Falling Mortgage Fraud Rates Hide an AI Problem: 5 Checks That Catch It
The fraud most likely to slip through a broker's office in 2026 arrives as a PDF that looks exactly right. Same bank logo, same font, same spacing as the real thing. Equifax Canada has noted that mortgage fraud detection rates appeared to dip slightly in 2024 and 2025, and its read is that much of the fraud has gone quiet and become harder to catch.
That should change how files get reviewed. Generative AI can now produce pay stubs, T4s and bank statements that mirror a real institution's layout and even its hidden file data. Squinting at a document for blurry logos stopped working. The checks that still work go around the document to whoever issued it. In Alberta, the know-your-client duty under RECA's rules makes the broker the first person who can do that, and the person whose name sits on the file if nobody does.
1. Verify income where it starts. Call the employer using a number you found yourself, never the one printed on the pay stub. Better still, use a verification service that confirms employment and pay directly with the payroll provider. People skip this because the stub looks clean and the deal is moving fast, which is exactly the moment a fake is built for. Edmonton's steady stream of out-of-province investors means busy weeks and rushed paperwork.
2. Take documents straight from the source. Have the client download statements from their banking portal while you watch, or use a secure third-party link to the bank. Some clients find this intrusive, so explain it once: it protects them too, and it is faster than a second round of questions from the lender. A file the client emails you can be edited. A file pulled from the bank in front of you can't.
3. Line up the whole story. The employer on the T4 should match the deposits in the bank statements, the employment on the credit bureau, and the address on the ID. AI makes each document convincing on its own, but keeping a dozen documents consistent with each other is harder. Mismatches between documents are where fakes show up.
4. Slow down on a thin file with a big ask. Fraudsters now build "synthetic identities," pairing a real Social Insurance Number with an invented name and address, then growing a credit history aimed at one large mortgage. Look at how old the oldest trade line is, whether the credit history fits the person's stated age and career, and whether a short, tidy history is suddenly asking for a large approval.
5. Have the honest conversation before the application exists. Equifax reports first-party fraud is rising: real people, using their real names, padding their own numbers. With the stress test still qualifying borrowers at the contract rate plus 200 basis points or 5.25%, whichever is higher, plenty of otherwise honest buyers are tempted to round their income up by a few thousand. Tell them plainly that this is fraud under the Criminal Code, that it can lead to a mortgage being called (repayment demanded in full) and to being shut out by major lenders. Then help them find a real path, whether that's a different lender, a co-signer or a few more months of saving.
Keep one caution in mind as automated screening spreads. Not every odd file is a fake. A commission-paid oil and gas worker or a self-employed contractor in Leduc can have income that looks strange to a software flag and is completely real. Honest mistakes on applications happen, and catching those before they reach the lender is part of the same job. The checks above work in both directions: they expose the fake, and they give a legitimate borrower a confirmed record that answers the lender's questions before they're asked.
Before you send the next file, ask one question about every income document in it: did this come to me from the person, or from the place that issued it?
The fraud most likely to slip through a broker's office in 2026 arrives as a PDF that looks exactly right. Same bank logo, same font, same spacing as the real thing. Equifax Canada has noted that mortgage fraud detection rates appeared to dip slightly in 2024 and 2025, and its read is that much of the fraud has gone quiet and become harder to catch.
That should change how files get reviewed. Generative AI can now produce pay stubs, T4s and bank statements that mirror a real institution's layout and even its hidden file data. Squinting at a document for blurry logos stopped working. The checks that still work go around the document to whoever issued it. In Alberta, the know-your-client duty under RECA's rules makes the broker the first person who can do that, and the person whose name sits on the file if nobody does.
1. Verify income where it starts. Call the employer using a number you found yourself, never the one printed on the pay stub. Better still, use a verification service that confirms employment and pay directly with the payroll provider. People skip this because the stub looks clean and the deal is moving fast, which is exactly the moment a fake is built for. Edmonton's steady stream of out-of-province investors means busy weeks and rushed paperwork.
2. Take documents straight from the source. Have the client download statements from their banking portal while you watch, or use a secure third-party link to the bank. Some clients find this intrusive, so explain it once: it protects them too, and it is faster than a second round of questions from the lender. A file the client emails you can be edited. A file pulled from the bank in front of you can't.
3. Line up the whole story. The employer on the T4 should match the deposits in the bank statements, the employment on the credit bureau, and the address on the ID. AI makes each document convincing on its own, but keeping a dozen documents consistent with each other is harder. Mismatches between documents are where fakes show up.
4. Slow down on a thin file with a big ask. Fraudsters now build "synthetic identities," pairing a real Social Insurance Number with an invented name and address, then growing a credit history aimed at one large mortgage. Look at how old the oldest trade line is, whether the credit history fits the person's stated age and career, and whether a short, tidy history is suddenly asking for a large approval.
5. Have the honest conversation before the application exists. Equifax reports first-party fraud is rising: real people, using their real names, padding their own numbers. With the stress test still qualifying borrowers at the contract rate plus 200 basis points or 5.25%, whichever is higher, plenty of otherwise honest buyers are tempted to round their income up by a few thousand. Tell them plainly that this is fraud under the Criminal Code, that it can lead to a mortgage being called (repayment demanded in full) and to being shut out by major lenders. Then help them find a real path, whether that's a different lender, a co-signer or a few more months of saving.
Keep one caution in mind as automated screening spreads. Not every odd file is a fake. A commission-paid oil and gas worker or a self-employed contractor in Leduc can have income that looks strange to a software flag and is completely real. Honest mistakes on applications happen, and catching those before they reach the lender is part of the same job. The checks above work in both directions: they expose the fake, and they give a legitimate borrower a confirmed record that answers the lender's questions before they're asked.
Before you send the next file, ask one question about every income document in it: did this come to me from the person, or from the place that issued it?
Read Next
Five Money Checks Canadian Snowbirds Should Finish Before Heading South
Canadian Bank Security Assumes a Human Is Logging In. AI Agents Break That
Before Storm Winds Arrive, Turn 5 Yard Hazards Into Protective Buffers
A 25 Basis Point Hike Doesn't Erase Variable's Long-Run Edge Over Fixed