The 28/36 rule: what AI gets wrong about mortgage affordability

AI chatbots often cite the 28/36 rule when they answer questions about how much house you can afford.

The 28/36 rule is a useful starting point for understanding mortgage eligibility, but it's not the final answer.

In this article we'll explain what the 28/36 rule actually measures and identify gaps in this number that AI doesn't always find.

Fortunately, there's a better way to see what payment you can afford, and it takes only a few minutes.

...in as little as 3 minutes — no credit impact

What the 28/36 rule actually says and what it doesn't

The 28/36 rule is a general lending guideline, not a hard-and-fast rule. 28/36 says two things:

• your total monthly housing costs — principal, interest, property taxes, and homeowners insurance, sometimes called PITI — should be no more than 28% of your gross monthly income.
• your total monthly debt payments, including that housing cost plus car loans, student loans, and credit card minimums, should be no more than 36% of gross monthly income.

Sometimes, AI underestimates a borrower's ability to repay a loan. It's true that a borrower whose loan fits the 28/36 framework will likely be able to afford the loan's monthly payments going forward. But loans that exceed this threshold may still be approved, depending on the borrower's unique situation.

Or, AI might overestimate affordability by not including all housing costs in the mortgage payment. Monthly payments include more than repaying the loan. They also include homeowners insurance, property taxes, and HOA fees.

Let's take a closer look at these and a few other holes in AI reasoning:

Five things AI gets wrong when it applies the 28/36 rule

When a buyer asks an AI "how much house can I afford on $85,000?", what comes back is a reasonably confident number derived from a reasonably simple calculation. Here is where that calculation breaks down.

1. It calculates from take-home pay, not gross income

The 28% figure is a percentage of your gross monthly income. That's your salary before taxes, FICA, health insurance premiums, and any retirement contributions come out. That is how lenders calculate DTI, and there is a reason for it: gross income is a stable, documented figure.

A buyer earning $85,000 a year has a gross monthly income of about $7,083. Twenty-eight percent of that is $1,983, which is the number a lender would likely use to cap a monthly house payment at 28/36.

After payroll deductions, monthly pay from an $85,000 annual salary might be as low as $5,500, and AI might use this number instead. 28% of $5,500 allows a house payment of $1,540, much lower than $1,983.

In this case, AI underestimates buying power.

Example is for illustrative purposes only. Rates, payments, and total interest will vary based on credit profile, loan terms, and market conditions.

2. It uses a generic national property tax rate

Property taxes are included a mortgage payment, so they count against your 28% ceiling. The problem is that AI tools typically apply a national average property tax rate, usually around 1.0–1.1% of home value annually. That average ignores wide regional variations.

Property taxes by state range from under 0.3% in Hawaii to over 2.1% in Illinois and New Jersey. On a $400,000 home, the difference between 0.5% and 2.1% is $667 versus $2,800 per year, or roughly $55 versus $233 per month. A $178/month gap in property taxes alone consumes nearly 2.5% of gross income on an $85,000 salary, and AI does not know which county you are buying in.

In high-tax metros, this single variable can shift a buyer's true affordability by $30,000–$50,000 in purchase price. The AI number may not account for it.

Example is for illustrative purposes only. Rates, payments, and total interest will vary based on credit profile, loan terms, and market conditions.

3. It uses a national average for homeowners insurance

AI tools typically apply a generic homeowners insurance figure, often in the range of $125–$175 per month. That figure is broadly accurate for low-risk areas in stable insurance markets.

It is not accurate for Florida, Texas, Louisiana, parts of California, or growing stretches of the Gulf Coast and Midwest where severe weather risk has fundamentally repriced coverage. Annual premiums in high-risk Florida counties now regularly exceed $4,000–$8,000. In coastal Texas, $3,500–$5,000 is common. Recent industry data shows that insurance costs cause a meaningful share of mortgage transactions to fall through in high-risk states.

On a monthly basis, $500 in insurance on an $85,000 salary consumes 7% of gross income — nearly a quarter of the 28% housing budget, before any principal and interest has been calculated. An AI tool applying $150/month in insurance is potentially off by $300–$350 per month in high-risk states, which represents tens of thousands of dollars in purchasing power it does not know is already spent.

Before making an offer in any region with known climate or insurance risk, get an actual insurance quote. What's included in your monthly mortgage payment depends on the real insurance cost, not a national average.

4. It doesn't reflect what lenders actually approve

The 28/36 rule and lender approval are two different things. A buyer whose numbers exceed 28/36 may still be approved without difficulty depending on their credit score, cash reserves, loan program, and rate. A buyer at exactly 28/36 with a low credit score or no reserves may face more scrutiny.

For buyers with strong credit scores (740+) and modest existing debt, a lender may approve a loan well above what 28/36 suggests. For buyers with credit in the 600s and student loan debt, the lender may approve less. Neither outcome is what the 28/36 rule predicts, because the rule does not know your credit score, your rate, your loan program, or your down payment.

5. It ignores the power of shopping around for better rates

An AI estimate will likely apply today's average mortgage rates to the loan scenario. In reality, few lenders get assigned the average rate.

Individual rates depend on the borrower's credit score, debt load, loan type, and other unique details. Your mortgage rate could be higher or lower than today's average, and even a small variation in interest rate can affect the dollar amount of a mortgage payment in big ways.

To see your unique rate, start with a pre-approval.

...in as little as 3 minutes – no credit impact

What the rule is actually useful for

This doesn't mean the 28/36 rule is broken. It means it is being asked to do something it was not designed for.

The rule does one thing well: it gives you a conservative budget ceiling before you know your rate, your credit-score-based pricing, or your specific property's tax and insurance costs. Use it to disqualify homes that are obviously too expensive. Use it to set a range before you start searching. Use it as a gut check when a lender says you qualify for more than feels comfortable. But don't use it as a final destination for your home buying journey.

How to get an accurate affordability number

Four steps replace the AI estimate with numbers you can actually use.

Step 1: Get pre-approved. A pre-approval uses your actual income documentation, credit score, existing debts, and current rate to calculate a real loan amount. It's the difference between pre-qualified vs. pre-approved — pre-approval is a verified underwriting assessment, not a ratio. Knowing how to get pre-approved for a mortgage is the most direct path to replacing an estimate with a real number.

Step 2: Look up the property tax rate for the specific county. County assessor websites publish current tax rates. For any home you are seriously considering, calculate the actual annual tax bill and divide by 12 to get your real monthly obligation.

Step 3: Get an insurance quote before you make an offer. In stable markets, a quick online quote is sufficient for budgeting. In Florida, Texas, Louisiana, or California, you may need to contact several carriers. The number you get will be far more accurate than any national average.

Step 4: Add HOA fees and a maintenance reserve. If the property has an HOA, its monthly fee is a fixed housing cost. A common starting point for maintenance budgeting is 1% of the home's value annually, hough newer homes typically run lower, older homes higher.

Together, these four steps produce a monthly housing cost figure that reflects your real situation and can help you decide whether now is a good time to buy a house.

Frequently asked questions

I asked an AI how much house I can afford and it told me $450,000. Should I trust that number?

Use it as a starting range, not a plan. AI affordability tools apply the 28/36 rule to your gross income, which typically overstates what buyers can sustain in practice because it does not account for your actual take-home pay, your specific property tax rate, current insurance costs in your area, or HOA and maintenance expenses. A pre-approval will give you a lender-verified number based on your real financial profile.

Is the 28/36 rule still accurate in 2026 when housing costs are so high?

The rule is still a reasonable conservative guideline. The problem is that housing costs in 2026 include property tax and insurance figures that have risen sharply in many markets, and the rule's math does not update automatically to reflect your specific location. In high-insurance states and high-tax metros, buyers who stay within 28% of gross income on PITI may still be stretched if their underlying tax and insurance inputs are based on outdated national averages.

I make $85,000 a year. Does the 28/36 rule actually tell me what I can afford, or is it just a rough estimate?

It is a rough estimate, useful for setting a range, not for evaluating a specific home. On $85,000 gross, 28% gives you approximately $1,983/month for housing. But your take-home pay, local property taxes, insurance costs, and any HOA fees will all shift that number. Getting pre-approved and looking up actual property tax and insurance figures for homes you are considering will give you a number you can rely on.

Does the 28/36 rule account for property taxes and homeowners insurance, or just the mortgage payment?

It does include property taxes and homeowners insurance. These are part of PITI, which is what the 28% front-end ratio is measured against. The issue is that AI tools typically apply a generic national average for both, rather than your actual local tax rate and a real insurance quote. In high-tax or high-insurance markets, the difference between the average and the reality can be several hundred dollars per month.

I live in Florida where insurance is really expensive. Does the 28/36 rule still apply to my situation?

The rule still applies as a framework, but accurate insurance costs are especially critical to plug in before you use it. Annual homeowners insurance premiums in many Florida counties now significantly exceed national averages — in some cases by $300–$400 per month or more. If you use a national average insurance figure in the 28% calculation, your apparent affordability ceiling will be higher than your real one. Get an actual insurance quote for the property you are considering before finalizing any budget.

What's the difference between what an AI says I can afford and what a mortgage lender will actually approve?

These are different calculations answering different questions. AI uses the 28/36 rule to suggest what is financially comfortable, a personal finance guideline. Lenders use your actual credit score, documented income, debt obligations, loan program, and rate to determine maximum loan amount, typically allowing up to 43–45% back-end DTI on conventional loans, and higher on FHA. Buyers with strong credit and low debt may be approved for significantly more than 28/36 suggests. Buyers with credit challenges may be approved for less.

Should I use the 28/36 rule or get pre-approved to figure out my home budget?

Use the 28/36 rule early to set a rough budget range and disqualify homes that are obviously out of reach. Then get pre-approved before you start seriously searching. Pre-approval replaces the estimate with a lender-verified number based on your real income, credit score, and debt profile, and it confirms to sellers that you are a qualified buyer.

Why does my AI affordability estimate seem higher than what my lender is actually approving me for?

A few likely reasons: AI calculates from gross income, and your take-home pay may be meaningfully lower after taxes and benefits. Your credit score affects both the rate you are offered and the DTI flexibility lenders will give you. A buyer in the 600s gets less flexibility than a buyer at 760. And if you carry significant student loans, car payments, or credit card debt, the back-end DTI calculation tightens faster than the front-end 28% figure alone suggests.

The bottom line

The 28/36 rule is a valid framework and a reasonable place to start thinking about affordability. What it is not is a reliable answer to "how much house can I afford?" And that is exactly the question AI tools are applying it to. The five gaps above are not edge cases. They affect most buyers in most markets, and they all point in the same direction: the AI number is typically too optimistic.

Getting pre-approved takes minutes, does not affect your credit score, and replaces a ratio estimate with a verified, lender-backed number you can actually use.

...in as little as 3 minutes — no credit impact