Illinois exam glossary
Automated valuation model: look beyond the price estimate
An automated estimate can be useful without being the final answer. Before relying on the number, check that the model found the right property, used relevant data and was designed for the decision being made.
Last reviewed:
The short answer
An automated valuation model, or AVM, estimates property value using data and mathematical methods. The output may include a value, range or confidence measure. Its usefulness depends on the property match, inputs, market evidence, model fit and intended use. A model result alone does not prove that a licensed appraisal, property inspection or lending requirement has been completed. Illinois exempts procurement of an AVM from its appraiser-license Act; that is not a general license to perform or advertise appraisal work.
Reviewed for October 2026. PSI identifies AVM at national III.C.2 within Valuation, which has 8 of the 100 scored national broker questions. The federal quality-control rule took effect October 1, 2025 and applies to defined actors and uses, with stated exclusions. Its scope differs from borrower-copy rules and Illinois licensing law. The practice numbers are invented examples, not estimates of actual properties.
Five questions to ask before using an AVM result
- Is this the right property? Match the parcel, unit and interest. A high-looking confidence score cannot repair a result generated for the wrong condominium or a record with the wrong building size.
- What facts and date does the model use? Recent sales cannot fix a stale subject record by themselves. A renovation, damage or unusual characteristic may not appear in the available data. The date of the model run and the dates of its inputs are different things.
- What does its uncertainty measure mean? Read the vendor's definition and tested performance. Do not assume scores from two models share a scale or that a score is the probability of a particular sale price. Population statistics are not guarantees for one home.
- Is the model suitable for this use and population? Check the relevant property type, market and validation sample. Coverage, accuracy, errors and model limits are separate. No estimate is sometimes more appropriate than forcing a result outside supported data.
- Which requirements apply to the decision? Identify the user, transaction and purpose. The federal AVM rule, appraisal requirement, Illinois licensing provisions, borrower-copy rule and a loan program's appraisal option answer different questions. Permission to obtain a model is not permission to ignore the others.
| Objective | What it addresses | Illustrative control, not one mandated algorithm |
|---|---|---|
| High confidence in estimates | Whether outputs are sufficiently reliable for the use | Validate relevant data and model performance |
| Protection against data manipulation | Improper changes that distort the result | Control inputs and review overrides |
| Avoidance of conflicts of interest | Incentives affecting model selection or use | Separate desired loan outcomes from valuation judgments |
| Random-sample testing and reviews | Problems outside the set of obvious exceptions or complaints | Review a random production sample |
| Compliance with applicable nondiscrimination laws | Lawful use across affected people and communities | Investigate relevant disparities and correct unlawful practices |
The distinctions that matter
| Distinction | How it works | What to notice |
|---|---|---|
| Model output and a professional assignment | An AVM estimates through a model; an appraisal assignment carries its own development, reporting and professional requirements. | Do not treat a website number as proof those requirements were met. Some technical standards use appraisal terminology for modeled valuation; context matters. |
| Procurement and regulated appraisal work | Illinois Section 5-5(h) exempts obtaining the AVM. Other appraisal acts and titles have separate requirements and exemptions. | Buying an estimate does not confer an appraiser credential. |
| AVM, CMA and BPO | Automation describes the model. Illinois broker analyses have their own licensing, permitted-purpose and reporting framework. | Renaming an output does not decide who may provide a service or what the lender may use. |
| Confidence, coverage and error | A score describes what its maker defines. Coverage counts returned estimates; error compares them with a defined benchmark. | A high return rate does not establish a low error rate, and neither guarantees an individual result. |
| Quality controls and copy rights | The AVM rule addresses covered principal-dwelling collateral uses. Regulation B separately addresses written valuations for covered first-lien dwelling applications. | Different definitions and coverage can apply to the same report. |
| Model estimate and value acceptance | Fannie Mae value acceptance is a casefile-specific DU offer under program requirements. | A public online estimate is not itself an appraisal-waiver offer or proof that all eligibility conditions are satisfied. |
Work through the facts
The wrong unit produces a confident number
The record supplied to the model is for the penthouse rather than Unit 12B. Its confidence display looks strong.
- The model is answering for a different property.
- A score is not independent proof that the property match is correct.
Answer: Do not rely on that result for Unit 12B. Correct the identity and evaluate any new result on its own merits.
The renovation has not reached the records
A home was substantially renovated, but the model's subject data still describes the earlier condition. Current photographs and verified renovation information conflict with that record.
- The run can be recent while an important input is old.
- The conflict needs investigation rather than an automatic adjustment by the renovation cost.
Answer: Review the relevant facts and suitable valuation process. Do not assume the model captured the work or that cost equals added market value.
Two confidence displays, two definitions
Vendor A shows 88 on one scale. Vendor B shows 0.92 on another. Their documentation does not establish a common metric or calibration.
- The figures are not comparable merely by converting one into a percentage.
- Read the score definitions and validated performance for the actual use.
Answer: Neither number alone proves which estimate is more reliable or the probability of a sale at that price.
A low median can hide a large miss
Five hypothetical absolute percentage errors are 1%, 2%, 3%, 4% and 30%. Treat these only as a small arithmetic sample.
- The ordered middle error is 3%.
- The largest observed error is still 30%.
Answer: A 3% median is not a maximum-error guarantee. The sample also does not establish a vendor's real-world performance.
Coverage is not the accuracy rate
A test requests estimates for 100 properties. The model returns 90. Of those 90, 63 are within the test's defined error band.
- Coverage is 90 of 100, or 90%.
- The share within the band among returned estimates is 63 of 90, or 70%.
Answer: Report the denominator. Do not label 90% coverage as 90% accuracy or claim every property received a reliable estimate.
A covered business-purpose mortgage
Assume a covered mortgage originator uses an AVM in a covered credit decision on a natural person's principal dwelling. The loan is primarily for business and no stated exclusion applies.
- The current AVM provision expressly reaches this principal-dwelling situation.
- Primarily business purpose alone is not the exemption assumed here.
Answer: Apply the covered quality-control requirements. Do not confuse a business-purpose mortgage on the principal dwelling with every commercial-property valuation.
Monitoring is a different purpose
An institution uses an AVM solely to monitor an existing mortgage portfolio's quality, with no credit decision or covered securitization determination.
- The rule expressly identifies monitoring as an excluded use.
- Being outside this particular rule does not eliminate every other legal or risk-management obligation.
Answer: Do not apply the five-factor AVM rule solely because a model appears in the workflow.
A written report and borrower copies
For a covered first-lien dwelling application, the creditor obtains a written AVM report estimating the property's value. No exception defeats the described copy requirement.
- Regulation B includes such a report among valuation examples.
- The question is about providing copies, not whether an appraisal license or inspection exists.
Answer: Apply the copy rule. A permitted timing waiver does not by itself eliminate the right to the copy.
DU offers value acceptance
The final loan casefile receives a Fannie Mae value-acceptance offer. The lender must still determine whether all applicable conditions for exercising it are met.
- The offer is specific to the program and casefile.
- It is not created simply by finding an online home estimate.
Answer: Follow the applicable loan requirements, including any relevant UAD 3.6 provisions. Do not turn a model result into a universal appraisal waiver.
An appraiser uses a model as one tool
An appraiser considers an AVM while developing an appraisal and finds it relies on properties unlike the subject.
- The professional assignment remains a separate responsibility.
- The federal AVM quality rule excludes the specified appraiser-development use; that is not an exemption from all appraisal standards.
Answer: Evaluate the tool appropriately within the assignment. Do not replace supported analysis simply because the automated number is higher.
Mistakes to avoid
- The estimate has many digits, so it must be precise. Display precision does not establish the accuracy or uncertainty of the underlying estimate.
- A confidence score is a standardized probability. Read its definition, scale, validation and intended interpretation.
- The latest run uses only current property facts. Check each material input and the time period represented.
- A median error is the worst possible error. The median is a middle observation, not a ceiling or guarantee for one property.
- A model that answers every request is accurate. Coverage and error measure different things.
- Every AVM use must follow this one federal rule. Check actors, purpose, collateral and stated exclusions; parallel agency rules can matter.
- Business purpose alone excludes the principal-dwelling mortgage. The current AVM provision expressly addresses those mortgages within its scope.
- An appraisal-development exclusion means no professional standards apply. The AVM-specific exclusion and professional appraisal duties are different questions.
- Permission to buy an AVM permits any appraisal service. Illinois procurement and other regulated acts or titles remain separate.
- A timing waiver eliminates borrower-copy rights. Read the distinct copy requirement and permitted timing waiver.
Check your understanding
These are original teaching questions, not recalled or predicted PSI questions. Answer first, then explain why the other choices fail.
1. What does AVM identify in this glossary?
- A mandatory property inspection
- A guaranteed sale price
- An automatic certification that all appraisal requirements were completed
- A model using data and analytic methods to estimate property value
Show answer and explanation
Answer: D. A model using data and analytic methods to estimate property value
The term describes modeled valuation. The output does not itself establish inspection, professional credentials, program eligibility or a guaranteed transaction result.
2. The model uses the wrong condominium unit but returns a strong confidence display. What should be addressed first?
- Average the number with the asking price
- Assume the larger unit proves a valid comparable adjustment
- Correct the property identity and assess the resulting valuation process
- Increase the confidence score
Show answer and explanation
Answer: C. Correct the property identity and assess the resulting valuation process
The result does not answer the intended subject question. Confidence cannot make a wrong property match relevant.
3. A vendor score is 88 and another is 0.92, with no common scale or definition established. What is supported?
- The first guarantees an 88% sale probability
- Both are necessarily the same metric
- The raw numbers do not establish a reliability ranking
- The second is necessarily more accurate
Show answer and explanation
Answer: C. The raw numbers do not establish a reliability ranking
Different methods and scales need their own documentation and validation. A conversion to percentage format does not establish common meaning.
4. Absolute percentage errors are 1%, 2%, 3%, 4% and 30%. Which statement is correct?
- The maximum is 4%
- The median is 3% even though one error is 30%
- Every error is at most 3%
- The sample proves every local property can be valued accurately
Show answer and explanation
Answer: B. The median is 3% even though one error is 30%
The middle of the ordered five is 3%. A median does not set a maximum, describe every individual property or validate a vendor's performance.
5. Ninety estimates are returned from 100 requests; 63 of those 90 meet the defined error band. Which denominator statement is correct?
- Coverage and within-band performance are both 90%
- Coverage is 90%; within-band performance among returned results is 70%
- Sixty-three of 90 means a 63% rate among returned estimates
- Every request produced a within-band estimate
Show answer and explanation
Answer: B. Coverage is 90%; within-band performance among returned results is 70%
Coverage uses all requests; the stated conditional within-band measure uses the returned estimates. Define both population and benchmark rather than relabeling one as the other.
6. A covered originator uses an AVM in a covered credit decision on a natural person's principal dwelling. It is primarily a business-purpose mortgage and no exclusion applies. What follows?
- A principal dwelling can never secure a business loan
- Business purpose automatically removes it from the AVM provision
- The stated AVM scope can include it
- All commercial-property models follow identical rules
Show answer and explanation
Answer: C. The stated AVM scope can include it
Section 1026.42(a) expressly addresses the AVM provision's principal-dwelling mortgage scope even when primarily for business, commercial, agricultural or organizational purposes. The other coverage facts still matter.
7. Use is solely monitoring mortgage-portfolio quality, with no relevant credit decision or securitization determination. What does the specific AVM rule state?
- This monitoring use is an identified exclusion
- No other law can ever apply
- Every monitoring model must be replaced by an inspection
- The monitoring result is an appraisal license
Show answer and explanation
Answer: A. This monitoring use is an identified exclusion
The stated exclusion concerns this AVM-specific rule. It does not remove other legal requirements or prudent controls.
8. Which is one of the five federal AVM quality-control objectives for covered uses?
- Random-sample testing and reviews
- A universal vendor confidence score of 95
- A ban on all third-party model services
- Guaranteed approval at the requested loan amount
Show answer and explanation
Answer: A. Random-sample testing and reviews
The rule requires controls designed around its five objectives. It does not supply this universal score, loan guarantee or third-party prohibition.
9. What does Illinois Section 5-5(h) exempt?
- Every model-based mortgage from all regulation
- Every unlicensed appraisal act after a model is purchased
- Procurement of an automated valuation model
- Every use of a State-certified title
Show answer and explanation
Answer: C. Procurement of an automated valuation model
The clause concerns procurement. It does not confer a credential or erase separately applicable service, title, lending and program requirements.
10. A creditor obtains a written AVM valuation for a covered first-lien dwelling application. Which issue does Regulation B address here?
- Automatic licensing of the report's purchaser
- Required valuation copies and their applicable timing
- Only whether the borrower's property is the principal dwelling under the AVM quality rule
- A promise that the modeled number equals the sale price
Show answer and explanation
Answer: B. Required valuation copies and their applicable timing
An AVM report can be a written valuation for the copy rule. That purpose and its dwelling/application scope differ from the AVM quality rule and Illinois procurement law.
11. A consumer finds a public online estimate. Does that itself establish a Fannie Mae value-acceptance offer?
- Yes, if the asking price is lower
- No, because automated tools can never inform any lending process
- Yes, any estimate is a DU casefile offer
- No, the program offer and applicable eligibility conditions are separate
Show answer and explanation
Answer: D. No, the program offer and applicable eligibility conditions are separate
Value acceptance is a specific DU option under applicable Fannie Mae requirements. A public model result is not itself that offer or proof of eligibility.
12. An appraiser uses an AVM while developing an appraisal. What does exclusion from this specific AVM quality-rule use imply?
- It does not erase the appraiser's separate professional responsibilities
- No professional duties remain
- Every model must be accepted without analysis
- The vendor automatically becomes the signing appraiser
Show answer and explanation
Answer: A. It does not erase the appraiser's separate professional responsibilities
The rule names the appraiser-development exclusion. That does not make every technical tool suitable or transfer responsibility for the assignment conclusion.
13. For arithmetic only, an AVM is $420,000 and the specified comparison benchmark is $400,000. Define signed percentage error as (estimate minus benchmark) divided by benchmark. What is the result?
- Plus 5%
- Minus 5%
- Plus 20%
- A guarantee of a 5% maximum error on every home
Show answer and explanation
Answer: A. Plus 5%
The difference is $20,000. Dividing by the expressly defined $400,000 benchmark gives 0.05, or plus 5%. The exercise neither proves the benchmark is actual market value nor establishes a vendor's accuracy.
Common questions
What does AVM stand for?
Automated valuation model. Illinois defines it through an automated system using available property records and analytic methods to derive value. Models can use sales, property characteristics and price changes. AVM does not necessarily mean generative AI or one particular algorithm.
Is an AVM the same as an appraisal?
Do not infer a completed licensed appraisal assignment from a stand-alone model result. Technical usage can differ: IAAO discusses modeled valuations using appraisal terminology. Illinois separately defines appraisal and exempts AVM procurement. Read the applicable law, service and assignment rather than assuming a model number proves credentials or that no automated work can ever be part of appraisal practice.
Can someone obtain an AVM without an Illinois appraiser license?
Section 5-5(h) says the Appraiser Licensing Act does not apply to procurement of an AVM. That limited clause does not authorize every appraisal service, misleading title or representation. The broker-product exemption and its permitted uses are separate provisions; calling a service automated does not answer every licensing question.
Can an AVM replace an appraisal required for a loan?
A model result alone does not remove an applicable appraisal requirement. Identify the transaction, institution, governing exception and any program conditions. A permitted evaluation or a program offer can involve different requirements. Do not assume either that every loan needs the same appraisal or that any automated estimate qualifies as a substitute.
What does an AVM confidence score mean?
Its documented meaning matters. Methods and parameters can differ between vendors, including similarly named measures. Read the scale, tested population and intended interpretation. A score is not automatically a percentage probability that the property will sell at the estimate, an inspection finding or a guarantee. The rule's high-confidence objective does not create one numerical vendor score that all models must use.
Why can a high-confidence estimate be wrong?
The model may have the wrong unit, stale condition data or inputs outside its supported population. A score does not independently verify all those facts. Investigate material conflicts rather than treating a display as conclusive or assuming every correction necessarily increases the property value.
Does a low national error statistic prove local accuracy?
No. Read the tested population, property types, period, benchmark and error definition. A national median can conceal large individual errors or differences across markets. Coverage also matters: a statistic based on returned estimates does not explain the properties for which the model returned no result.
Which federal quality rule applies in October 2026?
The interagency AVM quality-control final rule took effect October 1, 2025. It addresses defined mortgage-originator and secondary-market uses for principal-dwelling collateral credit decisions or covered securitization determinations. Its objectives concern confidence, manipulation, conflicts, random testing and nondiscrimination. It does not mandate one algorithm or a universal score cutoff.
Does the rule cover every website estimate?
No. Check the actor and use. The rule expressly excludes specified monitoring, review of completed collateral determinations and appraiser-development uses. The CFPB codification also distinguishes institutions regulated under parallel agency rules; that is not a blanket exemption for all banks. Other laws and requirements can still apply outside this particular provision.
Does it cover only consumer-purpose purchase loans?
No. The current AVM scope can include mortgages primarily for business, commercial, agricultural or organizational purposes when secured by the consumer's principal dwelling and the other conditions are met. Its credit-decision definition also addresses specified mortgage or credit-line changes. Actual actor definitions and exclusions still matter; do not treat every modification or commercial-property estimate identically.
Can a covered user outsource all responsibility to a vendor?
The rule addresses covered decisions made directly or through or with third parties or affiliates. Using a vendor does not eliminate the covered user's control obligations. It also does not make every vendor the same regulated actor or prescribe one universal validation contract or testing schedule.
Does random sampling mean only reviewing disputed estimates?
No. Reviewing complaints or obvious exceptions is different from random-sample testing and review. The rule names random testing as an objective. Useful performance review can also examine relevant segments and disparities, but one overall error statistic does not by itself establish compliance with every applicable nondiscrimination law.
Can an AVM be biased?
Automation does not prove that results are free from bias or unlawful discrimination. Data, model choices and deployment can produce patterns that need investigation. The covered federal rule includes compliance with applicable nondiscrimination laws. A disparity or one error statistic needs context; it does not by itself decide every legal claim or establish one mandatory testing method.
Does the borrower receive a copy of an AVM report?
For a covered first-lien dwelling application, Regulation B includes an AVM report estimating value among written valuations. Copies generally must be provided promptly on completion or three business days before consummation or account opening, whichever is earlier, subject to permitted timing waivers. A timing waiver is not a waiver of the copy itself. Coverage and exceptions must still be checked.
Is Fannie Mae value acceptance just another AVM name?
No. DU may offer a casefile-specific option under which an appraisal is not required when the applicable program conditions are met. The lender must follow the current requirements, including relevant UAD 3.6 policy where applicable. The offer, a model estimate and property-data collection are distinct concepts; none alone guarantees loan approval or a defect-free property.
Did technology guidance create a new 2026 USPAP edition?
The Appraisal Foundation identifies the current USPAP edition as 2024. It separately reports adoption of Advisory Opinion 41 on April 23, 2026, addressing technology in appraisal and review assignments. That guidance update should not be confused with a new edition or a claim that technology removes the appraiser's responsibilities.
Are these official PSI questions or real property estimates?
No. These are original teaching examples. PSI identifies AVM at III.C.2 within the whole 8-question national Valuation category. The small set has no validated readiness percentage. Explain what the number, tool and applicable rule establish, then practice unfamiliar mixed questions.
Keep practising
Your next practice step
Put valuation into practice
Choose the approach or value principle that fits the facts, then check your reasoning.
These web resources are free. The app has a separate free starter tier and optional paid full access.
Official sources and scope
Use each source for its own purpose. Illinois law addresses definitions, procurement and professional requirements; the federal AVM provision addresses defined collateral uses; Regulation B addresses valuation copies. IAAO's identified 2018 technical guidance explains model and uncertainty concepts, rather than supplying a new federal legal mandate. The Appraisal Foundation distinguishes its 2024 USPAP edition from 2026 technology guidance. Fannie Mae's program offer remains subject to the applicable current loan requirements.
- 225 ILCS 458/1-10: appraisal, reports and AVM definitions
- 225 ILCS 458/5-5: appraisal licensing and selected exceptions
- 225 ILCS 454/10-45: permitted uses and complete written-report contents
- 68 Ill. Adm. Code 1455.240: incorporated USPAP edition and qualified exception
- 12 CFR 1026.42(a),(i): current AVM scope, exclusions and standards
- CFPB final-rule fast facts: covered decisions and stated exclusions
- Federal Register 89 FR 64538: final AVM rule effective October 1, 2025
- The Appraisal Foundation: current 2024 USPAP and April 2026 technology guidance
- Regulation B 12 CFR 1002.14: valuation-copy coverage, timing and fees
- IAAO 2018 AVM standard: data, modeled appraisal terminology and uncertainty
- Fannie Mae B4-1.4-10: a DU offer and applicable program conditions
- PSI June 24, 2026: automated valuation at III.C.2
This is exam preparation, not advice for a specific transaction. State law and the applicable facts control. Our editorial and corrections process explains how to report an issue.