AI in Our Product

Artificial intelligence is core to how RaiseAppraisal works. We use AI to:

• Read your appraisal PDF and extract the structured data inside (subject property details, comparables, adjustments, dates, values)
• Identify potential issues worth raising in a Reconsideration of Value request
• Suggest alternative comparable sales that might support a different value
• Draft argument language for inclusion in your ROV documents
• Pre-fill lender ROV forms with your case information

When you upload an appraisal and click “Generate,” AI is doing most of the heavy lifting. We tell you this clearly, and we want you to understand both what AI is good at and where it can fall short.

We use leading commercial AI systems and secure processing tools to extract appraisal data, analyze documents, draft ROV arguments, and assemble submission-ready packages. We do not design our product around a single AI company or model; we evaluate the best available technology as the AI industry evolves.

• Provider-agnostic architecture. Our system is built to use capable, reputable AI services without being locked into one vendor.
• Privacy-first processing. Appraisal content is processed to generate your output, not to train third-party AI models.
• Continuous system improvement. We improve our own workflows, checks, strategies, and quality controls based on what we learn from outcomes and feedback.

We do not use:

• Free AI APIs that train on customer data
• AI providers without clear no-training commitments for API customer data
• Personal information to retrain public or corporate LLMs

• Reading documents fast. AI extracts structured data from a PDF appraisal in seconds. A human would take many minutes per page.
• Writing professional, neutral argument language. AI is excellent at composing the kind of formal, respectful, question-based language an ROV submission requires.
• Pattern recognition. AI can spot data inconsistencies, large adjustments, and other patterns that warrant attention.
• Comparable suggestion. AI can sort through large numbers of nearby properties and surface the ones that might be relevant.

We are upfront about this:

• Factual errors are possible. AI may misread a number, miscategorize a comparable, or state something inaccurately about a property. We've built validation systems to catch many of these (see “How We Prevent Errors” below), but we cannot guarantee 100% accuracy.
• Subjective judgment. AI does not have the experience or licensure of a professional appraiser. It cannot replace human professional judgment about value.
• Out-of-distribution cases. Unusual properties, unusual markets, or unusual appraisal types may produce unreliable AI output.
• Hallucination risk. AI can occasionally generate plausible-sounding statements that aren't supported by the underlying data.

1. Source-Grounded Data

Every comparable sale we surface is tied to a public-records or licensed-data source. We do not invent comparables. If AI suggests a comp, it must trace back to actual data we received from a real estate data provider.

2. Fact Validation Layer

Before any specific factual claim (comp address, sale price, sale date, square footage) appears in a generated document, our system cross-validates the claim against the source data. Claims that don't match the source data are rejected, not displayed.

3. Human Review of Output

We've built our software so that you, the user, see the generated documents before submitting anything. You are always in the loop. Our software does not submit anything to anyone on your behalf — every document goes through your review.

4. Compliance Layer

AI prompts include explicit instructions to avoid certain content categories (target values, accusatory language, references to protected class characteristics). Generated documents are scanned for prohibited content before delivery.

5. Continuous Improvement

We monitor outputs and update our AI prompts and validation systems regularly to address discovered issues.

Even with our safeguards, the user is responsible for reviewing every generated document for accuracy before submitting it to any third party.

Please:

• Verify that every factual claim about your property is correct
• Verify that every comparable sale and its details are correct
• Confirm dates, square footage, sale prices, and addresses match your records
• Read every page before signing or submitting
• If anything looks wrong, contact us at support@raiseappraisal.com

We provide tools and analysis. You are still the person submitting an ROV request.

We apply the same philosophy across the product: sensitive information and neighborhood context are not used to discriminate, score a borrower, or justify a value outcome. When context is relevant, we use it as a guardrail to keep comparable selection objective and inside the market area reflected by the appraisal.

• Protected characteristics are not value drivers. Race, color, religion, national origin, sex, age, disability, marital status, and similar characteristics cannot be used to support or reject a valuation argument.
• Context is a guardrail, not a conclusion. Crime, demographic, or other neighborhood context, if reviewed, is limited to checking whether proposed comparables remain in a reasonably comparable area and do not jump between materially different neighborhoods.
• The appraisal stays the frame. Our goal is to stay within the confines of the appraisal and evaluate the data, assumptions, market area, and comparable set on objective terms.

See our Fair Lending Statement for the full list of factors our AI cannot use to justify a value or recommendation.

When you upload an appraisal:

1. We send the document content to the AI provider for processing
2. The AI extracts the data we need
3. The AI returns the result to us
4. We display the output to you in your account

The AI provider holds your data only briefly (for abuse prevention) and does not train on it.

After 90 days, we delete your uploaded PDF. We may retain a de-identified version of the extracted data (with names, addresses, and other identifiers removed) to help us improve our software. See our Privacy Policy for details.

Every document our software generates carries a clear AI disclosure:

This disclosure appears so that anyone receiving an ROV submission generated through our software (lender, AMC, appraiser) knows exactly what they're looking at.

• We will continue to invest in accuracy. Hallucinations and factual errors are a real risk in AI products. We take them seriously and continuously work to reduce them.
• We will be transparent. When we make material changes to our AI providers, prompts, or validation systems, we'll update this page.
• We will not over-promise. We don't claim our AI is perfect. We don't claim it can replace professional judgment. We don't claim it can guarantee outcomes.
• We will not use your data to train third-party AI. Anthropic and Google have committed contractually that they don't train on API customer data, and we don't share your data with anyone else for AI training purposes.

• We don't use AI to determine your appraisal's value (we don't determine values at all)
• We don't submit AI-generated content on your behalf to any third party
• We don't sell AI-generated insights about you to anyone

If you notice an AI-generated error in a document we produced:

Email

support@raiseappraisal.com

Subject

“ROV Output Concern - {Summary of issue}”

We investigate every report. Where the issue indicates a systemic problem, we update our prompts or validation system to prevent recurrence.

We are building this product in an era when AI capability is improving rapidly. We commit to:

• Adopting better AI providers and models when they offer meaningful accuracy or capability improvements
• Maintaining our no-training commitment regardless of AI provider changes
• Continuing to operate within all applicable AI-related laws (CFPB algorithmic fairness guidance, FTC AI guidance, future state regulations)
• Communicating any material changes through this page

Questions about how we use AI? Email support@raiseappraisal.com or compliance@raiseappraisal.com.

See also: Privacy Policy · Fair Lending Statement · USPAP & Appraiser Independence Statement