AI Can Make a Flawless Fake Pay Stub in 60 Seconds. Here’s What Still Catches It.

For years, a fake pay stub gave itself away: mismatched fonts, math that didn’t add up, a blurry logo. Those tells are gone. Free AI tools now generate a pay stub in under a minute with perfect arithmetic, clean formatting that mirrors ADP or Gusto, a consistent pay schedule, and correct tax-withholding math. It will look better than a real stub — and if you’re screening by eye, you will approve it.

This isn’t a someday problem. AI-generated document fraud surged in 2025 — one industry report put the jump at roughly 5x (economy-wide, not rental-specific) — and by industry surveys more than 90% of property managers have seen a fraudulent application in the past year. The document you’re squinting at was very possibly made by a machine.

Why “spot the fake” no longer works

Every classic tip for spotting a fake pay stub — check the math, check the alignment, check the round numbers — assumes a sloppy human forger. AI is not sloppy. It nails the exact things you were taught to look for, because internal correctness is trivial for a model to produce. So a pay stub that looks perfect tells you nothing anymore. The forgers upgraded; the checklist didn’t.

Worse, the same tools spin up an entire fake employer to stand behind the stub — a company name, matching stub, sometimes a throwaway website and a phone number a friend answers. And if your plan is to “just call the employer,” remember you must never call the number on the stub: with an invented employer, that number rings straight back to the applicant.

What still catches an AI fake

Here’s the good news, and it’s the whole point: AI can perfect the document, but it cannot invent a real company in the real world. The defense has shifted from inspecting the paper to verifying what’s behind it — and that part AI can’t fake:

• Is the employer a registered business on file with the state?
• Does it have a genuine commercial address — one that exists on a map, not a house or a mailbox store?
• Does its phone and online footprint check out independently of what the applicant gave you?
• Do the tax lines follow the fixed statutory rates a real payroll system would produce?

A flawless AI stub attached to a company that evaporates under this scrutiny is the fraud — and it’s now the *most common* fraud, precisely because the document itself is beyond reproach. Confirming the employer is a real company is the check that survives the AI era.

Why this is where most tools — and most landlords — fall short

This is worth shouting, because it’s the difference that matters. Most fraud tools, including the enterprise leader behind many screening platforms, only detect whether a document was altered. An AI stub was never altered — it was born clean — so it passes. And the bank-connect verification services only work if the applicant *cooperates* and links a real account, which a fraudster simply won’t do, and which a fake employer has no account to link. The one fraud that’s exploding is the exact one these tools are blind to: a perfect document for a company that doesn’t exist.

That gap is what ProofSweep is built for. It re-runs the pay math, then independently verifies the employer — business registry, the address on a real map and street view, the phone, the web footprint, the statutory tax rates — without needing the applicant to connect anything. You get a plain-English report in 24 hours for $19.99. Weigh that against the cost of being wrong: a single bad tenant runs a small landlord $7,000 to $15,000 once you count unpaid rent, the eviction, and repairs. Nineteen dollars against fifteen thousand is not a close call.

AI made the fake document a solved problem for fraudsters. It also made document inspection worthless. The only thing left that a scammer can’t manufacture is a real employer — so that’s the thing to check. See everything a report verifies.