As of 13 August 2026, AI can only partly check whether a business invoice is fraudulent.

What to actually do

1. Hand it to a person

   The route this page recommends

   Someone with a licence or accountable authority has to sign this before it counts.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

   How to actually do it

   1. Open your accounting or procurement records and gather the invoice, purchase order, delivery evidence, previous invoices and the supplier's established contact details.
   2. Check whether the invoice requests new bank details, unusual urgency or a payment method that differs from your normal supplier process, and record those differences.
   3. Remove unnecessary personal data, account credentials and unrelated customer information, then paste the remaining invoice and records into a chatbot with the supplied prompt.
   4. Read the output's evidence table and compare every stated fact with the original invoice and your accounting records, deleting any observation the model cannot support.
   5. Contact the supplier using a telephone number or email address already held in your records, not the details on the invoice, and independently confirm the invoice number, amount and bank details.
   6. If anything remains unexplained, pause payment and send the evidence to your bank's fraud team and your accountant or solicitor before authorising the transaction.

   Prompt

   Act as an invoice-fraud triage assistant, not a decision-maker. Analyse the invoice and supporting information below for inconsistencies and warning signs, including changes to bank details, unusual urgency, mismatched supplier names, altered payment terms, duplicate invoice numbers, unexpected amounts, missing purchase-order references, formatting anomalies and discrepancies with our records. Separate what is directly visible from what is only an inference. Do not claim that the invoice is genuine or fraudulent, do not invent missing facts, and do not recommend payment. Return: 1) a table of observed details, 2) warning signs with the evidence for each, 3) benign explanations that should be considered, 4) information still needed, and 5) a cautious risk rating of low, medium or high with a short explanation. Include a specific verification checklist that uses contact details already trusted by our business, not contact details printed on the invoice. Flag any sensitive data that should be removed before sharing. Invoice: [paste invoice text or upload the invoice]. Supplier records and known contact details: [paste relevant records]. Purchase order, delivery and payment history: [paste relevant records].

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.