The Pulse
In Extend’s Benchmark, Verify Detected 96.3% of Fraudulent Documents
Extend announced Verify, an API that checks documents for signs of fraud before businesses or AI agents rely on them. In the company’s benchmark evaluation, Verify detected 96.3% of fraudulent documents with a 4% false-positive rate, according to the company, but the announcement does not disclose the test’s size or composition.

Extend’s new document-fraud detection API identified 96.3% of fraudulent documents in a benchmark evaluation, the company says, while falsely flagging 4% of legitimate ones. The company announced Verify on October 19, 2026, at Money20/20 USA, pitching it as a check developers can add before AI agents act on documents.
Source businesswire.com
The figures come from Extend’s own evaluation. The test measured how often Verify identified fraudulent documents and how often it flagged legitimate ones, but the announcement does not give the benchmark’s size, the document mix or other details that would let customers compare the result with other detection systems. The false-positive rate matters alongside the detection figure: those alerts can send legitimate documents for further review.
Verify looks beyond extracted text
Verify examines a document’s visual content, file structure and revision history, combining specialized models, forensic checks and agentic analysis. It is designed to flag fully fabricated documents, targeted changes to genuine records—including edits as small as one character—conventional digital alterations such as patched text or image overlays, and changes made after a digital signature.
The service returns detected signals, risk assessments and scores, explanations, and traceable evidence. Businesses can use those findings to decide which documents need additional review. Extend describes the service as a way to assess document integrity within processing pipelines, rather than relying only on the information extracted from a file.
One check for invoices, claims and lending files
Extend’s examples center on documents that can trigger consequential decisions. A bank statement might support a lending decision, an invoice can initiate a payment, and a receipt can substantiate an insurance claim. Verify could flag suspicious changes to an income document before it informs a credit decision, surface an altered invoice before payment is authorized, or identify a potentially fabricated receipt before a claim is settled, Extend says.
The product is aimed at workflows in which AI agents process documents. Extend argues that agents need a way to assess the integrity of their evidence, as well as extract fields from it. CEO Kushal Byatnal said:
“Businesses want to move beyond agents that search documents to agents that help get work done in the real economy. That requires trust in the evidence those agents act on. Verify helps developers build document integrity into their pipelines to prevent fraud-related losses, mitigate the difficulty of human review, and scale up agent deployments.”
Available by API, with benchmark details outstanding
Developers can use Verify independently through an API or add it to Extend pipelines that already parse, split and extract documents. The service is available now, and Extend invites developers to request access. Extend also demonstrated Verify at Money20/20 USA, where attendees could visit Booth 6085 and try Fraud Arcade, an interactive game about distinguishing legitimate documents from manipulated ones.
The launch places document-integrity checks alongside the extraction steps that feed automated workflows. But the headline performance figure remains a company-reported result: without a disclosed sample size, test composition or independent assessment in the announcement, buyers cannot tell how closely the evaluation matches the documents in their own queues.
Sources
- Original source · Press release Business Wire businesswire.com
Other pages read
- MarketMinute khsltv.marketminute.com