FAQ
Questions restaurant owners ask about fake complaints
Everything we get asked about doctored photos, delivery-app refunds and how FraudBite works. Every answer has its own link, so you can send one to a colleague.
Can FraudBite detect AI-generated food images?
Yes. FraudBite looks for the traces image generators leave (pixel and noise patterns, generative fingerprints and embedded metadata) and marks the suspect areas on a heatmap. Treat the result as a strong signal, not a verdict: very small edits are harder to catch, so look at the marked areas and the reasons before you decide.
Does FraudBite also detect manually edited photos?
Yes. Retouching, darkened or recoloured areas, added defects and pasted-in objects are what the heatmap is for. Detection is strongest on whole-image changes and AI-generated areas; a tiny, carefully blended edit can slip through, which is why every result shows you where and why, not just a number.
Is customer data secure?
Your photos and results are stored in Google Cloud’s EU region (Frankfurt), encrypted in transit and at rest, and only your account can open them: photos are shown through short-lived signed links. To run the check, each photo is also sent to our image-analysis provider, TruthScan, in the United States, which also uses the photos it receives to train and improve its detection models. Every provider we use is listed in our Privacy Policy.
How accurate is FraudBite detection?
No detector is right every time, and we won’t quote you a magic percentage. Independent research shows every current tool gets weaker on small, blended edits, and a genuinely burnt or damaged dish can look suspicious. That’s why FraudBite gives you a score plus the heatmap and the reasons, so you can weigh them against the order and the customer’s history before you refund or contest.
Do I need technical knowledge to use FraudBite?
No. Upload an image and receive a clear fraud score and report through a simple dashboard.
Does it work with photos from Uber Eats, Deliveroo or Glovo?
Yes, whenever you have the photo: save it from the dispute, the chat, the email or the review and upload it. FraudBite analyses the image itself, so it doesn’t matter where it came from. Not every app shows restaurants the customer’s photo (Glovo in Romania stopped in October 2024), which is why your own pass photo of each order matters too.
Can I use the result to dispute a refund or a chargeback?
You can attach the PDF report (the score, the marked areas and the metadata findings) to your reply to the platform or your bank. They still make the decision, but a documented, technical answer is far harder to dismiss than “the food was fine”.
What does a high risk score actually mean?
That the photo shows strong signs of editing or AI generation and deserves a closer look. It is a signal for your decision, not proof of fraud on its own: weigh it together with the order, the timing and the customer’s history.
How much does FraudBite cost?
You get a free trial of 3 checks, with an email-only account and no card. After that, Unlimited costs €4.99 a month or €47.90 a year and checks as many complaint photos as your restaurants receive (fair use applies). Both include the heatmap, the PDF report and your full history.
Which photo formats can I upload?
JPG, PNG and WEBP up to 10 MB each: the formats delivery apps, WhatsApp and email hand you. Screenshots work too, but the original file the customer sent gives the check more to work with, because a screenshot strips the camera metadata.
How long does a check take?
Typically well under a minute per photo. You upload, the analysis runs, and the score, the marked areas and the detected features appear in the dashboard; the PDF report is ready to download straight away.
Where are my photos stored, and for how long?
Your photos and results are stored in Google Cloud’s EU region (Frankfurt) for as long as your account exists or until you delete them. For the analysis itself, each photo is uploaded to TruthScan’s storage in New York (USA). TruthScan keeps its copy under its own terms and uses the photos it receives to train and improve its detection models, so deleting a photo in FraudBite doesn’t remove TruthScan’s copy. Crop out names, addresses or faces before you upload. You can ask us to delete everything we hold at any time at [email protected].
Are the photos I check used to train AI?
Yes, by TruthScan, the provider that runs the analysis: it uses the photos it receives to train and improve its detection models. FraudBite itself doesn’t train models on your photos. Upload only what the check needs, and crop out a customer’s name, address or face first. The details are in our Privacy Policy and Terms.
Is there an annual plan?
Yes. Unlimited can be billed yearly at €47.90 instead of €59.88 for twelve monthly payments: 20% off. Promo codes apply to the monthly plan only, since annual is already discounted. To move an existing subscription between monthly and annual, write to [email protected] and we switch it at your next renewal.
Does the free trial run out?
Yes, after 3 checks. It is a one-time allowance with no time limit and no card: use the 3 checks whenever you like, then move to Unlimited (€4.99 a month) to keep checking.
What if a genuine complaint gets flagged?
It can happen: a genuinely burnt dish can look suspicious to any detector. That’s why FraudBite shows the marked areas and the reasons, not only a score: you judge, with the order and the customer’s history in front of you. When in doubt, refund. The goal is to stop the fakes, not to fight real complaints.
Still unsure?
Check your own photos with a free trial, or write to us: we reply within one business day.
Or try it on a sample photo, no account needed