Insurance photo fraud has always evolved with technology. Now it’s evolved again – and this time, it’s harder to catch with the naked eye.
A recent BBC Wales investigation revealed that AI-generated insurance fraud is no longer a theoretical risk. Cardiff-based insurer Admiral reported a 71% rise in fraud during 2025, and pointed directly to generative AI as a driver. Customers and organised crime groups are now using AI tools to invent damage, fabricate documents, and duplicate claims – all with images that look convincing at a glance.
For auto glass and windscreen claims specifically, this shift matters more than most people realise. Here’s what the data shows, what it looked like in practice, and what claims teams can do about it.
What Is AI-Generated Insurance Fraud?
AI-generated insurance fraud is the use of generative AI tools to create or manipulate visual “evidence” submitted with an insurance claim. Instead of physically damaging a vehicle or staging a scene, a claimant can now generate a photo-realistic image of damage that never happened — in seconds, on a phone.
According to the Insurance Fraud Bureau, this is now “a trend across the entire insurance industry,” not an isolated Admiral problem. Two distinct groups are driving it:
- Opportunistic customers, who use AI to exaggerate a genuine claim
- Organised crime gangs, who use AI to fabricate documents and make large-scale fraud “more efficient”
How DriveX verifies glass damage in under 3 seconds
Real Cases: Fake Plates, Invented Watches, Exaggerated Damage
Admiral’s fraud team shared several documented examples with BBC Wales. Three stand out because they show how differently AI fraud can present itself:
1. A duplicated claim with a swapped number plate. A damaged Land Rover was submitted twice — the same vehicle, the same damage, but with an AI-altered number plate the second time, designed to make it look like a separate incident.
2. An invented luxury watch. A basic, clearly AI-generated image of a gold and diamond watch was submitted as “proof” of a stolen item — an object that never existed.
3. Exaggerated bumper damage. Real damage to the back of a car was digitally enhanced with AI to look far worse than it actually was — inflating the claim value.
All three were caught. But they were caught by a dedicated fraud team at a major insurer with the resources to investigate manually – not every claims desk has that.
Why Auto Glass Claims Are Especially Exposed
Glass and windscreen claims have three features that make them a natural target for this kind of fraud:
- High volume, low average value — which historically meant less scrutiny per claim
- Visual-only assessment — damage is judged almost entirely from photos, not physical inspection
- Fast turnaround expectations — customers expect same-day approval, which limits time for manual review
That combination — lots of claims, photo-based decisions, and speed pressure — is exactly the environment where a convincing AI-generated image can slip through unnoticed.
How the Industry Is Responding
The response so far has been twofold. Insurers are investing in AI-powered detection to counter AI-powered fraud, and the industry is starting to share intelligence rather than treating this as a competitive secret.
As John Davies of the Insurance Fraud Bureau put it: “It is a fast-moving issue, but I think what is positive is the collaboration across the industry… there are opportunities there in how we can share knowledge and best practice to help use AI in a positive way.”
That’s the key shift. The same generative technology being used to commit fraud can be countered by automated damage assessment systems trained specifically to spot manipulation, inconsistency, and duplication – at the speed claims volume actually requires.
What This Means for Glass Networks and Insurers
For claims teams handling glass damage specifically, three questions are now worth asking:
- Can our current process tell the difference between real and AI-manipulated damage photos?
- How long would a duplicated or altered claim currently take to catch — hours, days, or not at all?
- Is fraud detection built into the assessment step itself, or bolted on afterward?
This is the exact gap automated visual damage assessment is designed to close — verifying real damage against a trained model at the point of claim, rather than relying on manual review after the fact.
See how DriveX’s technology saved 4.1M$ for a Latin American Third Party Administrator!
Key Takeaways
- AI-generated insurance fraud rose sharply in 2025, with Admiral alone reporting a 71% increase
- Fraud now includes fabricated items, duplicated claims, and digitally exaggerated damage
- Glass and windscreen claims are structurally exposed due to volume, speed, and photo-only assessment
- The most effective counter is automated damage verification built into the claims workflow, not manual review after the fact
Fraud tactics will keep evolving. The claims processes built to catch them need to evolve just as fast.