← Back to Blog

AI Real Estate Deal Analyzer: Screen Rental Listings With Claude, Keep the Math in a Spreadsheet

An AI model can read a listing, pull out the numbers and argue with the seller's assumptions in a minute. It should not be the thing doing your cash flow math. Here is a setup that splits the jobs correctly.

Industry9 min read•

Forty listings, one evening, two worth a showing

If you buy small rentals, you know the routine. A wholesaler email, a few Zillow alerts and an agent's off-market list land in the same week, and each one says strong cash flow. Running every one through a calculator takes 15 to 20 minutes, so you skim, and the deal you skip is the one someone else buys.

An AI model is good at the reading part of that job: pulling price, rent, taxes and unit mix out of a messy listing, spotting the assumption that is doing all the work, and telling you what to ask the seller. It is bad at the part people most want it to do, which is the arithmetic. This guide sets up a deal analyzer that uses each tool for what it does well, with a worked duplex example and the real costs.

What the multi-agent version on this site does, and what it does not prove

The listing on this site, An AI Real Estate Investment Analyzer Using Multi Agent Architecture, describes a system built with OpenAI models. A coordinator agent sends each property to three specialists in parallel. A market agent looks at median prices, days on market, inventory and price trends. A financial agent runs cash flow projections, cap rate, ROI and mortgage analysis. A property agent reviews listing details, photos, condition reports and comparable sales. Each returns an assessment with a confidence score, and the coordinator writes one recommendation in a standard report.

The structure is sound: splitting the work means you can fix the financial piece without touching the others. But the source reports no investment returns and no time saved, and it does not say where the market data comes from. Treat it as an architecture, not evidence that AI picks better deals. For one investor screening a few dozen listings a month, you do not need four agents. You need one assistant with good instructions and one spreadsheet.

Why the math belongs in a spreadsheet

Language models predict text. They often get a mortgage payment right, and sometimes they get it slightly wrong in a way that looks exactly as confident. On a deal where the margin is $87 a month, slightly wrong is the whole answer. A spreadsheet formula is boring and correct every time, and you can see exactly what it did.

So the split is: the AI extracts inputs and questions them, the spreadsheet computes, and the AI reads the result back to you in plain language with the risks. Claude and ChatGPT can both run code on uploaded files, which helps, but a sheet you built and tested is still the thing you should trust with money.

A worked example: a duplex that looks better than it is

Take a duplex listed at $320,000. The listing claims each unit rents for $1,550, so $3,100 a month. You plan 25 percent down, $80,000, plus an estimated 3 percent in closing costs, $9,600, for $89,600 of cash in. The remaining $240,000 loan at an assumed 7.25 percent over 30 years is about $1,637 a month in principal and interest. Use a current quote from your lender, since the rate here is only for illustration.

Operating costs, using common rules of thumb you should replace with local figures: 5 percent vacancy ($155), property tax at 1.2 percent of price ($320 a month), insurance at $150, maintenance at 8 percent of rent ($248), capital expenses at 5 percent ($155), management at 8 percent ($248) and water and trash at $100. That totals $1,376 a month, leaving net operating income of $1,724 a month, or $20,688 a year.

The spreadsheet then gives you a cap rate of 6.5 percent, which sounds reasonable in many markets, monthly cash flow of about $87 and a cash-on-cash return of about 1.2 percent. Now drop rents 10 percent, to $2,790, because the listing's $1,550 a unit turns out to be the best unit on the block. Vacancy, maintenance, capital expenses and management fall with rent, but taxes, insurance and utilities do not, and the deal goes to about $143 a month negative.

That is the analyzer working. The AI's value was not computing 6.5 percent. It was reading the listing, noticing the rent claim was the seller's number, and telling you to check it before you drove out.

Worked example: $320,000 duplex, 25% down, 7.25% assumed rate Income and costs Rent: $3,100 a month Operating costs: $1,376 NOI: $1,724 a month Mortgage: $1,637 Cash in: $89,600 Base case Cap rate: 6.5% Cash flow: $87 a month Cash on cash: 1.2% Looks fine on cap rate Rents 10% lower Rent: $2,790 a month NOI: about $1,495 Cash flow: about -$143 The deal breaks Illustrative numbers. The spreadsheet computes these. The AI's job is to question the $3,100 rent and the 8% maintenance before you trust them.

Step by step: build the analyzer in an afternoon

Step 1: Build and test the calculation sheet

In Google Sheets or Excel, make one tab with input cells (price, down payment percent, closing cost percent, rate, term, monthly rent per unit, tax rate, insurance, and percentages for vacancy, maintenance, capital expenses and management) and output cells for loan amount, monthly payment, NOI, cap rate, cash flow and cash-on-cash. Use the PMT function for the payment. Test it against one deal you already own or a bank's mortgage calculator until the payment matches to the dollar. Add a second block that recalculates with rent 10 percent lower and rate 1 point higher.

Step 2: Write down your buy box

On one page write what you will and will not buy: price range, unit count, neighborhoods, minimum cash flow per door under the stress case, minimum cash-on-cash, maximum age of roof or systems you will accept without a discount. The AI will apply this every time, but only if it is written.

Step 3: Set up a Claude project with instructions

Create a Claude project called deal screening. Upload your buy box and a blank copy of your input sheet. In the project instructions say: when I paste or upload a listing, extract every input my sheet needs into a table; mark each value as stated in the listing, estimated by you, or missing; never compute payments, cap rate or cash flow yourself; list the three assumptions that most affect the result; and write five questions for the seller or agent. Claude Pro is enough for this.

Step 4: Paste the listing, check the extraction

Paste the listing text or upload the PDF or offering memo. Read the extraction table against the listing for the first ten deals. Models sometimes pick up the wrong tax figure or treat a pro forma rent as an actual rent, which is exactly the error that kills deals. Fix the table, then type the inputs into your sheet.

Step 5: Check rents and taxes outside the AI

Look up current rents for similar units on rental listings in that neighborhood or with a paid rent comps tool, and look up the actual tax bill on the county site. Do not let the AI estimate local rents from memory: it has no live market data unless you give it some. This step takes ten minutes and is the one that saves you.

Step 6: Paste the outputs back for a verdict

Paste the sheet's base and stress-case outputs into the same conversation and ask: against my buy box, pass, offer or reject, and if offer, what price makes the stress case break even. Then use the sheet, not the chat, to find that price by changing the price input until stress-case cash flow is zero.

Step 7: Keep a deal log

Log every screened deal with the verdict and the one reason. After two months you will see whether the assistant's flags match what you found on showings, and you can tighten the instructions.

What this costs

As of November 2026: Claude Pro is $20 a month, or $17 a month billed annually, per Anthropic's pricing page. Google Sheets is free. If you would rather buy than build, DealCheck has a free Starter plan with 15 saved properties, Plus at $10 a month and Pro at $20 a month with unlimited properties and comps, per its pricing page. A reasonable approach is to use DealCheck or your own sheet for the math and Claude for reading listings and questioning assumptions. Either way the total is $20 to $40 a month. The listing on this site puts its multi-agent build at under $50 a month, plus the weeks of work an advanced build takes.

When not to use AI for this

Skip it for the final decision on any deal you are about to put money into. Use an inspection, real rent comps, the actual tax bill, insurance quotes and your lender. Skip the multi-agent build unless you screen hundreds of listings a month or plan to sell the tool, because one assistant and a sheet do the same job for a single investor. And do not share a seller's confidential offering memo with any AI tool if your agreement with them forbids it.

Frequently Asked Questions

Can ChatGPT or Claude analyze a rental property?▾

They can read a listing, extract the numbers, flag weak assumptions and draft questions for the seller well. They should not be your calculator: put the payment, cap rate and cash flow math in a spreadsheet you have tested, because a small arithmetic error on a thin deal changes the answer.

What is a good cap rate for a rental property?▾

It depends on the market, property type and interest rates, so there is no single number. Compare a deal's cap rate to recent sales of similar properties nearby, and check cash flow under a stress case, since a fine-looking cap rate can still mean negative cash flow at today's borrowing costs.

Is there a free AI real estate deal analyzer?▾

DealCheck has a free Starter plan for up to 15 saved properties as of November 2026, and a free Google Sheets template plus a general AI assistant can do the screening. Most dedicated tools charge for comps and unlimited properties.

How accurate are AI rent estimates?▾

Without live data, not reliable. A general AI model has no current local rental listings, so check rents yourself on rental listing sites or a paid comps tool and give the model those numbers.

Do I need a multi-agent system to analyze deals?▾

No. A multi-agent setup splits market, financial and property analysis across specialist agents, which helps at high volume or if you are building a product. One investor screening a few dozen deals a month gets the same benefit from one well-instructed assistant and a spreadsheet.

How do I calculate cash-on-cash return?▾

Divide a year of cash flow after the mortgage by the total cash you put in, including the down payment and closing costs. In the worked example here, $1,041 of annual cash flow on $89,600 invested is about 1.2 percent.

Keep Reading

Explore Related Niches

Related Strategies

Real EstateReal Estate AgentsFreelance/Agency

A real estate agent grew to $400K GCI by building AI systems that run the business without a full team.

BeginnerUnder $50/moVery High ($10K+/mo)
View Strategy
Lead GenReal Estate AgentsFreelance/Agency

A first AI automation agency client worth $20,000 came from a casual networking conversation, not cold outreach or paid ads.

IntermediateUnder $50/moVery High ($10K+/mo)
View Strategy
Real EstateReal Estate AgentsInternal Tool

An Obsidian knowledge base connected to Claude Code that gives the AI full context on a three company, 50 plus property real estate operation from the first message.

BeginnerFreeHigh ($2K-10K/mo)
View Strategy

Keep Exploring the Directory

Browse more case studies, compare tools, or dig into the full tool stack behind these workflows.

Want more strategies like this?

Get weekly AI agent case studies in your inbox.