AI tenant screening that gathers the evidence — and leaves the decision to you.

What "AI tenant screening" actually means

The phrase covers two very different things, and the difference matters legally as well as practically.

The first is AI that collects and verifies: placing a reference call to a prior landlord and transcribing what they say, connecting to a payroll provider to confirm income, reading an uploaded document and noticing that its metadata does not match its contents. This is automation applied to work a landlord would otherwise do by hand, badly, at 9pm.

The second is AI that judges: a model that consumes an applicant's file and emits a score, a risk band, a colour, or an "approve/decline" recommendation. That is a fundamentally different product, and it is the one this platform does not build.

RentalApplication.ai is a reseller of consumer reports under FCRA § 603(u). The credit, criminal and housing-records data comes from consumer reporting agencies; we assemble and deliver it. A system that then produced its own judgement about the applicant would be doing something else entirely — and the landlord, not the software, is the party the Fair Credit Reporting Act makes responsible for the tenancy decision.

Where the AI does the work

AI reference calls to prior landlords

Prior-landlord references are the highest-signal, least-collected part of a rental application. Most landlords skip them because reaching another landlord by phone takes three attempts across two days.

The platform places the call for you, asks the questions you configured, and returns the prior landlord's answers. The response is transmitted as given: it is the prior landlord's statement, not a summary written by a model, and recorded calls are stored as transcripts rather than AI-condensed notes. If the reference says something unhelpful or contradictory, you see it in their words.

Income verification that survives an AI-generated paystub

Document forgery stopped being a craft skill some time in 2024. A convincing paystub, bank statement or offer letter is now a short prompt away, which makes the traditional "upload two recent paystubs" step close to worthless as evidence.

The durable answer is not better forgery detection — it is not asking for documents at all. Connecting directly to the applicant's payroll provider or bank returns income data from the source, where there is nothing for a generative model to fabricate. See income verification for rentals for how the connection works and what it returns.

Application-fraud signals

Alongside verification, the application itself carries signals: an employer that does not exist, a phone number that routes to the applicant rather than the company, an address history that contradicts the SSN trace. These are surfaced as observations with their source attached, so you can check them yourself. They are not aggregated into a fraud score. Our guides on detecting rental application fraud and application red flags cover what to look for.

What the AI is not allowed to do

• No applicant score or risk rating. There is no proprietary "RentalApplication score". The only score in your report is the one the consumer reporting agency furnished — TransUnion ResidentScore, or FICO if you choose that tier.
• No approve/decline recommendation. The platform will never tell you who to rent to, and no screen in the product implies a verdict.
• No ranking of applicants against each other. Reports are delivered per applicant; the software does not sort your pipeline by desirability.
• No inference about protected characteristics. Nothing in the product infers or proxies race, national origin, familial status, disability, religion, sex or age — and no automated message is sent to an applicant without a fair-housing screen having run on it first.
• No training on your applicants. Applicant data is not used to train consumer-facing or general-purpose models.

The practical consequence: because the decision is yours, so is the adverse-action obligation if you decline based on the report. The platform generates the compliant notice; it does not generate the decision behind it.

Why "the AI decides" is a bad deal for a landlord

An automated tenancy verdict sounds like it removes work and risk. It adds both.

Under the Fair Housing Act a landlord is responsible for the discriminatory effect of a screening policy, not merely its intent — and "the software recommended it" is not a defence. A model that has learned from historical tenancy outcomes has learned from historical tenancy decisions, which is precisely where disparate impact hides. Regulators have been explicit that automated screening tools are in scope.

A landlord who can point to the criteria they applied, consistently, to every applicant is in a far stronger position than one who deferred to a score they cannot explain. Written, objective screening criteria that you set and apply the same way every time remain the single best protection available — see screening laws by state for the local variations.

What comes back in the report

• Credit — TransUnion report and ResidentScore (FICO available), pulled as a soft inquiry so the applicant's score is untouched.
• Criminal records — state and county records, national database matches, sex-offender registry.
• Housing records — public housing-court filings the applicant was party to.
• Income — payroll-provider or bank-link verification rather than uploaded documents.
• Prior-landlord references — AI-placed calls, answers returned verbatim.
• Identity — SSN trace against historical address and name records on tiers that include it.

Pricing runs $24.99–$49.99 per screening depending on the bundle, with no subscription. The full breakdown is on the pricing page, and a sample report shows the output.