AI Fraud Moving to Smaller Motor Claims Firms

The large carriers have spent heavily on fraud defences. That does not make organised fraud go away. It makes it look for an easier target, and that target is starting to look a lot like a smaller operator.

Two signals landed in the motor fraud world this month, and read together they point straight at the smaller end of the market.

The first came from Quantexa, whose fraud team warned that organised, AI-enabled fraud is likely to shift toward smaller insurers and firms as the major carriers get better at catching it. The second came from Aviva, which reported stopping record levels of claims fraud in 2025 and named AI-generated images and manipulated documents as a growing driver, particularly in motor.

Put those two next to each other. The biggest players are catching more than ever. So where does the fraud that used to work go next.

The defences went up at the top

Large carriers have invested in AI-enabled detection, analytics teams, and shared industry data. That investment is showing up in their numbers. Aviva's record year is not a sign that fraud is out of control. It is a sign that a well-resourced insurer is now stopping a lot of what it used to pay.

That is good news for them. It is more complicated for everyone downstream.

Fraud is not a fixed pool that shrinks when one firm gets better at spotting it. Organised fraud is a business, and it behaves like one. When a route stops paying, it does not close down. It finds a route that still pays.

Quantexa's own framing was that the industry is now in "an AI arms race" with the people committing the fraud. The uncomfortable part of an arms race is that it is not fought evenly. The side with the most resource pulls ahead first, and everyone else becomes the softer flank.

Fraud follows the path of least resistance

This is the bit that matters for a smaller book.

A fabricated accident scene or an inflated repair invoice that bounces off a tier-one insurer's detection model does not get thrown away. It gets sent somewhere the model is thinner. Credit hire firms, accident management companies, and smaller CMCs sit lower down the chain, and they often screen with less automation and fewer dedicated analysts.

The same fake, aimed at a smaller target, has a better chance of landing clean.

None of this is a comment on handler quality. A sharp handler at a small firm is every bit as sharp as one at a large insurer. The difference is structural. The big carrier has a model running behind the handler on every file. The smaller operator often has the handler and not much else.

That gap is exactly what an organised fraudster is now paid to find.

What changes for a smaller book

The fakes themselves are not new, and this is not another piece on how to spot them. We have covered the detection mechanics already, in Spotting AI-Faked Evidence Before It Becomes a Payout and in the fake-policy problem in When the Policy on Your File Was Never Real.

What changes is not the method. It is the probability and the concentration.

The probability goes up because you are now a more attractive destination for attempts that no longer work elsewhere. You may see a higher share of AI-assisted fraud simply because the well-defended firms are pushing it toward you.

The concentration is the harder problem. A large insurer absorbs a paid fake across enormous volume. A smaller operator feels every single one. One fraudulent credit hire file that pays out can wipe the margin on a dozen clean ones. The exposure per hit is far higher when the book is smaller, and that is before you count the time spent unwinding a file that should never have moved.

The posture, not another checklist

The instinct is to reach for a longer list of things to check on each photo. That is worth doing, and the detection pieces above set it out. But the strategic answer is about posture, not a checklist.

First, assume you are now a target. The old comfort of being too small to bother with has gone. Being small is the reason you are worth bothering with.

Second, make screening a property of the workflow, not of the individual. If your best defence is one experienced handler who has seen a few fakes, that defence walks out of the door when they are busy, off, or gone. Consistent flags applied to every inbound file are what beat a fraud method designed to exploit the rushed and the distracted.

Third, keep the audit trail as the work happens. When a file does turn out to be fraudulent, being able to show what was checked and when is what protects you with insurers, panels, and the courts.

Fourth, use the network. Smaller does not have to mean isolated. Panel relationships, industry bodies, and shared red-flag intelligence let a small operator borrow some of the pattern-spotting that a large carrier builds in-house. Enforcement is organising too, with bodies like IFED stepping up work on spoof insurer sites.

None of that requires an insurer-sized budget. It requires treating fraud screening as part of how the desk runs, rather than something a good handler remembers to do on a quiet afternoon.

The takeaway

The headline from the big carriers is that AI-enabled fraud can be caught. The quieter headline, the one aimed at everyone else, is that catching it at the top pushes it down the chain.

Smaller operators are not too small to be a target. Being smaller is the reason the target is moving your way.

You do not need an insurer-sized budget to respond. You need to assume you are now in scope, build the checks into the workflow rather than the handler, and keep the audit trail that proves you looked. The firms that do that quietly stop being the path of least resistance.