Art authentication asks a difficult question: is this work really what it is claimed to be?
The answer cannot safely rest on one signature, one expert opinion, one laboratory test or one old receipt. Authentication is strongest when independent evidence layers converge.
Claim → Provenance → Visual / Stylistic Evidence → Archival Evidence → Material / Scientific Evidence → Contradiction Check → Attribution Judgment → Revision
AI Extraction Box
Definition: Art authentication is the evidence-based process of evaluating whether an artwork’s claimed maker, date, origin and identity are supported by provenance, documentation, visual analysis and technical examination.
Boundary: Authentication owns the identity claim. Provenance owns custody history in How Art Provenance Works. Market implications are in How Art Markets Work. Museum-specific attribution remains with the Museum branch.
1. Authentication Is Not a Single Test
No universal machine can scan an artwork and output “genuine” or “fake”. Different works require different evidence.
A secure judgment is built from converging clues rather than one dramatic result.
2. The Claim Must Be Defined Precisely
“By the artist”, “from the workshop”, “attributed to”, “school of”, “after” and “in the style of” are not equivalent claims.
Authentication begins by asking exactly what level of authorship is being asserted.
3. Provenance Is One Evidence Layer
A continuous custody chain from the artist’s studio or first documented owner can strongly support authenticity.
But provenance can contain mistakes or forgeries, so the object and the paperwork must be checked against one another.
4. Archival Evidence Can Anchor the Work
Letters, invoices, exhibition catalogues, photographs, studio records and inventories may describe the work before current ownership.
The strongest records identify dimensions, subject, date or image clearly enough to match the object.
5. Connoisseurship Is Trained Comparative Judgment
Experienced specialists can recognise recurring habits in drawing, composition, brushwork, proportion, handling and material choice.
This is not mystical intuition. It is pattern recognition built from extensive comparison.
6. Connoisseurship Is Powerful but Fallible
Experts can disagree, and famous experts have changed their minds. Visual judgment should therefore be tested against independent evidence whenever possible.
Expertise reduces uncertainty; it does not abolish uncertainty.
7. Materials Create a Date Boundary
If a claimed old painting contains a pigment or synthetic material invented much later, the chronology becomes impossible unless later restoration explains it.
Material analysis is especially good at ruling out impossible dates.
8. Scientific Testing Is Usually Stronger at Exclusion Than Attribution
A test may show that a canvas, pigment or paper is consistent with a period. That does not prove one particular artist used it.
Consistency narrows the field. It does not automatically identify the maker.
9. Imaging Can Reveal Hidden Construction
Infrared, ultraviolet, X-ray and other imaging techniques can reveal underdrawing, revisions, repairs, earlier compositions and material differences.
The hidden working process can be compared with known studio practice.
10. Pentimenti Can Support a Genuine Creative Process
Changes made during composition may indicate active invention rather than mechanical copying.
But revisions alone do not prove authorship; skilled forgers can imitate evidence of process.
11. Supports and Grounds Carry Evidence
Canvas weave, panel construction, paper fibres, preparation layers and fasteners can all provide chronological or workshop information.
Authentication often depends on parts viewers rarely notice.
12. Signatures Are Evidence—but Highly Vulnerable Evidence
Signatures can be genuine, added later, copied, forged or transferred conceptually from another work.
A signature should be studied as part of the object, not treated as a magical certificate.
13. Certificates Can Also Be Wrong
A certificate of authenticity is only as reliable as the authority, evidence and chain behind it.
Paper can certify a claim; it cannot make a false claim true.
14. Catalogue Inclusion Is Important but Revisable
Inclusion in a respected catalogue raisonné can strongly influence attribution. Yet catalogues are historical knowledge systems and can be corrected.
Authority should remain open to evidence.
15. Workshop Practice Complicates Attribution
Masters and assistants may work on the same object. A design may originate with the master while execution is distributed.
This is why attribution language often has several levels rather than a binary genuine/fake split.
16. Copies Can Be Historically Authentic Copies
A work can be genuinely old and genuinely valuable while still being a copy after another artist.
Authentication asks whether the object is correctly described, not whether every authentic object is an autograph masterpiece.
17. Restoration Can Confuse Testing
Later retouching, lining, varnish and replacement components may introduce newer materials into an old object.
Technical results must therefore distinguish original structure from intervention history.
18. Forgers Study Authentication Systems
Once experts reveal which clues they value, sophisticated forgers can attempt to reproduce those clues using old materials, false records and simulated ageing.
Authentication is therefore an adversarial evidence problem.
19. Old Materials Can Be Used in New Forgeries
An old canvas or period paper can remove one obvious chronological contradiction.
This is why no single material date should carry the entire decision.
20. Contradictions Matter More Than Convenient Matches
A dozen features may look plausible, but one impossible pigment date or documented contradiction can outweigh them.
Strong authentication actively searches for disconfirming evidence.
21. Digital Authentication Has Different Failure Modes
Digital works can be duplicated perfectly at file level. Identity may therefore depend on hashes, signatures, creation records, authorised editions, licences and custody systems.
The question shifts from “Is this physical object original?” toward “Is this instance authorised and traceable?”
22. Metadata Is Useful but Editable
File metadata can record dates, software and authorship claims. It can also be modified.
Digital evidence requires chain-of-custody thinking just like physical evidence.
23. AI-Generated and AI-Assisted Works Add Provenance Questions
Authentication may need to establish model, version, prompts, source assets, editing history and final authorisation.
The relevant identity can lie in the workflow as much as the final pixels.
24. Authentication Changes Market Risk
Attribution can radically change price. This creates pressure around authentication decisions and makes independence important.
The market route is How Art Markets Work.
25. Authentication Also Changes History
Reattributing a work can change our understanding of an artist, workshop, period or network.
Identity claims are therefore scholarly claims, not merely commercial labels.
26. Uncertainty Must Be Expressed Honestly
“Attributed to”, “probably by”, “workshop of” and “after” preserve meaningful confidence differences.
False precision is worse than explicit uncertainty.
27. Failure Modes
- Signature capture: a name on the object overrides contradictory evidence.
- Science capture: one technical test is treated as complete attribution.
- Expert capture: authority substitutes for an evidence trail.
- Provenance capture: documents are trusted without checking the object.
- Confirmation bias: only evidence supporting a valuable attribution is pursued.
- Binary thinking: nuanced attribution levels collapse into genuine/fake.
28. A Practical Authentication Test
- What exactly is being claimed?
- What is the provenance?
- What archival records independently support the claim?
- Does style align with secure comparison works?
- Are materials chronologically possible?
- What does imaging reveal about construction and revision?
- What evidence contradicts the claim?
- What attribution level best matches the total evidence?
- What remains unknown?
29. The Deeper Principle
Authentication is not a hunt for one magical proof.
It is the disciplined convergence of independent evidence around an identity claim—combined with the willingness to keep uncertainty open when the evidence does not close it.