How InspectionTriage analyzes an inspection report

1. Upload and payment. You upload a PDF of up to 25 MB and enter a ZIP code or state. Analysis starts after the $19 payment clears through Stripe.
2. Reading the report. Text is pulled from each page. Pages with little or no text, such as scans or photos of pages, go through optical character recognition (OCR) so they can be read too.
3. Finding extraction. AI models read the report in sections and list the defects, concerns, and recommendations the inspector recorded, along with the page each one came from. Boilerplate is skipped and repeated findings are merged.
4. Plain-English enrichment. Each finding gets a short explanation, why it matters, the system it belongs to, a severity (low, medium, or high), a timeframe (now, within 12 months, or in 1 to 5 years), next steps, and a specialist follow-up when one is warranted.
5. Cost ranges. Each finding receives a low, typical, and high planning range with a confidence level and stated assumptions. The section below explains how.
6. Summary and negotiation options. The findings are rolled up into a decision snapshot, a 7-day action plan, a budget roadmap by timeframe, and two negotiation approaches with suggested ask and settlement ranges.
7. Checks and delivery. AI output is checked against a fixed structure (a schema) before the packet is assembled. You get an interactive web view, a Decision Packet PDF, a Negotiation Playbook PDF, and a Findings CSV.

Every finding is placed in one of four priority tiers based on its timeframe, severity, the system involved, and its typical cost.

• Critical: needs attention now, is high severity, and either involves the structure, roof, water and drainage, or safety, or has a typical cost of $3,000 or more.
• Significant: needs attention now, or is a high-severity safety or water issue expected within the year.
• Moderate: high severity, one of the report’s largest cost drivers, a strong negotiation point, or calls for a specialist visit soon.
• Monitor: everything else. These are usually maintenance items or things to watch.

The tiers sort the report so you can see what to deal with first. They are not a judgment on whether to buy the house.

Cost ranges come from fixed rules rather than being written fresh by the AI model for each report. Each finding is matched to a repair category in InspectionTriage’s cost rules. Each category has a baseline low, typical, and high range, a starting confidence level, and written assumptions about the likely scope of work. When a finding does not match a specific category, it gets a general range for its system, such as plumbing or electrical.

The baseline is then adjusted with a coarse regional multiplier based on the state you enter or the first digit of your ZIP code. This shifts the whole range up or down for broad differences in labor and material costs. It does not reflect prices in your specific town.

Confidence is lowered when a finding needs more investigation before anyone can price it, or when it calls for a structural engineer, foundation contractor, or roofer. If the cost step fails for a report, every finding gets a deliberately wide range marked low confidence, and the packet tells you to get on-site quotes.

These are planning ranges for budgeting and for deciding which items need a quote. They are not contractor quotes, and they are not built from bids in your area. Actual costs depend on the real scope of work, local labor, materials, and what a contractor finds on site. The disclaimers cover this in more detail.

• AI models can miss a finding, misread one, or place it on the wrong page. Running the same report twice can produce slightly different results.
• The packet only knows what the report says. It cannot see the house, and it cannot find problems the inspector did not record.
• Poor-quality scans can lose text during OCR.
• Very long reports are read in sections with a size cap, so text near the end of an unusually long report may not be read in full.
• The regional adjustment is coarse. Costs in your area can fall outside the range.
• Results are automated and are not a licensed inspection, an appraisal, or professional advice.

Every finding includes a reference to the page of your report it came from. Read those pages yourself, ask your inspector about anything unclear, and get quotes for items marked high variance or low confidence before relying on a number. Your agent can use the Negotiation Playbook as a starting point for the conversation with the seller.

You can see a full example on the sample packet page. Questions about your packet? Use the contact page.

The free scan on the home page is a smaller version of the same idea. It reads up to six automatically selected pages, and no more than 18,000 characters, from a text-based PDF. It returns up to three findings that drive the estimate, with broad planning ranges, plus up to two lower-leverage items. It can miss findings elsewhere in the report, and it does not include page references, suggested asks, a budget roadmap, or the downloadable packet. Scanned reports need the paid upload.