AI Real Estate Investment Analysis: Vet a Rental Deal Step by Step

Use AI to organise rent comps, set out an assumptions table and write the spreadsheet formulas, then verify by hand what it cannot know or often gets wrong: achieved rent, property tax after sale, insurance, the loan payment and the NOI-to-cap-rate arithmetic. Any figure the assistant gives without showing its working counts as unchecked.

A rental deal that looks good on the seller's sheet often looks thin on yours, and the aim of this checklist is to reach a confident no, or a defensible offer price, quickly. Seller sheets flatter a deal by leaving out vacancy, reserves, management and reassessed taxes, and an assistant handed that sheet will repeat the flattering numbers in tidy prose. Every property and figure below is invented, for one imaginary house. None of it is financial or tax advice; a lender and an accountant should check yours.

Follow me on Instagram@sagnikteaches

Write your screen before you open the listing

The easiest way to be talked into a deal is to set your standards after seeing the numbers. Write three or four pass-or-fail rules first, in plain figures, and paste them into the Project or chat that holds your deal sheet so every analysis ends with a verdict against them. The rules below are an example, not market norms; your lender will have its own minimums, so ask for them in writing.

Connect on LinkedInSagnik Bhattacharya

• Debt service coverage ratio (DSCR) of at least 1.20. This is net operating income divided by a year of loan payments. At 1.00 the property only just pays its debt.
• Cash-on-cash return of at least 5%. This is a year's cash flow after loan payments divided by every dollar you put in, including closing costs and repairs.
• Still positive with rent $100 lower and vacancy at 10%. A deal that only works when everything goes right is a forecast, not an analysis.

Two more terms are used throughout. Net operating income (NOI) is the rent you actually collect minus the costs of running the property, before any loan payments. The cap rate is NOI divided by the purchase price. A property that costs $245,000 and earns $15,204 of NOI has a cap rate of 6.2%. If you are still deciding which listings deserve this much work, the screening tutorial for property investors comes first; this one is for the single deal that has passed it.

Subscribe on YouTube@sagnikteaches

The assumptions table: every number, its source, and who checks it

Keep one table that lists every input, where it came from and what you will do by hand to confirm it. Ask the assistant to maintain it as the deal changes, but type the first version yourself. The imaginary deal is a three-bedroom house with a 75% loan.

The tax row is a common trap. The seller's bill reflects what the previous owner paid, and in many places a sale triggers a reassessment, so the buyer's bill can be higher. The only reliable answer comes from the assessor, and an assistant that quotes the seller's figure back to you is repeating a number that may not apply to you. Mark every row you have not checked yet with a question mark; a deal with five question marks is not ready for an offer.

Rent comps: let AI organise them, not find them

An assistant is useful for the tedious part of rent comps: putting messy listings into one table, applying the same adjustments to each and showing the arithmetic. It is a poor source of the comps themselves. Unless it has live search, any rents it supplies come from training data or invention, and even with search it may quote a stale listing. Collect four to six comparable rentals yourself from listing sites and your own contacts, paste the details, and give it fixed rules.

Below are five rental listings I collected myself, with addresses removed.
Subject: 3 bed, 2 bath, 1,450 sq ft, built 1994.
Use only these rows. Do not search the web and do not add comps.
For each comp, adjust the asking rent to the subject using $0.50 per sq ft of
size difference and $50 per half bathroom, and show the arithmetic.
Flag weak evidence: still listed after 30 or more days, a different bedroom
count, or unknown status. Then give the median and range of the adjusted rents
for the comps you treat as comparable, and say which comp you dropped and why.

A: 3/2, 1,420 sq ft, asking $2,250, leased in 14 days
B: 3/2, 1,520 sq ft, asking $2,300, still listed after 31 days
C: 3/2, 1,450 sq ft, asking $2,250, leased in 6 days
D: 3/1.5, 1,380 sq ft, asking $2,150, leased in 19 days
E: 4/2, 1,600 sq ft, asking $2,450, leased in 11 days

A careful answer, illustrative like everything else in this tutorial, comes back as a table:

The four kept comps adjust to $2,235 to $2,265, with a median of $2,257.50. The base case uses $2,250, which matches the strongest comp and sits just below the median. That rounding down is deliberate: comp B has not leased, so its asking rent is a hope, not a result, and the manager's written opinion of $2,250 agrees.

Two things to do by hand. First, check that the adjustment rates are yours: $0.50 a square foot and $50 a half bath are placeholders invented for this example, and a manager who knows the area can replace them with something better. Second, ask whether each comp is a true asking-versus-achieved comparison. A listing that says $2,250 and leased in six days may have leased at $2,150 after a concession, and nothing in an advert shows that. When it matters, ask the manager what the comps actually signed for.

The sheet itself: inputs you can see, formulas you can audit

The safest way to use an assistant for the calculation is to ask for formulas, not answers. A formula sits in a cell where you can read it, change an input and watch the result move. A paragraph of numbers gives you nothing to audit.

Build a rental deal sheet as spreadsheet formulas, not as results.
Put every input in its own labelled cell: price, down payment %, interest rate,
monthly rent, vacancy %, property tax, insurance, turnover cost, combined
repairs, reserve and management %, closing cost %, prep cost.
Calculate: loan amount, monthly payment using PMT, effective income, operating
costs, NOI, cap rate, annual debt service, DSCR, annual cash flow, cash invested
and cash-on-cash return. Use no figure I have not given you.
Show the cell layout so I can type it in.

A good answer produces a layout like this one, which you can build in Excel or Google Sheets in a few minutes (the Google Sheets AI prompts tutorial helps if spreadsheets are new to you):

Inputs                         Calculations
B1  Price           245000     B13 Loan            =B1*(1-B2)
B2  Down payment %  25%        B14 Monthly payment =PMT(B3/12,360,-B13)
B3  Interest rate   6.75%      B15 Effective income =B4*12*(1-B5)
B4  Monthly rent    2250       B16 Operating costs =B6+B7+B8+B9*B15
B5  Vacancy %       6%         B17 NOI             =B15-B16
B6  Property tax    3100       B18 Cap rate        =B17/B1
B7  Insurance       1400       B19 Debt service    =B14*12
B8  Turnover        600        B20 DSCR            =B17/B19
B9  Repairs+reserve+mgmt %  20%  B21 Cash flow     =B17-B19
B10 Closing cost %  3%         B22 Cash invested   =B1*B2+B1*B10+B11
B11 Prep cost       5000       B23 Cash-on-cash    =B21/B22

PMT returns a negative number for a positive loan, which is why the loan sits behind a minus sign. Combining repairs (7%), reserve (5%) and management (8%) into one 20% input keeps the layout short; split them if you want to flex each one. The results for the imaginary house follow, per month and per year.

The headline ratios follow from those lines. The cap rate is $15,204 divided by $245,000, which is 6.2%. DSCR is $15,204 divided by $14,302, or 1.06. Cash invested is $61,250 down, $7,350 of closing costs and $5,000 of preparation, which makes $73,600, so cash-on-cash is $902 divided by $73,600, or 1.2%. Against the screen written earlier, the deal misses the DSCR and cash-on-cash rules at the asking price, and the stress test below shows it fails the third rule as well.

Checking the seller's cap rate sheet line by line

Suppose the listing agent's flyer for the same house claims a cap rate of 8.9%. The flyer shows rent of $27,000 a year and three costs: property tax of $2,450, insurance of $1,400 and maintenance at 5% of rent, which is $1,350. That leaves NOI of $21,800, and $21,800 divided by $245,000 is 8.9%. Nothing on the sheet is invented, and yet it is $6,596 higher than your NOI. A short bridge shows where the gap comes from, and each line gives you a question to ask.

The bridge lands on $15,204, and the cap rate falls from 8.9% to 6.2%. Each step is a legitimate difference of method, not a sign that the seller lied. The sheet used last year's tax bill, left out vacancy, reserves and management, and rounded maintenance down. A check of the arithmetic will pass, because the arithmetic on the sheet is correct; the problem is what the sheet leaves out. So ask a different question of the assistant: "List every cost an owner normally bears that does not appear on this sheet." Then rebuild the bridge yourself.

Where the assistant's arithmetic goes wrong

Assistants write fluent explanations, and fluency hides arithmetic slips. These are the kinds of error to expect when an assistant is asked for results rather than formulas, shown here as an illustration of one plausible run, set against the hand-checked figures.

Each error was small on its own, but together they turned a 1.2% return into a 6.6% one, which would pass the screen. Three quick sanity checks catch most of them without a calculator in sight. First, check the payment against a rule of thumb: a 30-year loan at 6.75% costs about $6.49 a month for every $1,000 borrowed, so $183,750 should cost about $1,192, and a figure near $1,034 is too low to be a fully repaying loan. Second, remember that a cap rate must be lower than the gross yield (rent divided by price), because it takes costs off. If the two match, or the cap rate is higher, something has gone wrong. Third, compare the expense total with the rough rule that costs run at about half of gross rent. It is only a rule of thumb, and a house where the tenant pays utilities can come in below it, but expenses far below it deserve a question.

The deeper fix is the working habit already described: formulas in visible cells, the assistant's own results treated as a claim to verify, and three numbers recomputed by hand every time (payment, NOI, cap rate). For a closer look at how reliably assistants handle yield calculations, see the tutorial on whether ChatGPT can calculate yield reliably.

Stress test: how thin is the margin, and what price fixes it

A base case is a single guess. A stress test changes one input at a time and shows how quickly the cash flow disappears. For the imaginary house at $245,000, the base case leaves $75 a month, and the table shows how little it takes to erase it.

A $100 drop in rent is enough to take the house to zero. The cap rate of 6.2% sits below the 6.75% loan rate, which is why the property cannot cover its own debt comfortably: borrowing costs more than the property earns. This is the usual shape of a deal that looks fine on a flyer and thin on a proper sheet, and the reasonable response is to ask what price changes it. Re-run the sheet at lower offers, keeping the 75% loan, the same rate and taxes in proportion to the price.

The first price on the list that clears a DSCR of 1.20 is $220,000, and the 5% cash-on-cash rule needs about $212,000, which is 13.5% below the asking price. Whether a seller accepts that is a negotiation question, not a maths question. A deal that only works at a price nobody will take is a no, and reaching that answer in an afternoon is the point of the exercise.

A one-page decision memo, filled in

End every analysis with a memo you could hand to a partner, a lender or your future self. Ask the assistant to draft it from your verified sheet, then read every figure against the sheet before you keep it. This is the memo for the imaginary house.

Verdict: pass at the asking price of $245,000. About $212,000 is the highest price at which all three rules pass, so open below it and walk away above it. ($221,000 would clear the DSCR rule alone, but not the 5% cash-on-cash rule.)

Why: at asking, NOI is $15,204, the cap rate 6.2%, DSCR 1.06, cash flow $75 a month and cash-on-cash 1.2%, so it misses all three screen rules. At $212,000, NOI is $15,622, the cap rate 7.4%, DSCR 1.26, cash flow $271 a month and cash-on-cash 5.0%. With rent $100 lower and vacancy at 10%, cash flow stays positive at $127 a month.

Still to verify by hand: achieved rents from the manager (comp B is unproven); the assessor's tax figure after sale; the written insurance quote; the lender's dated estimate; inspection findings on the roof and heating; local rules on renting the property out.

That last line is the part an assistant cannot complete. It can list the checks, but the rent, the tax bill, the quote, the estimate and the local rules each come from a person or an office that has to be asked. After an offer is accepted, the checking moves to documents, and the tutorial on AI due diligence for a rental purchase covers leases, surveys and searches.

A year later, close the loop. Keep the assumptions table and compare each row with what the property actually did: collected rent, vacancy, repairs, tax, insurance. If your 7% repairs figure turned out to be 4% or 11%, change it for the next deal. A landlord's monthly bookkeeping, covered in the rental bookkeeping routine tutorial, produces exactly the figures you need for that comparison, and an assumptions table that gets corrected each year is worth more than any template.

Questions investors ask about AI deal analysis

Can ChatGPT calculate a mortgage payment reliably?

Not unless it shows its working or runs code. Assistants sometimes return a plausible figure that is wrong, such as an interest-only payment. Ask for the spreadsheet formula (PMT) instead of a number, type it into your own sheet, and compare the result with your lender's written estimate. The lender's figure is the one that counts.

What is a good cap rate for a rental?

There is no universal figure, because it depends on the market, the property and your financing. One useful comparison is with your loan's interest rate. If the cap rate is below the rate you pay, borrowing reduces your cash flow, as in the example here. Set your own screen first and judge each deal against it.

How many rent comps are enough?

Aim for four to six genuinely comparable rentals, favouring ones that have actually leased over ones still advertised, and recent over old. An assistant can organise and adjust them, but it should not find them. Add a letting or management quote as a tie-breaker, and drop any comp with a different bedroom count.

Should I include a management fee if I plan to self-manage?

Yes. Budget a fee anyway, because you may hire help later, sell to someone who will, or simply value your time. Run the deal both ways, with and without the line, so you can see how much of the return is really payment for your own work.

Further reads

• Property Investment AI Tools: A Stack for a Five-Property Portfolio — A tool stack once you own several rentals.
• AI to Find Investment Properties: Searches, Filters and Sanity Checks — Find listings worth this much checking in the first place.
• AI Commercial Real Estate Underwriting: A First-Pass Model — The same discipline applied to a larger commercial property.
• AI for House Flipping: Rehab Budgets, Comps and Contractor Quotes — Budget repairs and quotes if the deal needs real work.
• AI Software for Landlords: Rent, Maintenance and Reminders Compared — Rent, maintenance and reminder tools for landlords compared.
• AI for Rental Properties: Screening Enquiries, Viewings and Tenant Replies — Handle tenant enquiries once the purchase completes.
• AI Automations for Real Estate Wholesalers: Sourcing to Contracts — A step-by-step map of what to automate in a wholesaling business, where a person must stay in the loop, and the consent and disclosure rules.
• AI Property Development Feasibility: A First-Pass Appraisal — A worked first-pass appraisal for a small developer: inputs, GDV, costs, sensitivity and planning risks, with AI as challenger and a spreadsheet as calculator.
• AI Rent Roll Analysis for Multifamily: Reconcile to the Bank — Clean a rent roll, summarise vacancy and delinquency, and tie billed rent to bank deposits with AI, using four prompts and a worked 48-unit example.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Microsoft's PMT and function documentation checked while building the examples, and standard definitions of NOI, cap rate, DSCR and cash-on-cash return. Every property, rent, rate and cost in the examples is invented for illustration.