AI for Real Estate Investors (2026): Deal Analysis and Underwriting, With a Worked Duplex Example

The short answer

How should a real estate investor use AI? Let it read and draft, and keep it away from the inputs and the decision. A general assistant like Claude or ChatGPT is good at screening listings against your buy box, pulling numbers out of a rent roll or a seller's operating statement, drafting a pro forma from inputs you supply, and running stress tests. It's bad at supplying rents, comps, taxes or insurance, and it won't stop you from leaving out the three lines that decide most small rental deals. Below is a worked duplex example where those three lines take the same deal from a 7.7% to a 0.6% cash-on-cash return.

This page is for investors with 1 to 50 doors and the small shops behind them: buy-and-hold, BRRRR, small multifamily. If you broker or own commercial buildings, AI for commercial real estate covers leases, offering memorandums and comps memos. If you're an agent, start at the AI for real estate hub.

What should an investor actually hand to AI?

Split every deal into inputs, math and judgment. AI is useful on the work around all three, and dangerous when it quietly becomes the source of any of them.

The comps row deserves its own warning. A valuation model has never seen the property, and a chat assistant hasn't either. Where automated valuations hold up and where they break is covered in can AI value a property, and the same limits apply to any rent or value an AI hands you.

A worked example: underwriting a duplex with AI

Here's a full pro forma on a hypothetical duplex. Every input is an assumption I picked to make the math readable, not market data. Replace each one with your own numbers before you believe any result. The math was run with the standard amortization formula, not by a chatbot.

• Price: $300,000. 25% down ($75,000) plus $9,000 of closing costs, so $84,000 of cash in.
• Loan: $225,000, 30-year fixed at an assumed 7.0%. Principal and interest: $1,496.93 a month, about $17,963 a year.
• Rent: two units at $1,400, so $33,600 a year gross.
• Taxes and insurance: $4,500 and $2,000 a year (assumed; get the county's number and a real quote).
• Repairs: 8% of gross rent, $2,688.

Version 1: the pro forma most first drafts produce

Ask an assistant to "analyze this duplex" with only the inputs above and nothing guarantees it adds a vacancy line, a capital reserve or management. Leave those three out and the deal looks like this:

• Net operating income: $33,600 minus $9,188 of expenses = $24,412
• Cap rate: 8.14%
• Cash flow after the mortgage: $6,449 a year, about $537 a month
• Cash-on-cash return: 7.68% · Debt service coverage ratio (DSCR): 1.36

Version 2: the same deal with the three lines that move this deal

Now add 5% vacancy ($1,680), a 5% capital reserve for roofs, HVAC and turns ($1,680), and 8% management on collected rent ($2,554). Even if you self-manage, your time has a price, and a lender or a future buyer will underwrite it.

Same building, same rents, same loan. The only difference is three lines a careful investor always includes. That's the whole argument for using AI as a drafter and a checker, not as the underwriter: it's very good at producing a clean, confident table from whatever you gave it.

Two stress tests worth running every time

• Rents come in 5% lower ($1,330 a unit): NOI drops to $17,248, cash flow goes to about -$715 a year, and DSCR falls to 0.96. The property no longer covers its mortgage from operations.
• The rate comes in at 7.5%, not 7.0%: the payment rises to $1,573.23 a month, cash flow goes to about -$380 a year, and DSCR is 0.98.

This is where AI earns its keep. Ask it for ten scenarios like these, have it lay them out in one table with each formula shown, and rebuild the two that worry you in your spreadsheet. On this example, the deal needs about $1,550 a unit, not $1,400, before it shows a 1.18 DSCR and roughly $268 a month of cash flow. That one sentence is a better offer strategy than the original pro forma.

Which AI tools do real estate investors use for deal analysis?

Two kinds, and they do different jobs. Prices are from each vendor's own pricing page, read October 8, 2026.

ChatGPT's paid plans do the same assistant job; pick the one your partners already use. The pairing that works is an assistant for the reading and drafting plus a calculator or your own spreadsheet for the math. A solo investor can run that for $17 to $30 a month. Per-deal AI underwriting products also exist; get the price in writing and compare it to what the two tools above already do.

Where AI underwriting goes wrong

• Invented inputs. Ask "what would this rent for?" and you'll get a number. It has no source you can check. Every input in your pro forma needs one.
• Today's taxes, not next year's. In many places the assessment changes after a sale. Use the county's rules for the new owner, not the seller's bill.
• Insurance and financing guesses. Both move fast and vary by property. Get a quote and a term sheet before the numbers matter.
• Arithmetic slips. Make it show each formula, then let the spreadsheet calculate.
• Seller documents taken at face value. AI reads a seller's P&L faster than you do. It can't tell you whether the P&L is true. Match it to bank statements and leases.

One plain note: this is an example of method, not financial, tax or legal advice. Run your own numbers and talk to your lender, CPA or attorney before you commit to a deal.

Who can help a real estate investor set up AI for deal analysis?

I'm Justin McKelvey, an AI consultant in Austin, Texas. I help small teams pick one assistant, write down what goes into it and what never does, and set up the first workflows, like the buy-box screen and the stress-test table above. The best first step costs nothing: the 6 AI tactics for real estate guide covers residential and commercial operators and lands in your inbox. If you run an investment shop and want a written plan for your own deal flow, the $2,500 AI Readiness Assessment is the paid path: two weeks and a roadmap.

What to do this week

Take the last deal you passed on and the last one you bought. Give an assistant the inputs for both, ask for a pro forma with vacancy, capital reserves and management as separate lines and every formula shown, then ask for five stress tests. Rebuild the result in your spreadsheet. If the AI's version and yours disagree, you've found either a mistake in the draft or a line you've been leaving out.

Sources

• Anthropic, Claude pricing, claude.com/pricing (read October 8, 2026)
• DealCheck, pricing, dealcheck.io/pricing (read October 8, 2026)
• Duplex pro forma: all inputs are labeled assumptions; payments computed with the standard 30-year amortization formula (October 8, 2026)
• DataForSEO, US search volume and keyword difficulty for "ai for real estate investors", "ai for real estate investing" and "ai underwriting real estate" (read October 8, 2026)