AI photo fraud: the alarming figures in 2026

Artificial intelligence has turned image creation into a process anyone can access. In 2026, generating a realistic photo of a product, a claim or a document takes just seconds with tools like Midjourney, DALL-E 3, Stable Diffusion XL or Flux. This democratization has a downside: fraud using artificial imagery is exploding at an unprecedented pace, affecting e-commerce, insurance, logistics and individuals.

How much does e-commerce fraud cost in France?

The French e-commerce federation (Fevad) estimates the cost of e-commerce fraud in France at 2.3 billion euros per year in 2024, a figure that has been rising steadily since the boom in generative AI. This amount includes fraudulent refunds, abusive chargebacks, fake returns and disputes involving falsified photos. At the European level, the total cost exceeds 18 billion euros a year according to the European E-commerce Report 2024 published by Ecommerce Europe. Image fraud accounts for a growing share of these losses, as it is particularly hard to detect and to prove without a certification tool.

What role do photos play in online disputes?

According to a Trusted Shops study published in 2024, 73% of disputes between buyers and sellers on marketplaces involve photographic evidence. Photos are used to document the condition of a product before shipping, on arrival, or to justify a claim. The problem is that these photos are no longer reliable: with AI editing tools, altering a detail in an image takes less than 10 seconds. A buyer can photograph a product in good condition, then generate an altered version showing a non-existent defect to obtain a refund. Without a certification mechanism, the platform cannot tell the real photo from the fake one.

How is generative AI multiplying photo fraud?

Juniper Research estimates that the number of fake photos used in a fraudulent context has quadrupled since consumer generative AI arrived in late 2022. This explosion is explained by the convergence of three factors: the growing quality of image-generation models that produce results indistinguishable from real photos, the accessibility of these tools available for free or for a few euros a month, and the increasing difficulty of automatic detection as the models improve. In 2026, the most advanced AI-image detectors show accuracy rates of only 65 to 80% on the latest generations of models, making after-the-fact detection less and less reliable as a sole line of defense.

Which sectors are most affected?

E-commerce and C2C marketplaces (Vinted, Leboncoin, eBay) are on the front line. Fake-return scams and photo disputes are a growing loss item for these platforms. But other sectors are also affected. Insurance faces claims backed by AI-generated or retouched photos, invented water damage, staged car accidents, fictitious valuables. Logistics and transport suffer delivery disputes with photos of supposedly damaged parcels. Real estate sees listings appear with AI-generated interior photos bearing no relation to the actual property.

Why is certification at the source the answer?

Faced with the gradual failure of after-the-fact detection, certification at the moment of capture is emerging as the only structurally reliable approach. The principle is simple: rather than trying to guess whether a photo is fake afterwards, you certify that it is authentic at the exact moment of its creation. This is the approach taken by CertiPix, which generates a timestamped cryptographic certificate for every photo taken via the app. This digital fingerprint (SHA-256 hash) guarantees that the photo hasn't been altered and that it genuinely comes from a real physical sensor, not from an AI generator. The C2PA standard (Coalition for Content Provenance and Authenticity), backed by Adobe, Microsoft and Google, moves in the same direction on a global scale.