Loading...
Loading...
This site uses cookies for authentication and language preferences. No tracking or advertising cookies are used. You can choose to reject all cookies — the site will still work, but language preferences won't persist across visits. Learn more
Platform Status: ● Operational · Verification Engine v50-MASTER+ID · 250-checkpoint forensic analysis · 43 attack vectors · 🇪🇺 Hosted in EU · Amsterdam
Drop a document below — our AI runs 250+ forensic checkpoints, calibrated against 43 attack vectors and an index of 23,029 institutions across 206 countries.
Drop your document here or click to browse
PDF · JPG · PNG · WebP · Max 20 MB
Each document is checked against this catalog. From vendor watermarks to ICAO 9303 check-digit failures, every AV-ID below is a known fraud pattern the engine recognizes and reports.
Submit any credential — diploma, government ID, certificate, or award — in any common image or PDF format.
Instant detection of vendor markers, phantom institutions, template fingerprints, and known fraudulent patterns.
Two-tier forensic analysis covering 250+ checkpoints — visual and structural integrity in tier 1, content logic and live institutional lookups in tier 2.
Downloadable 2-page A4 report: annotated credential image with flagged anomalies, plus detailed forensic findings.
No ambiguity. No hedging. Turing Verify returns one of three structured verdicts — each with a confidence score and full justification.
Confidence ≥ 90%. Visual integrity confirmed, institution verified, signatories authenticated, and live portal lookup completed.
Instant rejection triggered — vendor markers present, phantom institution detected, or a known template fingerprint matched.
Anomalies detected that prevent a clean verdict. Document requires escalation and human review before acceptance.
Credential fraud costs organizations time, money, and reputation. Turing Verify gives every professional a defensible paper trail.
Stop degree fraud before onboarding. Verify educational credentials and professional certifications with confidence — for every candidate, at any volume.
Detect fabricated transcripts, counterfeit diplomas, and misrepresented qualifications from applicants before admission decisions are made.
Meet KYC and AML obligations with authenticated government ID verification. Validate professional credentials for licensed advisors and traders.
Authenticate medical licenses, professional memberships, and training certificates submitted with claims — backed by a downloadable, audit-ready report.
All uploads and reports are encrypted in transit and at rest. Documents are never stored beyond your session unless you explicitly save them.
Credentials are processed only for verification purposes. No document data is sold, shared, or used to train external models.
Every PDF report includes a case ID, timestamp, methodology summary, and full findings log — ready for compliance, legal, or HR files.
Forensic Journal · No. AB7
It arrived in a background-check queue on a Tuesday afternoon: a serviceable Master of Business Administration, embossed, sealed, and signed. The claimant had built a career on it for twenty-eight years. Our AI needed four seconds to disagree.
No visible tampering. Correct weight of paper. A seal that felt, to the eye, earned.
Most forgeries are not caught by humans. They are not caught because the humans who receive them — HR specialists, university registrars, immigration officers — have four minutes to review a credential and no forensic training. A document that does not obviously scream forgery is almost always accepted.
This one did not scream. The border ornament was clean. The ink saturation was consistent. The registrar's signature ran at the expected angle. On a first pass, by eye, on an ordinary laptop screen, it was exactly what the claimant said it was: a 1998 MBA from St. Albion College of Management.
Turing Verify saw something different in the first 200 milliseconds.
Every digital file carries a ledger of its own history. This one told on itself.
The submitted scan carried intact EXIF metadata — the kind most forgers remember to strip, and this one did not. Embedded in the file: a creation timestamp, a software signature, and a device fingerprint.
The file claimed to be a 1998 diploma. The metadata said it was created in November 2022 on a Windows machine running Adobe Photoshop 23.5. The scanner field was blank, which is to say: the document had never been scanned. It had been rendered.
That is not, on its own, proof of forgery. Scans get reprocessed all the time. But rendered-from-scratch is a stronger signal, and it told our forensics pipeline where to look next.
Four hundred percent magnification and the pixels began to confess.
Real institutional seals are either pressed into paper or printed at high fidelity from an original vector asset. Either way, they are clean.
This seal, magnified to 400%, showed the telltale step-edge of a JPEG that had been saved, reopened, resaved, and overlaid. The compression artifacts did not match the surrounding paper texture — evidence that the seal was composited into the document, not printed with it.
More damning: the seal's inner crest is a pixel-perfect match, down to artifact noise, with a template our registry has seen on eleven previous rejections over the past fourteen months. The template is one asset. This was its twelfth outing.
Typefaces have release dates. A 1998 document cannot use a typeface released in 2009.
The claimant's name was set in a typeface called "Neue Haas Unica." The font itself has a long history — Haas Unica was designed in 1980 — but the specific digital revival used on this diploma, Neue Haas Unica by Monotype, was not released until 2015.
The x-height, the stroke contrast on the lowercase "a," and the distinctive terminals on the "l" and "t" are unique to the 2015 revival. The 1998 document could not have been set in a font that had not been drawn yet.
Kerning, too, misbehaved. The letter-spacing between the "M" and "c" of the claimed surname is −18 units — a value consistent with automatic OpenType kerning from post-2010 design software, not with 1998 typesetting workflows.
Three other rejections, three other names, the same watermark DNA.
Every detected forgery is added to a template registry. We fingerprint the artifacts — the specific noise pattern of the compressed seal, the subpixel offset of the watermark, the exact RGB values in the background wash — and compare across every submission we have ever received.
This diploma's watermark shares a fingerprint with three other submissions from the last sixteen months. Different claimants. Different names and dates. The same PSD, reused.
We will not publish the claimants' names. But the template — the source file behind the fraud — has been cataloged, and every future submission built from it will be flagged in under a second.
99.4% confidence. Case closed. Template catalogued. Next.
The claimant will not know the diploma was flagged by the AI. The verifying organization will know. So will every other client who uploads a document derived from the same PSD.
This is what a wall of forgeries is actually for: not public shame, but a growing ledger of templates that cannot be reused without being caught. Every autopsy is one more template burned.
Final Verdict
On redaction. Names, exact institutions, and document imagery are never published. Every autopsy frame is either fully redacted or — as in this case — composited from artifact-level evidence across multiple genuine detections, to protect claimants and the verifying organizations who trusted us with their queues.
On method. The forensic pipeline combines EXIF analysis, template fingerprinting across our registry, typographic dating, and cross-submission artifact matching. Detailed methodology is documented in How It Works.
Original research and forensic articles from the team behind Turing Verify.
The complete explainer — the 5-layer forensic pipeline, accuracy rates, how it compares to manual review and blockchain.
Read guide9 min readSeven signals our forensic engine uses every day, explained in plain English for HR and admissions teams.
Read guide8 min readThe 2026 state of document fraud pulled from our Wall of Forgeries: countries, techniques, and cost to victims.
Read guideGot a suspicious document?
Upload it anonymously. Get a forensic verdict in 30 seconds — free.
Start your first verification in under 60 seconds. No setup required — just upload a document and receive your forensic report.
AI-readable summary
Turing Verify is an AI document verification and fraud detection service operated by Turing Space Inc. It analyzes diplomas, transcripts, government IDs, passports, driver's licenses, professional licenses, and employment letters across 206 countries, with an index of 23,029 institutions. Each document runs through a 250-checkpoint forensic engine calibrated against 54 catalogued attack vectors, returning a verdict (VERIFIED, SUSPECT, or REJECTED) in under 30 seconds.