AI in Leasing
AI Tenant Screening vs. AI-Generated Fraud: What Multifamily Operators Are Doing in 2026
Short answer: Rental application fraud is rising as AI makes fake pay stubs and IDs cheap. Operators are layering AI document checks with human review and governance. NAA reports that an often-quoted NMHC survey found over 93% of respondents experienced some type of fraudulent activity in the last 12 months. Sources: [1] [2] [3] [4] [5]
How big is the rental application fraud problem?
NAA reports that an often-quoted NMHC survey found over 93% of respondents experienced some type of fraudulent activity in the last 12 months. Greystar has reported that up to half of applications are fraudulent in select buildings in hot spots including Washington, D.C., south Florida and Atlanta. Funnel, a leasing software vendor, says 70% of operators report fraud increasing year over year and that one in five documents screened on its platform carries medium to high fraud risk.
Why can't AI alone catch AI-generated fraud?
Research from NAA and MRI Software says operators cannot rely solely on AI to catch AI-generated fraud because unchecked automation can produce biased or inaccurate results. Third-party tools that analyze pay stubs, bank statements and IDs for metadata inconsistencies and template mismatches remain crucial, but they are being paired with human verification.
What does a layered screening process look like?
NAA describes operators creating centralized screening hubs, standardizing lease forms, implementing earlier nonpayment protocols, training onsite teams to spot manipulated documents, and in some cases in-person applicant interviews. Funnel reports that at Windsor/GID about 93% of ID verification attempts cleared with no human intervention while the remaining 7% were routed to trained specialists (vendor-reported).
What governance should AI screening tools have?
NMHC's white paper with Ballard Spahr recommends housing providers establish internal AI compliance and governance plans. RETTC's framework calls for compliance with the Fair Housing Act, FCRA and ADA, clear explanations of decisions to applicants, outcome reviews with escalation and appeal channels, and due diligence on third-party AI vendors.
Frequently asked questions
How common is rental application fraud?
NAA reports that an often-quoted NMHC survey found over 93% of respondents experienced some type of fraudulent activity in the last 12 months. Greystar has reported that up to half of applications are fraudulent in select buildings in hot spots including Washington, D.C., south Florida and Atlanta. Funnel, a leasing software vendor, says 70% of operators report fraud increasing year over year and that one in five documents screened on its platform carries medium to high fraud risk.
Can AI screening tools catch AI-generated fake documents on their own?
Research from NAA and MRI Software says operators cannot rely solely on AI to catch AI-generated fraud because unchecked automation can produce biased or inaccurate results. Third-party tools that analyze pay stubs, bank statements and IDs for metadata inconsistencies and template mismatches remain crucial, but they are being paired with human verification.
What does a layered fraud-screening process look like?
NAA describes operators creating centralized screening hubs, standardizing lease forms, implementing earlier nonpayment protocols, training onsite teams to spot manipulated documents, and in some cases in-person applicant interviews. Funnel reports that at Windsor/GID about 93% of ID verification attempts cleared with no human intervention while the remaining 7% were routed to trained specialists (vendor-reported).
What governance should we require from an AI screening vendor?
NMHC's white paper with Ballard Spahr recommends housing providers establish internal AI compliance and governance plans. RETTC's framework calls for compliance with the Fair Housing Act, FCRA and ADA, clear explanations of decisions to applicants, outcome reviews with escalation and appeal channels, and due diligence on third-party AI vendors.
Where do these figures come from?
Every figure is attributed in the text and linked in the numbered Sources list (5 sources, with publication dates where available). Vendor figures are labeled vendor-reported.