Mortgage Fraud Rates Are Falling, and Equifax Says AI May Be Why
Consider a couple in the Cowichan Valley in the spring of 2026, shopping for a detached house somewhere between Duncan and Lake Cowichan. One works full time at a regional employer. The other left a salaried job the previous year to take contract work, and the income is real but uneven. Under the federal stress test, they have to qualify at their contract rate plus 200 basis points, or 5.25%, whichever is higher. On paper, they come up a little short.
Then someone offers to help. It might be a relative who "does paperwork," or an online service that promises to "format" self-employed income so lenders can read it. The pitch sounds reasonable: the money is genuinely being earned, the T4 from the old job is messy, and lenders only want a clean summary. The couple hands over their documents and gets back a tidy package with pay stubs, an employment letter, and a year-end summary. Every column lines up. No typos, no fuzzy scans. It looks more professional than anything they could have produced themselves.
That is the decision worth slowing down on. At the time it did not feel like fraud. It felt like presentation.
What changed on the other side of the desk
Equifax Canada's fraud and identity team has been pointing out something uncomfortable this fall. Reported mortgage fraud rates have been falling, and that may not mean there is less fraud. Generative AI now makes it cheap to produce pay stubs, bank statements and employment letters that pass a visual check. A fake that gets through is never counted. Equifax's warning is that the lower number may partly reflect how much is going undetected.
Falsified income is still the most common kind of mortgage fraud in Canada. The difference now is that it no longer looks like a doctored PDF with mismatched fonts. Investigators have started to treat a document that is too perfect as a reason to look harder. A pay stub from a small Duncan contractor that looks like a design agency built it is an anomaly in its own right.
So lenders have changed what they check. Validating a document only tells you that it exists and looks right. Verifying it means confirming that the numbers match the source: the Notice of Assessment in CRA My Account, a call to the employer, deposits that actually appear in the bank account. More lenders in British Columbia now run that cross-check automatically.
The turning point
In a case like this, the file usually comes apart at verification, not at the first review. The application gets a conditional approval. Then the lender asks for the CRA Notice of Assessment, pulled directly by the borrower. The income on it is lower than the "formatted" summary. The gap may not be large. It doesn't need to be.
The deal is declined. The bigger problem is that the file may now be flagged in the databases lenders share, and that flag follows the borrowers. Mainstream lenders may be closed to them for years. Under the Criminal Code, fraud over $5,000 carries a maximum sentence of 14 years. Most cases never get near a courtroom, but the borrowers' names are attached to the application either way. "Someone else prepared it" is a hard argument to make once their signatures are on it.
Honest borrowers feel the effects too. They now get more requests for screenshots straight from CRA My Account, more calls to employers, and slower conditions. That friction is the cost of an arms race in which AI tools produce the fakes and AI tools are also the most practical way to catch them. When enough fraud gets through, lenders build that risk into their pricing, and every borrower pays for it.
A broker who knows the local market still catches things software misses, like a claimed salary that doesn't fit the role in this region. But the more useful protection comes earlier, before anyone hands their documents to a stranger.
The couple in this scenario had a legitimate path the whole time. It involved two years of tax returns, an honest debt-service calculation, and possibly a smaller first house. The household that comes back later with clean, verifiable documents pulled from CRA themselves usually finds the door still open, just at a lower price point than the one they first wanted.
Consider a couple in the Cowichan Valley in the spring of 2026, shopping for a detached house somewhere between Duncan and Lake Cowichan. One works full time at a regional employer. The other left a salaried job the previous year to take contract work, and the income is real but uneven. Under the federal stress test, they have to qualify at their contract rate plus 200 basis points, or 5.25%, whichever is higher. On paper, they come up a little short.
Then someone offers to help. It might be a relative who "does paperwork," or an online service that promises to "format" self-employed income so lenders can read it. The pitch sounds reasonable: the money is genuinely being earned, the T4 from the old job is messy, and lenders only want a clean summary. The couple hands over their documents and gets back a tidy package with pay stubs, an employment letter, and a year-end summary. Every column lines up. No typos, no fuzzy scans. It looks more professional than anything they could have produced themselves.
That is the decision worth slowing down on. At the time it did not feel like fraud. It felt like presentation.
What changed on the other side of the desk
Equifax Canada's fraud and identity team has been pointing out something uncomfortable this fall. Reported mortgage fraud rates have been falling, and that may not mean there is less fraud. Generative AI now makes it cheap to produce pay stubs, bank statements and employment letters that pass a visual check. A fake that gets through is never counted. Equifax's warning is that the lower number may partly reflect how much is going undetected.
Falsified income is still the most common kind of mortgage fraud in Canada. The difference now is that it no longer looks like a doctored PDF with mismatched fonts. Investigators have started to treat a document that is too perfect as a reason to look harder. A pay stub from a small Duncan contractor that looks like a design agency built it is an anomaly in its own right.
So lenders have changed what they check. Validating a document only tells you that it exists and looks right. Verifying it means confirming that the numbers match the source: the Notice of Assessment in CRA My Account, a call to the employer, deposits that actually appear in the bank account. More lenders in British Columbia now run that cross-check automatically.
The turning point
In a case like this, the file usually comes apart at verification, not at the first review. The application gets a conditional approval. Then the lender asks for the CRA Notice of Assessment, pulled directly by the borrower. The income on it is lower than the "formatted" summary. The gap may not be large. It doesn't need to be.
The deal is declined. The bigger problem is that the file may now be flagged in the databases lenders share, and that flag follows the borrowers. Mainstream lenders may be closed to them for years. Under the Criminal Code, fraud over $5,000 carries a maximum sentence of 14 years. Most cases never get near a courtroom, but the borrowers' names are attached to the application either way. "Someone else prepared it" is a hard argument to make once their signatures are on it.
Honest borrowers feel the effects too. They now get more requests for screenshots straight from CRA My Account, more calls to employers, and slower conditions. That friction is the cost of an arms race in which AI tools produce the fakes and AI tools are also the most practical way to catch them. When enough fraud gets through, lenders build that risk into their pricing, and every borrower pays for it.
A broker who knows the local market still catches things software misses, like a claimed salary that doesn't fit the role in this region. But the more useful protection comes earlier, before anyone hands their documents to a stranger.
The couple in this scenario had a legitimate path the whole time. It involved two years of tax returns, an honest debt-service calculation, and possibly a smaller first house. The household that comes back later with clean, verifiable documents pulled from CRA themselves usually finds the door still open, just at a lower price point than the one they first wanted.
Read Next
Home Prices in 2027 Hinge on a Handoff From Population to Business Investment
Bank Rate Hikes Lift Your Qualifying Rate: 3 Moves on Term and Rate Shopping
Steady Policy Rate, Rising Fixed Rates: Reading the Gap Before You Lock In
The 39% Limit: How the Stress Test Sets Your Real Buying Power