On 2026-03-19, schema.org shipped release v30.0, and with it something the physical-supply-chain and web worlds had never had: formal equivalence annotations between GS1's vocabulary and schema.org's. If you run digital for a brand whose identifiers were born in the barcode world, this is the release where those identifiers became web citizens. It is also the release that makes the remaining gap impossible to unsee.

What v30.0 actually shipped, counted

The claims here are counted from the published schemaorg-all-https.ttl file, not paraphrased from a blog post:

  • 1 owl:equivalentClass annotation — gs1:Organization declared equivalent to schema:Organization.
  • 15 owl:equivalentProperty annotations — GS1 organization-vocabulary properties formally mapped to their schema.org counterparts.

Two corrections belong beside those counts, because a GS1-literate reader will check:

  1. hasGS1DigitalLink was not made core in v30.0 — it carries schema:isPartOf: https://pending.schema.org, meaning it remains pending.
  2. hasGS1DigitalLink is not new in v30.0 — it has shipped since v27.0.

The news in v30.0 is the equivalence weld, and only the equivalence weld. Counted, dated, checkable.

Why "welded at the company grain" matters

An equivalence annotation is a strong statement. It says a GS1 Organization — the thing behind a GLN, the party on an EPCIS event — is a schema.org Organization, the thing behind a knowledge-graph entry, a search result, an answer an agent assembles. One identity fabric, two vocabularies, formally joined.

Joined, precisely, at the grain of a company. And that is exactly one level too high for the question your traceability program actually needs answered.

EPCIS 2.0 §7.2.2 defines five event dimensions — what, when, where, why, how — and no performer is among them. The standard's party fields are organisation-grain: CBV 2.0 §8.7.1 defines the PGLN, a party identifier that names a company, never a person or an agent. So the stack, end to end, can now say with formal web-grade semantics that a company did a process. What it structurally cannot say is that a worker — this person, this software agent, this robot — did this task, at the moment of this scan. There is no field to fill in better. That is a structural absence, not a data-quality problem, and it is stronger for being checkable: EPCIS 2.0 §7.2.2 · CBV 2.0 §8.7.1 — ten minutes, the specs are public, don't take our word for it.

The seam, made visible

Before v30.0 you could argue the whole identity layer was loose enough that the missing grain was one gap among many. The weld removes that excuse. The company grain is now solved twice — once in GS1's vocabulary, once in schema.org's, with a formal bridge between them. When both vocabularies agree perfectly at one grain, the grain neither of them carries stops looking like noise and starts looking like the seam.

For your brand this lands in two ways. First, your identifiers — GTINs resolved through Digital Link, GLNs naming your organization — participate in the same graph the web's answer engines read. The class-grain versus instance-grain split that your data-pool feeds already struggle with is covered in GDSN vs. EPCIS; v30.0 raises the stakes on getting it right. Second, the unit-level signal your team wants — which lots got scanned, in which markets, how long after they shipped — depends on event records, and event records are only as attributable as their who. The record we keep carries the two grains the standard omits: who, the attested observer — human, agent, or embodied agent — and capturedBy, the warrantor account that stands behind the capture. They are different parties with different failure modes, and they are never collapsed into one field. The observer grain itself — agent, human, thing — is the problem id.org.ai exists to name.

The watch-list

What would close the seam from the standards side? An equivalence — or any formal vocabulary at all — at the worker-and-agent grain: a way to say, with the same web-grade semantics v30.0 gave companies, that this identified person or agent performed this event. Watch three things: whether hasGS1DigitalLink graduates from pending to core; whether any future release annotates below the organization class; and whether EPCIS's own extension mechanism (the conformant path for carrying a performer today) acquires standardized vocabulary. Until then, the seam is open, the company grain is welded shut, and the difference between the two is where your event layer either answers or doesn't.

The decision that puts a performer-complete event layer behind your marks belongs to the same programme office deciding your 2D artwork — the Sunrise 2027 brand-owner read covers that boundary. When you want the read for your own catalog, get started. The interview branches for brand owners, and its questions are past-tense facts: your retailer questionnaires, your last launch's first thirty days, the dated items on your 2026–2027 plan.