A decade ago, most consumer lenders priced loans off a small handful of variables: credit score, income, debt-to-income ratio, loan amount. Today, AI-driven underwriting models can ingest thousands of features — including how long you spent on the application page, what device you used, and where your phone has been. Two applicants with identical credit profiles can now be quoted meaningfully different APRs and repayment schedules. The industry calls this personalization. Whether borrowers benefit depends almost entirely on which side of the model you sit on.
What “personalized” actually means in lending
Modern underwriting models optimize for a lender’s expected profit per applicant, not for the borrower’s best terms. That means the algorithm is looking for the highest rate it can offer that still gets the loan accepted. Variables that correlate with lower price sensitivity — urgent need, limited shopping behavior, older age, lower financial literacy proxies — can quietly nudge an offered APR upward. Repayment schedules can be customized too, with due dates timed to payday and minimum payment structures designed to maximize total interest paid over the life of the loan. None of this is illegal. Most of it is invisible to the borrower.
The case for: better access, fewer one-size-fits-all rejections
It would be unfair to call AI lending uniformly predatory. Fintech models have extended credit to thin-file borrowers who would have been auto-rejected by traditional scorecards, and some have demonstrably reduced racial pricing gaps relative to legacy underwriting. Cash-flow underwriting that looks at bank account activity often serves people more accurately than a FICO score alone. The technology is genuinely capable of expanding access. The question is whether the access comes with terms that the borrower would have accepted in a transparent market.
The dynamic pricing problem
Where personalization slides into predatory territory is dynamic pricing — APRs that adjust based on behavioral signals unrelated to creditworthiness. If a borrower applies at 2 a.m. from a payday-loan referral, models can read that as desperation and price accordingly. If they comparison-shop across three lenders, models can detect that and offer better terms. The result is a market where the same person can get materially different prices depending on how stressed they look. Regulators are catching up: the CFPB has signaled increased scrutiny, and the EU’s AI Act treats credit scoring as high-risk. Enforcement remains thin.
How to protect yourself in an AI-priced market
Shop more than you think you need to. Pull offers from at least three lenders, preferably including a credit union, before accepting anything. Don’t apply when you’re rushed — models can tell. Read the repayment schedule, not just the APR; long-amortization, low-monthly-payment loans often hide enormous total interest. And know that pre-qualification soft pulls let you see real offers without damaging your score, so there’s no excuse not to comparison shop.
The bottom line
Personalization in lending is a tool, and like most tools it serves whoever holds it. Borrowers who shop, slow down, and read the fine print get something close to a fair price. Borrowers who don’t are paying for the privilege of being predictable.
Leave a Reply