How to Get the Best Results from AIpraisal: A Complete Guide
Tips and tricks for appraisers to get the most accurate comparable sales, better comp results, and faster workflows using AIpraisal.
How to Get the Best Results from AIpraisal: A Complete Guide
The difference between a defensible appraisal and a vague one often comes down to comp quality. Here's how to make AIpraisal work harder for you.
AIpraisal searches dozens of sources simultaneously — sold transaction databases, auction records, specialist marketplaces, and retail listings — and ranks the results by relevance. But like any research tool, the quality of what you put in determines the quality of what you get out.
This guide walks through every stage of the workflow: from taking the right photo to selecting your final comps.
The AIpraisal Workflow at a Glance#
Before diving into the details, here's the full workflow from photo to finished comp set:
Table of Contents#
- Start With a Good Photo
- Let AI Analyze First, Then Review
- Understand Your Comp Results
- Use Focus Search for Complex Items
- Know When to Re-Analyze
- Use Optimize List Order
- Select the Right Comps
- Tips by Item Category
1. Start With a Good Photo#
The photo is the foundation of everything. AIpraisal uses your image for two things: AI analysis (to identify the item) and visual similarity search (which finds visually matching listings across the web). A poor photo degrades both.
What makes a good photo:
- One item per photo. If you're photographing a room full of donated furniture, photograph each piece individually. Multiple items in one frame causes the AI to guess which one to analyze.
- Show the whole item. Don't crop off legs, edges, or the top. The full silhouette helps the AI categorize it correctly.
- Capture brand and model markings. If there's a manufacturer's label, serial plate, or model number visible anywhere on the item — photograph it clearly. When AIpraisal can read a specific brand and model from the image, it can locate an exact product match instead of approximating from description alone.
- Neutral background when possible. A cluttered background with other items adds noise to the visual search. A plain floor, wall, or drop cloth produces cleaner results.
- Good lighting, no flash glare. Natural light or diffused lighting shows true color and material. Flash creates hotspots that obscure surface texture and finish — both of which matter for condition assessment.
What to avoid:
- Blurry or out-of-focus shots
- Photos taken from too far away
- Images where the item is partially hidden or boxed
- Screenshots of other listings (use the original item photo)
Pro tip: For deconstruction donations — windows, doors, cabinets, flooring — lean the item against a wall and photograph it squarely from the front. This gives the AI the clearest possible view of the item type and condition.
2. Let AI Analyze First, Then Review#
After uploading, click Analyze Item and let the AI do its initial pass before you touch anything. The AI extracts:
- Item name and category
- Brand and model number (when visible)
- Materials, style, and era
- Condition estimate
- Optimized search queries tuned for the item type
Then review what it found. The editable fields on the item page are there for a reason. If the AI got the brand wrong, fix it before searching — brand and model number are the highest-confidence signals for finding exact product matches. A corrected brand and model will dramatically outperform a generic description search.
Common things to check:
- Is the category correct? (furniture vs. decor, for example — they're searched differently)
- Did it pick up the brand name from the label?
- Is the condition assessment accurate?
You don't have to fix everything — just the fields that are clearly wrong or missing a key detail.
3. Understand Your Comp Results#
Not all comps are equal. AIpraisal pulls from multiple source types, and each tells a different story:
| Badge | What it means | Best used for |
|---|---|---|
| Sold | Completed transaction — item actually changed hands at this price | FMV evidence — strongest for IRS purposes |
| Retail | Current asking price from a retailer | Replacement cost — useful for context, not FMV |
| Auction | Realized price from an auction sale | High-value items where auction records are authoritative |
| Visual Match | Found by visual similarity search | When text search misses — confirms item type |
For charitable donation appraisals establishing fair market value, prioritize Sold comps. The IRS definition of FMV is what a willing buyer would pay a willing seller — that's a completed transaction, not an asking price.
Retail comps are still useful context: they establish replacement cost and help explain why the FMV is lower than new. But don't use a current retail listing as your primary comp for a used item.
Reading the similarity score: Each comp shows a match score. Higher scores mean the AI found strong similarity between the comp and your item. When you see the identified product displayed at the top of the results, the comps below were searched using that precise product name — which makes the match scores more meaningful and the comps more defensible.
4. Use Focus Search for Complex Items#
Focus Search (the crop tool on the item image) is your manual override when the standard search returns irrelevant results.
When to use it:
- The item was photographed in a room with other items and the visual search is matching the wrong thing
- You want comps specifically for one component of a multi-piece set
- The background is cluttered and pulling the visual search off target
- You want to search specifically for a brand label or model plate visible on the item
How it works: Draw a box around just the part of the image you care about. AIpraisal re-runs the search using only that crop — isolating it from the surrounding noise.
This is especially useful for deconstruction appraisals where items are photographed in-situ (still installed in a building). Crop out the door, the cabinet, or the light fixture and let the search focus on just that piece.
5. Know When to Re-Analyze#
Re-analyzing costs one AI call and takes about 10 seconds. It's worth doing when:
- You've corrected a major field (wrong category, missing brand)
- The initial analysis produced overly generic queries with no specifics
- You've photographed a clearer image of the item
- The search results look completely wrong for the item type
Re-analyzing regenerates all search queries from scratch based on your corrections and the updated image. If you fixed the brand name and model number, the new queries will be far more targeted.
Don't re-analyze when:
- The analysis looks basically right and you just want to tweak one small thing
- You've already run a good comp search — re-analyzing will reset your results
6. Use Optimize List Order#
The Optimize List Order toggle on the project page sorts your items by comp readiness — items with the strongest, most complete comp sets rise to the top.
Use this when you're working through a large project (20+ items) and want to tackle the easy wins first. Items with weak or no comps will sink to the bottom, letting you focus your manual review time where it's actually needed.
For items at the bottom of an optimized list, that's your signal to:
- Re-examine the photo
- Correct the category or brand
- Run a Focus Search
- Manually enter a comp you found through your own research
7. Select the Right Comps#
AIpraisal finds comps — you select which ones make it into your appraisal. Here's how to choose well:
Aim for 3–5 strong comps rather than 10 mediocre ones. A defensible appraisal doesn't need volume — it needs specificity. Three verified sold listings for the exact same item are more persuasive than ten vague "similar style" retail listings.
Prioritize:
- Same item, same condition (sold) — ideal
- Same item, different condition (sold, with condition adjustment noted)
- Same brand/model, similar condition (sold)
- Comparable item, same category and approximate age/condition (sold)
- Retail pricing as replacement cost reference only
Avoid:
- Comps where the price seems anomalous (one listing at 10x the others — check if it's a different size, rare variant, or listing error)
- Retail listings as FMV evidence for used items
- Comps that are obviously a different item despite a high visual score
Geographic considerations: For items that sell locally (building materials, large furniture), local market pricing may differ significantly from national online marketplace comps. Note this in your appraisal where relevant.
8. Tips by Item Category#
Furniture#
Brand matters enormously. A "mid-century modern walnut credenza" and a "Herman Miller Nelson credenza" are the same description to a layperson but worlds apart in value. If you can identify the manufacturer — look for labels on the underside, back, or inside drawers — enter it before searching.
Building Materials (Deconstruction)#
Items like doors, windows, cabinets, and flooring are often commodity-priced. Search by material, size, and condition rather than style. For reclaimed and salvage items, the specialist salvage and reclaimed materials markets provide the most relevant comps — AIpraisal prioritizes these sources automatically for building materials categories.
Electronics#
Model number is everything. A general description returns thousands of irrelevant results. The exact model number returns precise matches. Photograph the model sticker (usually on the back panel) and make sure it's captured in the analysis.
Artwork#
Original artwork is the hardest category — no two pieces are identical, so "comparable sales" is inherently approximate. Focus on the artist's sales history if the artist is known. For unknown artists, use style, medium, and size to find comparable works. Realized auction prices are the most authoritative comps for fine art.
Clothing and Textiles#
Condition and brand dominate pricing. A designer jacket in excellent condition and a generic equivalent of the same weight are not comparable. Always capture the brand label, size, and any visible condition issues (stains, wear, repairs) in your description.
Jewelry and Watches#
For watches, the reference number is the exact identifier — it's engraved on the case back or casebook. Dedicated watch markets are the most liquid and relevant source for timepiece comps. For jewelry, metal type, gemstone specs (carat, cut, clarity, color for diamonds), and brand all matter.
Getting Consistent Results#
The appraisers who get the best results from AIpraisal share a few habits:
- Photograph with intention. They think about what the AI needs to see, not just what looks good.
- Fix the category and brand before searching. Two minutes of review before a search beats re-running it three times.
- Use sold comps for FMV, retail for context. They don't confuse asking price with market value.
- Re-analyze when the first pass misses. They treat analysis as a starting point, not a final answer.
- Don't settle for weak comps. If the results don't look right, they investigate — a Focus Search, a manual correction, or a manual entry solves most gaps.
The goal is a comp set that's specific, defensible, and honest. AIpraisal handles the search — you bring the judgment. For a broader product overview, see our personal property appraisal software page.
Have questions or feedback on how AIpraisal is working for your practice? Reach out at support@aipraisal.io — we read every email.
Ready to streamline your appraisal workflow?
AIpraisal helps appraisers find comparable sales in seconds, not days. Try it free today.