We build receipt fraud detection infrastructure for expense teams.
AI-generated fake receipts went from a curiosity to a real line item on finance teams' fraud reports in under two years. airegco exists to give expense-management platforms a fast, explainable signal for "is this receipt real — and does the math add up?"
Why receipt authenticity matters.
Anyone can generate a photorealistic fake restaurant receipt in seconds with freely available AI image tools — no template, no Photoshop skill required. Independent reporting on enterprise fraud data puts AI-generated receipts at over 70% of flagged expense fraud by mid-2026, up from effectively zero a year earlier. Traditional OCR-and-keyword review was never built to catch this.
We built airegco around two independent signals instead of one opaque score: AI-image forensics that checks whether the receipt photo itself shows signs of generation, and an OCR arithmetic-consistency check that verifies the subtotal, tax, and total actually reconcile. Either signal alone can flag a receipt — and you see which one tripped, not just a number. v1 is focused specifically on restaurant receipts, not a generic all-document detector.
We charge per check because we want alignment: we get paid when you use it, not when you sign a contract. Credits never expire. There is no subscription. Start with a free check and scale to millions.
Principles we build with.
Engineering excellence
We ship fast but never cut corners on correctness. Every model claim is backed by a test set.
Privacy by default
Images are never stored. No training on customer data. No third-party analytics on the detection path.
Open and verifiable
Our detection methodology is documented. We publish what we can. You should be able to audit what you rely on.
Three things that don't change.
Engineering excellence
We ship fast but never cut corners on correctness. Every model claim is backed by a test set.
Privacy by default
Images are never stored. No training on customer data. No third-party analytics on the detection path.
Open and verifiable
Our detection methodology is documented. We publish what we can. You should be able to audit what you rely on.