Why collectors are about to demand provenance files

The acquisitions counsel at a London gallery moving a $42,000 digital piece in February 2026 sent a one-paragraph email to the artist before the wire cleared. The paragraph asked for a manifest. Not a certificate of authenticity, which is a paper convention from another century. Not a wallet token, which the gallery's counsel had stopped accepting as evidence by 2023. A structured file recording who signed the work, when, with which protocol, and against what content hash. The artist did not have one. The wire still cleared, with an additional clause in the bill of sale acknowledging the gap. The clause was specific: this work was acquired absent a creator-side provenance manifest.

The gallery's counsel sent the same clause four times in the next eight weeks. By April they had a standard template.

What are acquisitions teams actually asking for?

The artifact has converged across the half-dozen galleries that began requiring it in 2025. It is a JSON sidecar attached to the work, structured according to either the C2PA specification (the Coalition for Content Provenance and Authenticity, ratified by Adobe, Microsoft, the BBC, and a dozen other parties between 2021 and 2024) or a comparable open schema that resolves to the same fields. The fields are:

The creator's signed attestation, generated with a key the creator controls. The content hash of the work at the moment of signing, computed by SHA-256 against the file bytes. The timestamp, recorded against a trusted clock. The protocol or device identifier that performed the signing operation. References to any subsequent modifications, each carrying its own signature and timestamp.

The acquisitions team does not need to understand the cryptography. They need to be able to verify, in front of an art lawyer, that the work was signed by the named creator on the date claimed. The manifest is the artifact the lawyer's paralegal can check by querying a public registry. The query returns either a confirmation or a finding. The acquisitions team makes a decision on the basis of the query result.

Why is this becoming standard now, and not in 2018?

Two changes converged. The first is the volume of generative content entering primary sale. Galleries that moved one or two AI-adjacent pieces a year in 2022 are now seeing twenty or thirty in a quarter. The 2018 due-diligence process was tolerable when the failure mode was rare and the failure cost was containable. At current volumes, the failure cost is not. A 2025 lawsuit against a gallery that acquired and resold a work the original creator subsequently disputed cost the gallery roughly $180,000 in legal fees, plus the unwind cost of the original sale.

The second is the regulatory floor. The EU AI Act's Article 50 transparency obligations, with operative force from August 2026, require providers of certain AI systems to mark synthetic outputs in machine-readable form and to make the marking legible to users. Galleries selling into European jurisdictions do not bear the obligation directly, but their counsel reads the regulation as an indicator of which side of a future dispute the regulatory consensus will land on. A gallery that has acquired and sold a synthetic work without provenance is exposed in a way it was not in 2022.

The two pressures together are the reason the request appeared roughly synchronized across acquisitions teams in the second half of 2025. It was not coordinated. It was the same set of inputs hitting the same set of senior counsel within a four-month window.

What does an artist actually have to do?

The operational change is small. The publication workflow that signs each finished piece before posting is a five-second addition. The signing step generates the manifest, anchors it to a registry, and produces the receipt the artist can hand to a gallery on request. Most working illustrators in 2026 have already integrated this step into Photoshop, Procreate, or Affinity via plugin support added in late 2024.

The manifest is not separate from the work. It travels with the file. A gallery that receives the work file receives the manifest along with it, by default, in the file's metadata. The acquisitions team's verification step is a single registry lookup. If the query returns a confirmation, the acquisition proceeds. If it returns a finding (the signature does not match the claimed creator, the timestamp is inconsistent with the publication record, the registry has flagged the entry), the acquisition is paused and the artist is notified.

The artist's part is the signing. The gallery's part is the verification. The registry's part is the third-party witness that neither party can manipulate.

What is the cost of not producing a manifest?

Not catastrophic, today. A working artist without a manifest can still sell. The cost is differential. Galleries that require manifests offer slightly better pricing, slightly faster commission cycles, and a measurably reduced rate of post-sale disputes. The galleries that do not yet require manifests will, on the timeline the acquisitions data suggests, within eighteen months. By the end of 2027 the artist without a manifest is selling to platforms that auction, not to galleries that buy. The two routes are not the same career.

The shift is happening underneath the conversation, in the legal departments of acquiring institutions, on a timescale that does not announce itself on the front page of trade publications. By the time it is visible in headline revenue data, the artists who adapted are in a different business than the artists who did not. The manifest is cheap to produce. The career consequence of not producing one is not.

Sign the work, anchor the signature, hand the receipt to the gallery on request. The work product is the same as it has been for a century. The acquisitions counsel is just asking, finally, for the same chain of custody they have always asked for on the physical side. The chain is the product.