Independent writing on tax-smart planning, mortgage strategy, and retirement building. For Canadian professionals who want the whole picture, not just a piece of it.
Falling Mortgage Fraud Rates May Reflect What AI Has Learned to Hide
Equifax Canada's fraud data shows an odd split in 2026: fewer confirmed mortgage fraud cases, while the suspicious-activity flags in its automated systems stay high. Equifax calls the gap "silent fraud." Applications are still getting through. They just aren't being caught and counted.
A fraud rate can only measure the fraud someone noticed. For years, noticing meant spotting something off on the page: a misaligned font, a T4 with the wrong spacing, a Notice of Assessment that didn't quite match what the CRA issues. Generative AI has removed most of those tells. If you place mortgages or build plans around them, the check you've relied on longest, looking closely at the paper, now catches less than it used to, and a calm year in the statistics can hide that.
A clean document now proves very little
Forgeries used to take skill and time. Today a pay stub, T4 or NOA can be produced in seconds with the right fonts, watermarks and layouts of major Canadian banks and the CRA. Manual verification of a CRA document can still take days. Speed and polish are now on the forger's side.
Credit files have the same problem. Synthetic identity fraud blends real details, such as a genuine social insurance number, with invented ones to create a borrower who doesn't exist. Given enough time, that borrower can build a spotless credit history. A strong score used to be reassuring. Now it's one more thing that can be manufactured.
The practical lesson: confirm income and identity through a channel the client doesn't control. Have the client pull their NOA from CRA My Account while you watch, or use the lender's direct verification. Find the employer's phone number yourself instead of using the one on the letter. A forger can copy a document perfectly. Faking the source behind it is much harder.
Read the file as one story
Lenders are fighting AI with AI: liveness checks during ID verification, and digital footprint tests that look at how long an email address or online profile has existed. Those tools help, but they don't sit across the table from the client. You do.
Every honest file tells one consistent story. Stated income matches the tax return, the tax return matches the deposits, the deposits match the lifestyle, and the down payment has a history you can trace. Fraud tends to break down where two of those meet. Picture a salaried applicant whose bank statements show none of the usual payroll rhythm, or a large down payment that appeared in one transfer a month before closing. Neither one proves fraud. Both are worth a question, because each document on its own would pass.
Pressure raises the risk too. Uninsured borrowers must qualify at their contract rate plus 2% or 5.25%, whichever is higher. That rule pushes some honest buyers toward "fraud for shelter," a small tweak to income to make the numbers work. Organized "fraud for profit" rings, using straw buyers, look for files nobody reads closely. Both depend on no one asking how the story fits together.
The business owner may be the target
Self-employed clients usually draw extra scrutiny as possible fraudsters. More and more, they're the victims. Corporate registries are public, and a fraudster with a scraped director's name, a business address and AI-made paperwork can try to borrow against a property the real owner holds free and clear. That's why land title protections matter so much, and why they're only as strong as the application filed behind them.
The opposite mistake causes harm as well. Honest business owners with layered income, farm operations or dividends from a holding company can trip automated flags. When the story is complicated but consistent, your job is to explain it to the lender, not drop the client.
For clients with long-term structures, such as a readvanceable mortgage set up for the Smith Manoeuvre™, a fraud flag on the file can stall years of planning. A clean mortgage is the base every later step stands on.
Treat the next fraud statistic with healthy doubt. The useful question for each file is simple: did I confirm this through a source the client couldn't have shaped?
Equifax Canada's fraud data shows an odd split in 2026: fewer confirmed mortgage fraud cases, while the suspicious-activity flags in its automated systems stay high. Equifax calls the gap "silent fraud." Applications are still getting through. They just aren't being caught and counted.
A fraud rate can only measure the fraud someone noticed. For years, noticing meant spotting something off on the page: a misaligned font, a T4 with the wrong spacing, a Notice of Assessment that didn't quite match what the CRA issues. Generative AI has removed most of those tells. If you place mortgages or build plans around them, the check you've relied on longest, looking closely at the paper, now catches less than it used to, and a calm year in the statistics can hide that.
A clean document now proves very little
Forgeries used to take skill and time. Today a pay stub, T4 or NOA can be produced in seconds with the right fonts, watermarks and layouts of major Canadian banks and the CRA. Manual verification of a CRA document can still take days. Speed and polish are now on the forger's side.
Credit files have the same problem. Synthetic identity fraud blends real details, such as a genuine social insurance number, with invented ones to create a borrower who doesn't exist. Given enough time, that borrower can build a spotless credit history. A strong score used to be reassuring. Now it's one more thing that can be manufactured.
The practical lesson: confirm income and identity through a channel the client doesn't control. Have the client pull their NOA from CRA My Account while you watch, or use the lender's direct verification. Find the employer's phone number yourself instead of using the one on the letter. A forger can copy a document perfectly. Faking the source behind it is much harder.
Read the file as one story
Lenders are fighting AI with AI: liveness checks during ID verification, and digital footprint tests that look at how long an email address or online profile has existed. Those tools help, but they don't sit across the table from the client. You do.
Every honest file tells one consistent story. Stated income matches the tax return, the tax return matches the deposits, the deposits match the lifestyle, and the down payment has a history you can trace. Fraud tends to break down where two of those meet. Picture a salaried applicant whose bank statements show none of the usual payroll rhythm, or a large down payment that appeared in one transfer a month before closing. Neither one proves fraud. Both are worth a question, because each document on its own would pass.
Pressure raises the risk too. Uninsured borrowers must qualify at their contract rate plus 2% or 5.25%, whichever is higher. That rule pushes some honest buyers toward "fraud for shelter," a small tweak to income to make the numbers work. Organized "fraud for profit" rings, using straw buyers, look for files nobody reads closely. Both depend on no one asking how the story fits together.
The business owner may be the target
Self-employed clients usually draw extra scrutiny as possible fraudsters. More and more, they're the victims. Corporate registries are public, and a fraudster with a scraped director's name, a business address and AI-made paperwork can try to borrow against a property the real owner holds free and clear. That's why land title protections matter so much, and why they're only as strong as the application filed behind them.
The opposite mistake causes harm as well. Honest business owners with layered income, farm operations or dividends from a holding company can trip automated flags. When the story is complicated but consistent, your job is to explain it to the lender, not drop the client.
For clients with long-term structures, such as a readvanceable mortgage set up for the Smith Manoeuvre™, a fraud flag on the file can stall years of planning. A clean mortgage is the base every later step stands on.
Treat the next fraud statistic with healthy doubt. The useful question for each file is simple: did I confirm this through a source the client couldn't have shaped?
Read Next
Three CREA Downgrades in One Year Should Retire the 'Temporary Dip' Theory
Mortgage-Free in Retirement? The 4 Costs Your Home Still Charges You
Higher Rates, Softer Hiring: What 2026's Housing Shift Means for Your Equity
The Toronto Penthouse Premium: What Holds Value and What Is Just Status