Tenant fraud has existed as long as rental markets have. Someone presents false income documentation, a fabricated reference, a borrowed identity. The operator approves the application, the tenant moves in, and the problems begin. Eviction timelines run three to six months. By the time the unit is recovered, the damage and lost rent often exceeds $30,000.
That version of fraud still exists. What has changed is the capability available to the person running it. AI-generated synthetic identities can now produce a complete, internally consistent application package: employment verification letters with matching phone numbers answered by AI voice agents, pay stubs with accurate formatting and plausible numbers, credit files assembled from real but thin-file individuals whose data was acquired cheaply. Standard screening software, built to catch human fraud, cannot distinguish this from a legitimate application. It was not designed to.
Tenant screening was built to verify that the documents you were shown match the data in the bureaus. It was not built for an environment where both the documents and a version of the bureau data can be fabricated simultaneously, at scale, by an automated process that costs less per application than the screening fee itself.
The numbers behind the exposure
93.3% of large residential operators report experiencing tenant fraud in 2025, according to the National Multifamily Housing Council. The per-incident cost ranges from $25,000 to $50,000 when you account for lost rent, legal fees, unit damage, and retenanting costs. For a mid-sized operator running 500 units with a 10% annual turnover, even a 5% fraud rate in applications represents a material exposure.
Synthetic identity fraud is now the fastest-growing category within that number. Unlike traditional identity theft, which involves stealing a real person's complete identity, synthetic fraud constructs a new person from fragments: a real social insurance number from one source, fabricated name and address history, AI-generated employment and rental history designed to pass screening models trained on real data. The constructed identity has no prior fraud flags because it has never existed before.
Of large residential operators report experiencing tenant fraud, per NMHC 2025.
Average per-incident cost including lost rent, legal, damage, and retenanting.
Growth in synthetic identity fraud cases in rental applications since 2023, per TransUnion.
Why standard screening cannot catch it
The fundamental problem is that most screening software verifies consistency rather than authenticity. It checks whether the name on the application matches the name associated with the social insurance number in the bureau's records. It checks whether the stated income is plausible given the employment history returned by the verification vendor. It looks for prior evictions, derogatory marks, and identity flags.
All of these checks can be passed by a well-constructed synthetic identity, because the checks were designed for a world where fabricating a complete, consistent identity package was difficult and expensive. It no longer is. The marginal cost of generating a convincing synthetic application has fallen to near zero. The marginal cost of the fraud detection system that catches it has not moved commensurately.
The insurance dimension
Renters insurance conversion in the residential market sits at 2 to 5 percent among standard operator relationships. Part of that gap is product awareness. A significant part is the carrier's rational unwillingness to write policies in a data environment they cannot verify. A carrier pricing renter's insurance for a population of applicants that includes a meaningful percentage of synthetic identities is not pricing actuarial risk. It is pricing unknown risk with models built for a different threat environment.
The fraud problem and the insurance problem are the same problem. Both are symptoms of an information architecture that has not kept pace with the tools now being used against it. A carrier that could verify identity, payment history, and risk profile through a single regulated data spine would price differently. More accurately, more competitively, and with far less adverse selection exposure.
What actually addresses this
The screening vendors are aware of the problem and are working on AI-based detection tools. Some will be effective at the margins. The deeper issue is that detection-based approaches are reactive by nature. They are constantly catching up to the fabrication capability, and the fabrication capability is improving faster.
The more durable solution is to restructure what screening is checking. An operator who sees a tenant's actual payment behavior, sourced directly from banking data through a consented open-banking integration, is not relying on a document they were shown. They are looking at the underlying financial reality. An identity verified through a regulated financial account, rather than through documents that can be fabricated, carries a fundamentally different evidentiary standard.
This is not a product feature. It requires a regulated entity with the bank sponsorship and bureau relationships to access that data and present it through a compliant consent flow. The operators who will have meaningful protection against synthetic identity fraud in three years are the ones connecting to infrastructure built at that level today, not the ones adding another screening layer on top of the existing stack.