AI Repair Estimates: How They Work
AI can estimate repair costs from photos: the software identifies the rooms and items in each photo, judges their condition, picks a repair or replacement scope, and prices it with cost data for the property's area. It's useful for screening deals and getting an itemized first draft fast. It can't see behind walls or under the house, it's only as good as the photos, and its prices can miss local labor rates, so verify the big items with a contractor before you commit.
Disclosure: Deal Run is our product, and it makes photo-based repair estimates. We describe the other tools only from their own sites and say plainly where photos fall short, ours included.
For what repairs cost at each rehab level, see our guide to rehab cost per square foot. This page covers how the AI version of an estimate gets built and how far to trust it.
How a photo-based repair estimate gets built
Tools differ in the details, and most don't publish them, but a photo-based estimate generally goes through the same stages a person does:
- Sort the photos. Each photo is tagged as a room or area: kitchen, bathroom, bedroom, exterior, roof, utility.
- Read the condition. In each photo the model looks for things a person would note: dated cabinets and counters, worn or damaged flooring, ceiling stains, peeling paint, broken fixtures, old appliances, siding and roof wear.
- Choose a scope. For each category it decides between leaving it, repairing it and replacing it, ideally at the finish level of the exit (a rental scope is lighter than a flip scope).
- Estimate quantities. Square footage, room counts and what's visible become quantities: square feet of flooring, linear feet of cabinets, number of fixtures.
- Price it. Quantities are multiplied by unit costs for materials and labor, adjusted for location.
- Show the breakdown. The useful tools give you categories or line items you can change, not a single total.
Bricked, for example, says on its site that it "reads the listing photos of each home," uses "a national database of material costs by zip code and aggregates local labor pricing from the web," and "generates itemized repair estimates based on the property's size, condition, and location," which you can adjust in its chat.
What photos can and can't show
| Usually visible in photos | Usually not visible |
|---|---|
| Kitchen and bath finishes, cabinets, counters, fixtures | Wiring inside walls, panel capacity, aluminum branch wiring |
| Flooring type and wear | Subfloor rot, slab cracks under flooring |
| Paint, drywall damage, ceiling stains | The leak causing the stain, mold inside the wall |
| Appliance age and condition (roughly) | Whether the HVAC, water heater and plumbing actually work |
| Roof surface and siding from the ground | Roof decking, attic condition unless photographed |
| Obvious cracks and sagging | Foundation movement, sewer line, termite damage |
That right-hand column is where budgets blow up, and photos are weaker there than a person in the house. Even a professional home inspection is, in InterNACHI's Standards of Practice, "a non-invasive, visual examination of the accessible areas," and the same standards say "an inspection will not identify concealed or latent defects." A model looking at photos is working with less than that.
Is it accurate?
On visible, standard work (paint, flooring, kitchens, baths, fixtures) with a full set of photos, a photo-based estimate can land close to what a contractor would price. Accuracy drops for reasons you can predict:
- Photo coverage. Five listing photos of a three-bedroom house can't produce the estimate thirty photos can. Listing photos are picked to sell the house, so the worst rooms and the utility areas are often missing.
- Photo quality. Dark, blurry or zoomed-in photos hide condition.
- Hidden systems. Anything in the right-hand column above is a guess from age and clues, not something the model saw.
- Quantities. Roof area, linear feet of cabinets and exact room sizes are estimated, not measured.
- Local labor. Labor rates differ by metro and by contractor; cost data can't know what your crew charges.
- Finish level. If the tool assumes a flip finish and your buyer is a landlord, the number is right for the wrong job.
We looked for published accuracy numbers on October 29, 2026. Bricked, ChatARV and Propera call their results accurate in marketing copy and customer quotes, but none of their sites publishes a measured error rate for repair estimates, and we don't quote one for Deal Run here either. The honest way to know how a tool does in your market is to test it.
How to test an AI repair tool yourself
- Pick three to five houses where you already have a contractor's bid or the final rehab cost.
- Run each one with the same photos you had before the work started.
- Compare category by category, not just the total. A tool can hit the total by being high on kitchens and low on systems.
- Note where it's consistently off (for example, always light on roofs in your area) and correct for that on every new deal.
When to verify with a contractor
- Screening leads: the AI estimate on its own is fine for go or no-go.
- Making an offer: AI estimate plus a contingency sized to what the photos didn't show; our guide to estimating repairs without seeing the property has a contingency table by access level.
- During the inspection period: a contractor or specialist on anything structural or system-level (foundation, roof, electrical, plumbing, sewer) before you're committed.
- Marketing a wholesale deal: put the estimate in your deal package with the breakdown and say it's an estimate pending the buyer's inspection.
- Keeping the house to renovate: get a full written bid before you close. Don't set a renovation budget from photos alone.
Getting better results from any AI estimate
- Shoot every room wide from a corner, in daylight with the lights on.
- Photograph the systems: the electrical panel and its label, the AC and furnace data plates, the water heater, under every sink, the attic and the crawlspace.
- Add what you know. If the seller says the roof is three years old, or you have an HVAC quote, change that category. The model only knows what the photos show.
- Set the exit before you read the number, so the scope matches the buyer.
For the full step-by-step method that these tools automate, see the best way to estimate repairs.
AI repair estimate tools today
Here is how three AI deal-analysis tools describe their repair estimates on their own sites, in alphabetical order with our product last. For a ranked comparison of repair apps, see our best repair estimator apps roundup.
| Tool | How the repair estimate is made (per its site) | Price (checked Oct 29, 2026) |
|---|---|---|
| Bricked | Reads listing photos; itemized repair estimate from size, condition and location; adjust in chat or save your own rates | Basic $89/mo or $69/mo billed yearly (100 comps a month); Growth $199/mo or $149/mo yearly (300 comps). No free trial shown on its pricing page. |
| ChatARV | Site doesn't say how; you can adjust repair costs, assignment fees and MAO percentages | Not published on their site (checked October 29, 2026). At my September 20, 2026 signup: Basic $89/mo for 100 comp reports, after a 7-day trial. |
| Propera | From a text description of the property's condition, not photos | Starter $49/mo (25 properties), Pro $149/mo (100), Business $299/mo (500); run packs from 5 runs for $29; 7-day trial covering 3 properties. |
| Deal Run (our product) | AI repair estimate from the property's photos, for flip, rental and wholesale exits | Analyze $29/mo; Analyze + Sell $99/mo or $83/mo billed annually. Or pay per deal: $2 to analyze, $20 with 100 buyers. |
Bricked's strength is speed: comps, repairs and an offer from an address, with the repair line items editable in chat. Propera's text approach suits you when you have no photos but know the condition. ChatARV centers on comps and lets you change the repair number yourself. When I signed up for ChatARV on September 20, 2026, the paywall offered 7 days free, then $89/mo, with a card or wallet up front; on our ChatARV alternatives page we compare the ARV side of these tools.
Doing This in Deal Run
Deal Run is our product. It makes AI repair estimates from the property's photos, next to sold and rental comps, ARV and a max offer for flip, rental and wholesale exits. The same limits on this page apply to it: it only sees what the photos show, so you still add contingency and get a contractor on the big systems. See how Deal Run's repair estimates work.
The public demo on the homepage is free: enter an address and see the analysis before you pay. Its repair step uses sample photos, so an estimate for your own house comes when you add that house's photos in the app. Analyze is $29/mo for deal analysis and includes one disposition deal. Analyze + Sell is $99/mo (or $83/mo billed annually). Or pay per deal: $2 to analyze, $20 with 100 buyers.
Frequently Asked Questions
Can AI estimate repair costs from photos?
Yes. Photo-based tools identify the rooms and items in each photo, judge their condition, choose a repair or replacement scope, and price it with cost data for the property's area. The result is an itemized first draft you can edit. It only covers what the photos show, so it works best with a full set of photos and as a starting point, not a contractor's bid.
Are AI repair estimates accurate?
They can be close on visible, standard work such as paint, flooring, kitchens and baths when the photos cover the whole house. They miss what the photos do not show, such as wiring, pipes, the sewer line and structure, and their prices can be off where local labor costs differ from their data. None of the vendor sites we checked publish a measured error rate. Test any tool on a few houses where you already have real bids.
What can't AI see in property photos?
Anything behind walls, under floors or underground: wiring, supply and drain pipes, the sewer line, hidden water damage, termite damage, and most foundation problems. It also cannot test whether the HVAC, water heater or plumbing actually work, and it cannot see rooms nobody photographed. Even a home inspection is a visual exam that will not find concealed defects.
Should I still get a contractor bid if I use an AI estimate?
Yes, before you commit money on the big items. Use the AI estimate to screen leads and make a first offer, then get a contractor or specialist to look at anything structural or system-level during your inspection period: foundation, roof, electrical, plumbing and sewer. If you are keeping the house to renovate, get a full bid before you close.
Which tools give AI repair estimates for investors?
Bricked says it reads listing photos and builds itemized repair estimates. ChatARV includes repair costs you can adjust but does not say on its site how they are made. Propera builds repair estimates from a text description of the condition. Deal Run, our product, makes AI repair estimates from the property's photos. Prices and limits differ; check each vendor's own site.
Sources and Check Dates
- Bricked homepage — bricked.ai — reads listing photos, cost data by zip code and local labor pricing, itemized estimates, adjust in chat; no measured accuracy figure for repairs. Checked October 29, 2026.
- Bricked pricing — bricked.ai/pricing — Basic and Growth monthly and yearly prices, comps per month, AI repair estimates on every plan, no trial shown. Checked October 29, 2026.
- ChatARV homepage and pricing — chatarv.ai — adjustable repair costs, assignment fees and MAO percentages; no method or accuracy figure; no prices on the public site. Checked October 29, 2026.
- ChatARV plan price and trial — my own ChatARV signup, September 20, 2026, hands-on.
- Propera homepage — getpropera.com — repair estimates from a text description of condition; no accuracy figure. Checked October 29, 2026.
- Propera pricing — getpropera.com/pricing — plan prices and properties per month, run packs, 7-day trial. Checked October 29, 2026.
- InterNACHI Home Inspection Standards of Practice — nachi.org/sop.htm — definition of a home inspection as a visual examination; concealed or latent defects not identified. Checked October 29, 2026.
- Deal Run plans and features: our own product and pricing page. Checked October 29, 2026.