Every screening service now sells the same promise: submit the application, get a score or a green check, approve or deny in one click. For a landlord with a handful of doors and a day job, that is genuinely tempting. Here is the part the sales page skips: when the algorithm gets it wrong, the liability lands on you, not on the software.
The short version
Federal regulators have been explicit that a housing provider is responsible for screening decisions even when the screening is outsourced to a company using algorithms or AI. Disclaimers in the vendor contract do not shield you. If a tool denies an applicant and you cannot explain why, you are carrying risk you did not price. The fix is not avoiding these tools. It is choosing ones you can explain, and keeping the paperwork rights the law gives applicants.
Who is liable when an AI screening tool gets it wrong?
You are, and usually the vendor too. HUD’s guidance on tenant screening says housing providers are responsible for their rental decisions even when they have largely outsourced screening to a third party, and that screening companies can be liable as well, even when the landlord formally makes the final call. The screening report is an input. The decision, legally, stays yours.
That principle has teeth in court. A screening company that sold an algorithmic score was sued under the Fair Housing Act over how its scores treated applicants with housing vouchers, and the case ended in a settlement that included changing how the product works. The claims did not require anyone to prove the algorithm intended to discriminate.
One thing worth saying plainly: parts of the federal guidance landscape have shifted since 2024, and some HUD documents have been withdrawn. The Fair Housing Act itself has not changed, and private lawsuits run on the statute, not on guidance memos. Do not let a vendor tell you the rules went away.
The screening practices that create the most risk
The recurring problems are specific and avoidable:
- Unlimited criminal lookbacks. Screens that treat a decades-old record the same as a recent one, or count arrests that never became convictions, are exactly what regulators have flagged. Look for tools that let you set the nature, severity, and age of records that matter.
- Eviction records without outcomes. Filings where the tenant won, settled, or the case was dropped say very little, and eviction databases are notoriously error-ridden. A tool that counts raw filings is counting noise.
- Name-only matching. Cheap data matching attaches strangers’ records to your applicant, and the error rate is worst for common surnames. This exact practice is behind some of the largest federal settlements ever paid by screening companies.
- A score you cannot explain. If the tool returns “denied, 4 out of 10” with no reasons, you cannot write a lawful denial explanation, and you cannot defend the decision later.
The paperwork the law requires from you
Tenant screening reports are consumer reports under the Fair Credit Reporting Act, and that triggers duties for the landlord, not just the bureau. If you deny an applicant, require a co-signer, or ask for a bigger deposit based even partly on a screening report, the applicant is owed an adverse action notice: which company supplied the report, that the company did not make the decision, and that they have the right to a free copy and to dispute errors. Screening companies have paid multimillion-dollar federal settlements over sloppy data, and landlords who skip the notice hand a tenant’s lawyer the easiest claim on the menu.
Texas adds a wrinkle worth knowing: state law expects landlords to make their tenant selection criteria available in writing at application time, and rejecting an applicant without having done so can obligate you to refund the application fee. Oklahoma has almost no state overlay on screening, which means the federal rules are doing all the work there.
A word about the small-landlord exemption
Many small landlords believe the Fair Housing Act does not apply to them. The actual exemption is narrow: it covers an owner living in their own building of four or fewer units, or an owner selling or renting up to three single-family houses without a broker. It never exempts discriminatory advertising, and race discrimination in property transactions is barred by a separate statute with no exemptions at all. An investor with a few scattered rentals, screening through a commercial tool and advertising online, is fully covered by the Act. Plan accordingly.
Frequently asked questions
Can I still use an AI screening service?
Yes, and a good one beats gut feel, which has its own bias problem. Choose a tool that lets you set your own criteria, explains every denial in plain language, updates its data, and gives applicants a correction path. Then apply the same written criteria to everyone.
What about AI rent-pricing tools?
Be careful with any tool that sets your rent using competitors’ private data. Federal antitrust enforcers took action against exactly that model, and several states and cities now ban those tools outright. Texas and Oklahoma have no such ban today, but a small landlord pricing off public comps loses nothing and avoids the tail risk entirely.
Does this apply to my one rental house?
If you advertise publicly and use a screening service, effectively yes. The exemptions are far narrower than the folklore version, and the advertising rules apply to everyone.
Where we land on it
Good screening is written criteria, applied evenly, with tools you can explain. That protects you and it is fairer to applicants, which is not a coincidence. If you are building a rental portfolio in DFW or Oklahoma and want deal flow that pencils, that is our world, and our note on what AI can and cannot underwrite on a deal covers the acquisition side. Nothing here is legal advice; for your specific criteria and notices, have a landlord-tenant attorney look at what you are actually sending.
Enjoying these? We publish straight talk like this twice a week. Follow us on Linktree to keep up with everything we are working on, from new tools to new markets.