still manages eligibility, data, collateral, condition, fraud, legal, and repurchase risks under the program.
- Trap
- Ignoring copy rights
- Correction
- Regulation B includes AVM reports as valuations for covered first-lien dwelling application copy rules.
- Trap
- Treating fair lending as a final checkbox
- Correction
- Nondiscrimination controls belong in data selection, development, validation, deployment, monitoring, overrides, and remediation.
- Trap
- Letting technology replace professional responsibility
- Correction
- A broker, appraiser, lender, reviewer, or model owner remains responsible for lawful use within that person's role.
Can you answer these original practice questions?
These questions are original study items aligned to the published outline. They are not copied, recalled, or predicted PSI questions.
1. What is an AVM?
- A computerized model that estimates property value
- A guaranteed appraisal
- A title-insurance policy
- A home inspection
Show answer and explanation
Answer: A computerized model that estimates property value
The estimate comes from data and mathematical relationships rather than a complete appraisal assignment by itself.
2. What does Illinois section 5-5 say about AVMs?
- The Appraiser Licensing Act does not apply to procurement of an AVM
- Every AVM is a certified appraisal
- Only brokers may buy an AVM
- AVMs are prohibited
Show answer and explanation
Answer: The Appraiser Licensing Act does not apply to procurement of an AVM
Procurement is exempt, but the output cannot be misrepresented as licensed appraisal work.
3. Which property is most likely to challenge a typical residential AVM?
- A unique rural estate with few comparable sales
- A typical tract home with many recent similar sales
- A condo with correct unit data and many same-project sales
- A standardized home in a dense market
Show answer and explanation
Answer: A unique rural estate with few comparable sales
Sparse data and unusual features can place the subject outside strong model coverage.
4. Which is one of the five federal AVM quality-control factors?
- Random-sample testing and reviews
- Guaranteed loan approval
- Use of the highest estimate
- No human oversight
Show answer and explanation
Answer: Random-sample testing and reviews
The other factors address high confidence, manipulation, conflicts, and nondiscrimination.
5. What does a confidence score prove?
- Only what the vendor defines and validates it to indicate
- That the estimate is a certified appraisal
- That the property has no defects
- That every lender will accept the result
Show answer and explanation
Answer: Only what the vendor defines and validates it to indicate
Confidence scales are not standardized and must be read with coverage and error documentation.
6. An AVM uses the wrong parcel. What should the user do?
- Reject the result and correct identity before reuse
- Trust the confidence score
- Average it with list price
- Call it an appraisal
Show answer and explanation
Answer: Reject the result and correct identity before reuse
A subject-matching failure makes the estimate irrelevant regardless of apparent precision.
7. How does Regulation B treat an AVM report for a covered application?
- As a written valuation subject to copy rules
- As no document at all
- As a deed
- As a credit score only
Show answer and explanation
Answer: As a written valuation subject to copy rules
Section 1002.14 expressly lists an AVM report as a valuation example.
8. What is Fannie Mae value acceptance?
- A DU offer allowing an eligible loan casefile to proceed without an appraisal
- Every public website estimate
- An Illinois broker license
- A property-tax appeal
Show answer and explanation
Answer: A DU offer allowing an eligible loan casefile to proceed without an appraisal
It has casefile eligibility, timing, property, transaction, law, and lender conditions.
9. Why should AVM tests examine subgroups and geography?
- Portfolio averages can hide unequal coverage or error patterns
- Protected classes determine property value
- Every neighborhood must have one price
- Accuracy never changes by market
Show answer and explanation
Answer: Portfolio averages can hide unequal coverage or error patterns
Fair lending and model risk require more than one overall accuracy statistic.
10. What happens when an appraiser uses an AVM in an appraisal?
- The appraiser remains responsible for understanding and using it appropriately
- The vendor becomes the signing appraiser
- USPAP no longer applies
- Reconciliation is unnecessary
Show answer and explanation
Answer: The appraiser remains responsible for understanding and using it appropriately
Technology assists the analysis but does not transfer professional responsibility.
How should you study this area?
- Session
- Session 1
- Focus
- Map product and use
- Proof you are ready
- Classify 30 outputs as AVM, appraisal, evaluation, CMA, BPO, value acceptance, property data, tax assessment, or hybrid appraisal, then name the permitted use.
- Session
- Session 2
- Focus
- Audit data lineage
- Proof you are ready
- Trace subject identity, sales, area, units, condition, permits, listing, geography, price index, effective date, missing values, and overrides for 15 model files.
- Session
- Session 3
- Focus
- Read performance
- Proof you are ready
- Interpret 20 cases using coverage, hit rate, confidence, median absolute error, directional bias, tail error, local validation, score definition, and drift.
- Session
- Session 4
- Focus
- Apply the five federal controls
- Proof you are ready
- Design high-confidence, manipulation, conflict, random-testing, and nondiscrimination controls for 12 mortgage model-use scenarios.
- Session
- Session 5
- Focus
- Build human-review triggers
- Proof you are ready
- Route 24 properties based on confidence, property match, data conflict, market density, uniqueness, condition, renovation, value range, dispute, and regulatory use.
- Session
- Session 6
- Focus
- Run M-O-D-E-L
- Proof you are ready
- Score at least 90 percent and explain every miss through mission, origin, design, equity and errors, or launch limits.
Do not count recognition as mastery. Close the notes and explain the rule, apply it to a new fact pattern, and identify why each distractor fails.
Practice the topic in Pass Illinois
From concept to decision
Drill this topic, then review the explanation
Pass Illinois gives you original national and Illinois questions, topic-by-topic study, clear explanations, timed practice, flashcards, progress tracking, and Math Coach. Start free, find the weak distinction, and focus the next session there.
Questions students ask about What Is an AVM? Illinois Real Estate Exam Guide
What is an automated valuation model in real estate?
An automated valuation model, or AVM, is a computerized model that estimates real property value from data and mathematical methods. It can use recorded sales, property characteristics, location, tax and parcel records, market trends, and statistical or machine-learning relationships. The output may be a point estimate, range, confidence measure, or risk flag.
Is an AVM an appraisal in Illinois?
No. Illinois section 5-5 says the Real Estate Appraiser Licensing Act does not apply to procurement of an AVM. That exception permits obtaining the model product; it does not make the output a licensed appraisal or authorize an unlicensed person to represent model output as appraisal work.
How does an AVM calculate a home value?
The exact method is model-specific. Common approaches relate the subject's verified features and location to sales and market patterns, then estimate the most likely value and uncertainty. Some models use repeat-sales indexes, hedonic regression, comparable-selection algorithms, geospatial features, ensemble methods, or combinations. A precise-looking number can still rest on incomplete or stale inputs.
What is an AVM confidence score?
It is a model-specific indicator of expected reliability or uncertainty. A higher score can mean the subject fits the model's data and coverage better, but scores are not standardized across vendors. Users must read the vendor's definition, validation results, population, effective date, and error measure rather than comparing raw scores as though they share one scale.
When does an AVM work best?
AVMs often perform better where there are many recent, verifiable sales of reasonably similar properties and accurate standardized subject data. Performance can weaken for unique homes, rural or thin markets, rapidly changing areas, mixed-use property, unusual rights, major unrecorded renovations, condition extremes, new construction, and places with sparse or inconsistent records.
What federal AVM quality standards are current in 2026?
The interagency final rule effective October 1, 2025 requires covered mortgage originators and secondary market issuers to maintain policies, practices, procedures, and control systems designed for high confidence in AVM estimates, protection from data manipulation, avoidance of conflicts, random-sample testing and reviews, and compliance with applicable nondiscrimination laws.
Does the federal AVM quality rule apply to every online estimate?
No. The final rule addresses covered use by mortgage originators and secondary market issuers to determine collateral worth in certain mortgage credit decisions or securitization determinations involving a consumer's principal dwelling. A public website estimate, tax model, broker tool, portfolio screen, or commercial-property model may fall outside that specific rule while remaining subject to other laws and controls.
Can an AVM discriminate?
An AVM can reproduce or amplify unequal historical data, coverage gaps, proxy effects, geospatial patterns, or model-design choices. The current federal quality rule expressly includes compliance with applicable nondiscrimination laws. Responsible use requires representative validation, subgroup and geographic testing, data governance, monitoring, exception review, and a path for material data errors.
Does a borrower get a copy of an AVM report?
Current Regulation B lists an AVM report estimating property value as a valuation example. For a covered application secured by a first lien on a dwelling, the creditor generally must provide copies of appraisals and other written valuations promptly upon completion or three business days before consummation or account opening, whichever is earlier, subject to the timing-waiver rules.
Is Fannie Mae value acceptance the same as an AVM?
No. Fannie Mae describes value acceptance as a Desktop Underwriter offer under which an appraisal is not required for an eligible loan casefile. It uses Fannie Mae's data and modeling framework and can depend on prior appraisal information, casefile facts, eligibility, and lender representations. It is a program decision, not a generic consumer AVM estimate.
Are these official PSI questions or a property valuation?
No. The questions are original, and the primary sources were checked through August 1, 2026. A live decision requires the actual model documentation, input data, effective date, intended use, coverage, validation population, error metrics, confidence definition, fair lending tests, federal and state rules, program requirements, and qualified human review.
Primary sources
- PSI Illinois Candidate Information Booklet effective June 24, 2026
- 225 ILCS 458/5-5(h), current Illinois exemption for procurement of an automated valuation model
- CFPB interagency final rule overview, AVM quality-control coverage and five required control factors
- CFPB current AVM compliance resource for Regulation Z section 1026.42(i) and official interpretations
- 89 FR 64538, interagency AVM quality-control final rule effective October 1, 2025
- Regulation B, 12 CFR 1002.14, current AVM report treatment as a valuation and copy timing
- Regulation Z, 12 CFR 1026.42, current valuation independence definitions and covered transaction context
- 12 CFR 34.43, current OCC appraisal exceptions, institution evaluation duties, and credential thresholds
- The Appraisal Foundation, current 2024 USPAP and Advisory Opinion 41 on technology adopted April 23, 2026
- Fannie Mae Selling Guide B4-1.4-10 updated June 3, 2026, value acceptance eligibility and use
- Fannie Mae Selling Guide B4-1.4-11, current value acceptance plus property data and collector controls
- U.S. Department of Housing and Urban Development, current Fair Housing Act protected-class overview
The current official outline controls the tested scope. Statutes, regulations, and official agency materials control when a general study rule and a jurisdiction-specific rule differ.
Editorial status
Checked against primary sources
The Pass Illinois editorial team last checked this guide on August 1, 2026. Every practice question is an original study item, and the source links above let you verify the rules that support the lesson.