A 74-year-old LA landlord named Michael Renkow lost $90K and seven months of his life to a guy who used a fake name with his real photo. The case got solved because the fraudster got sloppy. Most don’t.

The story ran in Bisnow this weekend. Worth reading. Link in below.

The industry framing: AI is making fake pay stubs and IDs too good to spot. We need better screening tech. Trade groups cite a 93% fraud-experience rate, $15K average cost per case, and 24% of evictions linked to fraudulent applications.

I read the article and asked a different question. Who benefits from this framing?

Pull on the 24% thread and it unravels. The source is an NMHC pulse survey of 75 self-selected member respondents. Self-reported. No audit. No verification. Average bad debt write-off was $4.2 million. Median was $800K. When the average is five times the median, a few REIT-scale portfolios are dragging the math, and the headline isn’t describing the typical owner.

The 24% is post-hoc attribution. A tenant stops paying. The property manager decides, with no documentation requirement, that the original application must have been fraudulent. Some of those cases are real fraud. Others are job loss, medical emergencies, or weak underwriting that approved a marginal applicant under occupancy pressure.

Notice who fills out the survey. The same property managers whose performance looks better if delinquencies get labeled fraud instead of bad approvals.

That’s the first incentive problem. The bigger one sits one level up.

A leasing agent earns commission on a signed lease. A property manager earns a bonus on occupancy. A regional VP earns on portfolio NOI. Where does fraud detection sit in any of those compensation plans?

Nowhere.

RealPage’s own 2024 fraud survey of 402 property managers across Atlanta, Houston, LA, Miami, and Seattle confirmed it from inside the industry. 97% said preventing fraud is a top priority. Only 17% had a portfolio-wide fraud prevention program. Only 22% had any metric tracking fraud at all. 43% of site staff had no financial incentive to prevent it. 73% of fraud was discovered after move-in.

Almost everyone says it matters. Almost nobody measures it. Half admit they don’t pay anyone to catch it.

That’s not a tech failure. That’s an incentive plan failure.

The fix isn’t another screening subscription. It’s smaller and less profitable to vendors than a SaaS contract.

Run every applicant ID through a DMV database lookup. Cost is a few dollars. MRI’s own published study of 200 AI-generated fake IDs found that OCR alone caught 24%. The same vendor’s database lookup against DMV and national records caught 99.6%. The tech works. Most operators bought the cheap tier and stopped.

Verify income from the bank, not from a PDF. Plaid integrations exist at every major screening platform. Document uploads are the back door fraudsters use. Close it.

Do the cheap procedural cross-checks nobody is paid to do. Look up the employer phone number on your own, not from the application. Check county property records to confirm a previous landlord actually owned the property they claim. Reverse image search the photo ID. Run the LLC on the pay stub against the state corporations database. None of this requires AI. It requires twenty minutes and a person whose comp plan rewards finding the discrepancy.

Centralize the override authority. Strip site staff of the power to approve marginal applications. Put the override in a regional role with no occupancy quota and a paper trail that gets reviewed quarterly.

Build fraud metrics into the comp plan. Track which evictions trace back to which screening exceptions. Pay leasing agents and property managers something on a clean book of business, not just on lease velocity.

On the back end: We never accept cash at our properties. Money orders from USPS are preferred for cash-equivalent payments because verification is a phone call to a federal agency. ACH from a verified bank is another solid path. Cashier’s checks held until cleared. The principle is simple. Funds settle before keys change hands.

The deeper issue is the law. Application fraud is a misdemeanor or nothing in most states. Eviction takes months. Renkow’s case isn’t a tech failure, it’s a policy one. The trade group’s lobbying energy currently goes toward preempting AI screening regulation. It could go toward expedited eviction for fraud-based applications and felony charges for document forgery. That fight would actually move the number.

Fraud is real. The distribution isn’t even. Class A urban with employer-verified W-2 income sees less of it than Class C garden-style with hospitality workers and gig economy income. Single-family and small landlords see more still, because they have the thinnest verification stack and the deepest principal-agent problem of all (the owner is the screener, the screener is exhausted, and there’s no team to enforce a policy).

The fraudsters are getting better. The industry has decided that means owners need more software. The data says owners need a different compensation plan.

Watch where the attention goes. And who profits when you give it.

https://www.bisnow.com/national/news/multifamily/ai-rent-fraud-apartment-renters-landlords-pay-price-134391?rt=109026