During the Football World Cup, insurers reliably face a wave of claims reports: damaged TVs, monitors and electronic devices. This is not a new phenomenon. A much-cited GDV analysis of nearly 1,455 submitted TV damage claims concluded that one in four reports contained inconsistencies. The description of the damage and the damage pattern did not match. At the time, the average claim amounted to 745 euros. Insurers continue to observe this effect ahead of major tournaments to this day.
What is new is not the motive, but the tool. Where a television used to “unfortunately fall off the table,” an AI-generated image is enough today. In a recent analysis from early June 2026, the GDV reports cases that would have been unthinkable just a few years ago: entirely artificially generated photos of allegedly defective smartphones and AI-generated X-ray images submitted to a pet health insurer that were intended to document operations that never took place.
That may sound like two curious episodes. In fact, they describe the same question – only around twelve years apart: When and how do insurers recognise that something is wrong?
In claims settlement, timing is crucial
Insurance fraud is not a marginal issue for the industry. The German Insurance Association (GDV) now estimates the losses at more than six billion euros per year. Around one in ten claims reports is considered suspicious and worthy of investigation, according to a special GDV analysis of more than 600,000 cases over three years. The fundamental problem lies less in the volume than in the moment of detection.
In many claims processes, suspected fraud is only identified shortly before payment, when the case has long since been processed. Documents have been reviewed. Customer communication has been initiated. Claims handlers have invested time. Internal costs have been incurred.
If suspicion is then identified, the entire process must be rolled back. This is costly, time-consuming and avoidable.
“Fraud detection must not take place only shortly before payment,” says René Weseler, Senior Executive Manager at Buildsimple. “By then, the case has long since been processed, documents have been reviewed, customer communication has been initiated and internal costs have already been incurred.”
Deceptively real: Digital evidence under pressure of trust
In addition to timing, there is another factor increasing the pressure on insurers: fraud methods are becoming more sophisticated. Generative AI is also changing the framework conditions for claims processing. Today, images can be manipulated, generated or taken from other contexts more easily.
For insurers, this means that traditional visual inspection is becoming less meaningful. The question is no longer whether digital evidence can be used, but how its authenticity can be verified efficiently. Modern image forensics can detect whether photos have been manipulated, reused multiple times or taken from another context. Buildsimple integrates such checks – for example, through our forensic partner VAARHAFT – directly into the claims process. Here too, timing is crucial: the check is carried out early and in the context of the entire case, rather than in isolation shortly before payment.
It is also important to note that AI does not determine fraud. It identifies patterns, inconsistencies and risk signals, and routes conspicuous cases out for controlled handling. The final assessment remains the responsibility of claims handlers, assessors and the responsible specialist departments.
Detect early instead of reworking late
A claim does not consist of a single document. It consists of documents, photos, emails, free-text entries, policy information, processing history, rules and risk signals.
If this information is only brought together late in the process, or manually, blind spots arise. This is precisely where intelligent document processing comes in.
Buildsimple classifies claims documents directly upon receipt, extracts the relevant specialist data and transfers it automatically to downstream target systems and processing paths. An early decision is made:
Can this case continue through the process automatically?
Are documents missing that need to be requested?
Are there anomalies that justify a special review?
Unremarkable cases proceed quickly and smoothly. Conspicuous cases reach the right people early, before costs arise. Early review does not hold up the honest majority. It gets straightforward claims to their destination faster, while the few conspicuous cases are directed specifically to where experts can assess them.
“Modern claims settlement is more than an online form,” says Weseler. “A claim consists of documents, photos, emails, free-text entries, policy information, history, rules and risk signals. This information must be brought together early and routed into the appropriate processing path.”
What applies in cases of proven fraud
If a case of fraud can be proven, it has noticeable consequences: according to the GDV, the insurer does not have to pay, may terminate the contract, reclaim expert costs and file criminal charges. Identifying suspicion early therefore protects not only against the direct loss, but also against the effort involved in a late review.
Fair play with Buildsimple
Why Buildsimple? Image forensics only comes into play once someone examines the individual photo. Buildsimple starts earlier: upon receipt, all components of a case – documents, photos, invoices, free-text entries – are classified, extracted and routed into the appropriate processing path. This allows specialised checks to run in the context of the entire case rather than as an isolated individual step. That is the difference between “detecting a fraud” and “setting up the process to be fraud-proof and fast from the outset”
A multi-AI approach ensures that the right method is used for every task. For stable, cost-effective and traceable results. The platform is GDPR-compliant, TÜV-certified, ISO 27001-certified and suitable for use in BaFin-regulated environments.
And so the supposedly curious TV damage claims during the World Cup are ultimately more than an anecdote. They illustrate what matters in modern claims settlement. Anomalies are not identified only at the end of the process, but as early as possible, before effort, costs and risks arise.
Would you like to find out how Buildsimple accelerates your claims settlement?
The fastest way to assess this is by looking at your own claims portfolio. In a demo, we show you, based on your document and claim types, which cases could be processed automatically, where documents are missing and which anomalies would trigger an early review – specifically using your cases, not in the abstract. See from your own claims how much faster straightforward cases can proceed and how early conspicuous cases are identified.
FAQ
AI can identify patterns, inconsistencies and risk signals in claims reports and flag conspicuous cases for more in-depth review. The final decision on a case of fraud is always made by a human (claims handlers, assessors or specialised departments).
As early as possible and ideally directly upon receipt of the claim. If suspicion is identified only shortly before payment, the case has already been fully processed and the effort required to reverse it is considerable.
Image forensics checks whether submitted claim photos have been manipulated, AI-generated, reused multiple times or taken from another context. The result is a forensic indication for further review, but not a final judgment.
According to the GDV, fraud causes losses of more than six billion euros per year for the German insurance industry. Around ten percent of all claims reports are considered suspicious.
