TL;DR: Yes, AI can make a fake certificate or diploma that looks perfect, and looking at it will rarely tell you otherwise. A TRUE Original document carries a fingerprint anchored on a public blockchain on the day it was issued, so an AI-made copy or an edited version fails the check in TRUE Verify, which is free and needs no account. The check does not care how good the fake looks: it asks whether the document matches what the issuer anchored.
Yes. Generative AI can produce a certificate or diploma with the right names, dates, logos, seals and signatures, in the layout of a real issuer, and the result can pass a glance from a trained eye. Professional Security magazine, reporting in December 2025 on a webinar by Hedd, the UK degree verification and fraud service, wrote that "artificial intelligence is making fake certificates such as diplomas faster and cheaper to make, and more convincing than ever." (Professional Security, 8 December 2025) An AI-made copy of a TRUE Original document still fails the check in TRUE Verify, the free service that compares a document with the fingerprint its issuer anchored on a public blockchain.
The quality is improving rapidly. What looked obviously fake two years ago now passes casual inspection. What passes casual inspection today may pass expert inspection tomorrow.
The turning point was text. In March 2025 ChatGPT gained an image generator that, in TechCrunch's words, is "a lot better at generating text inside images". By the end of that month people were posting restaurant receipts it had made. One of them was Deedy Das, an investor, who posted a fake receipt for a real San Francisco steakhouse:
"You can use 4o to generate fake receipts. There are too many real world verification flows that rely on “real images” as proof. That era is over." (Deedy Das on X, 29 March 2025, reported by TechCrunch, 31 March 2025)
A certificate is a receipt with better paper. Everything on its face can now be generated:
A fake made this way is not a copy of one certificate. Each one can carry its own name, date and serial number, so there is no duplicate to trip a pattern check.
A human forger might produce a few high-quality fakes per week. AI systems can generate thousands per hour, each one unique and convincing. The economics have inverted: generating fakes is now cheaper than verifying them manually.
Employers already report them. In a YouGov poll of more than 500 HR decision makers for Hedd in April 2025, 67% of large companies said they had seen an increase in job application fraud and put it down to AI tools used to enhance or fabricate experience or qualifications. In the same poll, 45% of large companies had caught an applicant giving false information about their qualifications.
Checking lags behind. According to the same 2025 Hedd survey, 52% of large businesses check all academic credentials, against 37% of medium-sized firms and 29% of small businesses, and 26% of small companies do not check qualifications at all (Prospects, July 2025).
Chris Rea, who leads the Hedd qualification fraud service, put it this way:
"The rise of AI-generated CVs makes it easier than ever for candidates to misrepresent their qualifications and it can be harder for employers to spot." (Chris Rea, Hedd, July 2025)
A well-made AI-generated fake certificate usually cannot be spotted by eye, so the reliable test is to check the certificate against a record the issuer created when it was issued. For a TRUE Original document, that record is a fingerprint anchored on a public blockchain, and TRUE Verify compares the two in seconds. Hedd's 2025 advice to employers was blunt: "Simply asking to see a paper or digital qualification certificate is not enough."
Early fakes do slip. When TechCrunch tested the new ChatGPT image generator, its fake receipt had "a couple of dead giveaways": the total used a comma instead of a full stop, and the maths did not add up. TechCrunch also wrote that "it wouldn’t be hard for a fraudster to quickly fix a few of the numbers".
The older tells fail for the same reason. Trained eyes once caught forgeries by misaligned text, the wrong font or an off colour, and AI output does not have to carry any of them. Holograms and embossed seals protected paper; a digital image can reproduce what they look like. A lookup in the issuer's database helps only if the verifier reaches the real database, and a phone call to the issuer is slow and depends on finding the real number.
A few signs are still worth a second look:
Each sign raises a question. The check answers it.
To check if a certificate is real, compare it with a record the issuer controls instead of judging how it looks. For a TRUE Original document, paste the link, the document ID or the PDF into TRUE Verify, or scan its QR code, and TRUE Verify checks it against the blockchain anchor. TRUE Verify is free, needs no account and returns a security level in seconds.
Take a hiring manager in Leeds with a cybersecurity certificate attached to an application. She drops the PDF into TRUE Verify. If the document is a TRUE Original document and nothing in it has changed, TRUE Verify finds the blockchain record and rates it Secured. If the name or the date has been edited, the fingerprint no longer matches.
She does not have to email anyone.
TRUE Verify checks documents against blockchain anchors, embedded digital signatures, issuer databases and more than 80 external credential platforms, and the level it returns reflects the proof it finds. A document that exists only in an issuer's database is rated Unsecured, with a high risk of fabrication, and a document with no proof at all is rated No Origin. For a certificate from an issuer with no such record, contact the awarding organisation through contact details you look up yourself, never the ones printed on the document. More on this in how to check if a certificate is real.
Blockchain stops fake certificates from passing as real by giving every genuine document a fingerprint that is anchored on a public ledger at the moment of issue. A forged or edited document does not match any fingerprint the issuer anchored, so the check fails. Blockchain does not stop anyone from making a fake image; it makes the fake fail the check. SSF Stöldskyddsföreningen, which has worked on crime prevention since 1934, issues blockchain-secured certificates through TRUE Original, and anyone who opens the link or scans the QR code gets the answer in seconds without calling SSF.
The fingerprint is a SHA-256 hash, a 64-character code calculated from everything in the document. The US National Institute of Standards and Technology defines it in the Secure Hash Standard:
"Any change to a message will, with a very high probability, result in a different message digest." (NIST, FIPS 180-4, Secure Hash Standard)
When an organisation issues a TRUE Original document, four things happen:
Checking runs the same arithmetic backwards. Anyone who holds the document can do it, years later, without the issuer's help:
TRUE is blockchain-agnostic. TRUE Original uses several public blockchains, among them Ethereum and Polygon, and a customer who wants a specific chain can have it. You can read more about how blockchain certificates work.
AI can fake the look of a blockchain-verified certificate, but the copy does not carry a fingerprint that matches a record the issuer anchored, so TRUE Verify finds no matching proof. To pass, a forger would need a document whose SHA-256 fingerprint equals one already on the chain, which NIST describes as computationally infeasible, or would need to rewrite a public blockchain.
AI can't modify blockchain records. Blockchain entries exist across thousands of independent computers worldwide. To alter a record, an attacker would need to simultaneously modify copies on more than half of these systems, a mathematical impossibility.
The check shows two things. The document is the one that was anchored, and it has not changed since the moment it was anchored. A TRUE Original document also opens on the issuer's own domain, so the reader sees whose record it is.
Time is the forger's other problem. A fingerprint written to a blockchain carries the moment it was written, and nobody can go back and add one to the day a diploma was awarded.
Yes. A QR code is only a link, and a forger can print one that leads to a lookalike verification page. The defence is to check where the link goes: a TRUE Original document opens on the issuer's own domain, and TRUE Verify's QR scanner and paste field run the check on TRUE's side, not on a page the forger built.
AI can't spoof the verification process. The verification system checks against the actual blockchain, not against a database that could be compromised. The blockchain itself is the source of truth.
The same goes for verification sites an AI can mock up in minutes. Type the address of the issuer or of TRUE Verify yourself instead of following a link the document hands you. Make it a rule: no certificate is accepted before it has been checked.
No. A blockchain anchor proves that a document existed in a given form at a given time and has not changed since, but by itself it has no status under the EU eIDAS regulation. TRUE's article are blockchain timestamps eIDAS qualified timestamps? sets out the difference.
The two layers can sit on the same document. TRUE Value Software AB holds its own eIDAS qualified electronic seal certificate, and TRUE offers eIDAS-sealed university diplomas. TRUE Verify calls a document with both a blockchain record and an eIDAS signature "Signed + Secured", its highest level.
AI fake detection estimates whether a document looks forged; blockchain verification checks whether a document matches the record its issuer anchored. A detection model learns from fakes it has seen and returns a probability, while a blockchain check returns match or no match and works the same way however good the forgery is. TRUE Verify gives the second kind of answer: a TRUE Original document that matches its blockchain record is rated Secured, and a copy with one changed character finds no match.
| Question | AI fake detection | Blockchain verification |
|---|---|---|
| What it asks | Does this document look forged? | Does this document match the fingerprint the issuer anchored? |
| What it returns | A probability score | Match or no match |
| A new kind of fake | The model may need retraining | Same check, same answer |
| A genuine document with one edited date | May pass if the edit is clean | Fails: the fingerprint changes |
| Works on | Any image or PDF | Documents the issuer anchored when issuing them |
| Who can run it | Whoever has the model | Anyone, free, in TRUE Verify |
| False positives | Possible (good docs flagged) | None: binary verification |
| False negatives | Possible (fakes missed) | None: fakes cannot verify |
| Speed | Varies (model inference) | Sub-second |
| Scalability | Compute-bound | Unlimited |
| Maintenance | Continuous model updates | Minimal |
| Cost trajectory | Increasing (compute + expertise) | Decreasing (mature infrastructure) |
| Attacker adaptability | High (can probe and iterate) | Zero (mathematical barrier) |
Detection has built-in limits. A model only flags patterns it has learned, so a new generation technique can produce fakes it has never seen. It outputs a probability, so someone has to choose a threshold and accept that some genuine documents will be flagged and some fakes will pass. And a determined forger can test fakes against a detector until one gets through.
Processing overhead: Detection requires running inference on every credential, consuming compute resources that scale with volume.
AI has fundamentally tipped the balance. The cost and speed of generating fakes has dropped dramatically while quality has increased. Detection systems are now playing catch-up against an opponent that can iterate millions of times faster.
Blockchain verification doesn't have this problem. The mathematical properties that make it secure are absolute. No amount of AI advancement changes the fundamental impossibility of forging blockchain-verified credentials.
The blockchain check has no threshold to tune, and it covers every document the issuer anchors when issuing it.
Attack-proof verification: The verification process itself cannot be gamed. The blockchain is the source of truth.
Constant-time verification: Verification speed is independent of credential complexity or volume.
Security bodies, certifiers, training providers and a Swedish government agency issue blockchain-secured documents with TRUE Original, because their documents are checked by people who have never met them. TRUE Original has issued over 900,000 documents for 200+ issuers in 15+ countries.
SSF Stöldskyddsföreningen, Sweden's leading security authority for over 80 years, issues blockchain-secured credentials for security professionals. SSF Stöldskyddsföreningen has worked on crime prevention since 1934 and issues blockchain-secured certificates through TRUE that anyone can verify without contacting SSF (SSF case).
"It has been particularly important for us to be able to ensure that our Proofs of Education, in the form of Certificates and Diplomas are correct and secure." (Maria Dahlstedt, Program Manager, SSF)
OneMore Secure, a Swedish company that helps organisations control their cyber hygiene, issues certificates on its own domain, secured in the blockchain with TRUE. OneMore Secure certifies cybersecurity professionals who understand document security at a technical level. They chose blockchain credentials because they could evaluate the cryptographic guarantees. "As a cybersecurity company, it felt natural for us to work with blockchain secured certificates," OneMore Secure says in TRUE's OneMore Secure case.
"For us, it's only natural to collaborate with the player in secure document management that has the highest quality, and stands for world-class security." (Matti Olofsson, CEO, OneMore Secure)
Safe Cert Group certifies hotels and restaurants for safety and hygiene compliance.
"The certificate is not just a piece of paper... it is a symbol of a commitment to quality and safety that both guests and staff can trust." (Joachim Törngård, CEO, Safe Cert Group)
Bolagsverket (the Swedish Companies Registration Office) piloted blockchain for business registration documents. Bolagsverket, the Swedish Companies Registration Office, built Proof of Business with TRUE, a concept that turns official documents into records anyone can verify. "With the help of TRUE, setting up our concept Proof of Business took eight weeks," says Erika Sjödin, Innovation Officer at Bolagsverket. The concept led to a government assignment for a second phase covering licences, trademark certificates and company registrations (Bolagsverket case).
An organisation protects its certificates from AI fakes by issuing them as documents that a stranger can check against the issuer's own record without phoning anyone. With TRUE Original, every document is anchored on a public blockchain at issue, opens on the issuer's own domain and can be checked in TRUE Verify.
Book a demo and TRUE will show you your own certificates, designed, anchored and live on your domain.
An AI model can draw the seal, the signature and the QR code in seconds. It cannot go back to the day the diploma was supposedly awarded and put a fingerprint on the blockchain.
Yes. Generative AI can make a fake certificate or diploma with the right names, logos, seals and signatures that passes a visual check. A TRUE Original document is checked against a fingerprint anchored on a public blockchain, so an AI-made copy fails the check in TRUE Verify.
A well-made AI-generated fake certificate usually cannot be spotted by eye, so check it against the issuer's own record instead. For a TRUE Original document, paste the link, the document ID or the PDF into TRUE Verify, free and without an account.
Blockchain gives every genuine document a fingerprint that is anchored on a public ledger at the moment of issue. A forged or edited document does not match any fingerprint the issuer anchored, so the check in TRUE Verify fails.
Compare the certificate with a record the issuer controls instead of judging how it looks. For a TRUE Original document, paste the link, document ID or PDF into TRUE Verify or scan its QR code; TRUE Verify is free, needs no account and returns a security level in seconds.
No. A blockchain anchor proves that a document has not changed since it was anchored, but by itself it has no eIDAS status. TRUE Value Software AB holds its own eIDAS qualified electronic seal certificate, and a TRUE document can carry both layers.
Every organisation we work with started with the same question. Give us thirty minutes and we will show you what your own documents would be worth.
Book a callNot sure where to start? Let us help!
Trusted by leading organisations worldwide
