Cotality data shows purchases jumped to 72% of Q2 volume, and where purchases go, fraud risk follows
While many in the mortgage industry were hoping for a rate decline in 2026, leading to a wave of refinances, that has not happened so far. In addition to cutting off the refi market, higher rates are also increasing mortgage fraud risk.
Cotality's National Mortgage Application Fraud Risk Index reached 132 in the second quarter of 2026, up 11 points or 9.1% from the first quarter, with an estimated 1 in 119 mortgage applications showing indications of fraud risk. The index remains down 4.6% year-over-year from 138 in Q2 2025, but the quarter-over-quarter jump reflects how quickly the purchase market has grown.
Purchase loans accounted for 72% of overall mortgage volume in Q2 2026, up from 59% in Q1, not because purchase demand surged but because rates did not fall as many had expected, leaving fewer borrowers in a position to refinance.
The relationship between the purchase mix and fraud risk is well documented. Purchase loans require more documentation than refinances, and more documentation gives fraudsters more to work with.
Matt Seguin (pictured top), senior principal at Cotality Mortgage Fraud Solutions, said the jump in purchase share was larger than he anticipated.
"I was a little surprised about how big the jump was from 59% purchases up to 72%," Seguin told Mortgage Professional America. "Obviously, rates have an impact there. And really, the opportunity to commit fraud in purchases is larger than refis."
Why fraud risk increased
Seguin said the difference comes down to documentation. Government-backed streamline refinance programs often do not require income, asset, or appraisal documentation, whereas purchase loans require all of those as standard, which creates more surface area for fraud.
He said when you break it down, the conditions for mortgage fraud are not that different from any other crime.
"Like any crime, you need a means — and now with AI and things of that nature, it's a little easier to do that," he said. "And then you need really the motivation. And it's usually one of two things: the borrower wants to get into that dream house, or somebody on the inside, a mortgage insider, wants to make an extra commission. So similar to any crime, you have the contributing factors from that aspect."
The category showing the largest year-over-year increase in Q2 was undisclosed real estate debt, which rose 2.6%. Cotality's data shows these alerts are 2.5 times more likely to fire on an investment property than on an owner-occupied property, with investment applications showing 1 in 44 with fraud risk indicators and multi-family at 1 in 27, compared to the overall rate of 1 in 119.
Seguin said that while most fraud risk categories showed year-over-year declines, that should not be read as a signal to relax.
"Don't become complacent," he said. "The fraud is still out there, and we're just reporting that the risk of that category is dropping a little. Having the data to know when a loan is potentially riskier than another and knowing what to pull off the conveyor belt — that's the old work smarter, not harder kind of thing."
The AI arms race
Seguin said AI is changing fraud on both sides of the equation. Lenders are deploying it to detect suspicious patterns, and fraudsters are using it to make fabrications harder to catch.
"The days of whiteout on a document are pretty much gone, and maybe misspellings or major math errors — the obvious stuff to an eye," he said. "So using AI tools is great from a fraud detection standpoint, but I haven't heard a lender really ready to turn it all over to the machines. Nothing is yet able to replace a really good underwriter's intuition, that gut feel that something's wrong."
He said AI document review tools from third-party vendors are still developing, and several lenders told him that when they tested tools designed to detect altered PDFs, the tools did not always catch the alterations.
He said what holds up better over time is pattern recognition across loans rather than document-by-document review.
"Does that machine recognize that the McDonald's worker making $500,000 a year is out of whack with reality, or does it say the pay stub looked good, the calculations line up, keep it going?" he said. "An underwriter's not going to see a pattern on a one-off loan basis. But trying to spot that pattern sooner rather than later can be really beneficial."
He said the right mindset for brokers going into any application should be trust but verify.
"Make sure you double-check some of those income docs yourself or make those phone calls to validate the employer, just to make sure you feel pretty good about what you're submitting in your name," he said.
Stay updated with the freshest mortgage news. Get exclusive interviews, breaking news, and industry events in your inbox, and always be the first to know by subscribing to our FREE daily newsletter.
