Standard Improvement Proposal (07.04.2026): Identity‑Layer Semantic Anchor for llms.txt (JSON‑LD Binding)

[marin-popov]

I’m sharing a related but structurally different concept that may be relevant to anyone working on standards or machine readable interfaces for LLMs.
On 07.04.2026, I published a protocol level extension for the llms.txt standard called Identity Layer Semantic Anchor.
The full specification and canonical files are here:
Specification (GitHub):
https://github.com/marin-popov/semantic-anchor
The proposal addresses the structural “Identity Gap” in llms.txt — the file describes content, but it does not prove who is making the declaration.
It introduces a single mandatory header that binds llms.txt to a canonical JSON LD identity resource:
Identity: https://1euroseo.com/identity.jsonld
Live reference implementation:
https://1euroseo.com/llms.txt
https://1euroseo.com/identity.jsonld
This mechanism provides:
• verifiable, machine readable provenance
• persistent entity coherence
• zero drift between declarations and canonical identity
• a structured identity node that LLM retrieval systems can resolve
• token efficient identity reuse across agents and tools
This is not a glossary, taxonomy, or conceptual anchor.
It is a protocol level identity binding that operates below content and terminology layers.
I’m sharing it here as related work, since both projects use the term “Semantic Anchor” but in different domains:
• this project → semantic anchors as conceptual vocabulary
• my proposal → semantic anchor as identity binding for llms.txt
If useful, I can provide a short comparison, interoperability notes, or a minimal example of how both concepts can coexist without collision.

[marin-popov]

Additional Notes on Conceptual Separation and Interoperability
For clarity, the two concepts operate at different layers:
• Semantic Anchors (this project): conceptual vocabulary, domain specific terminology, and prompt level context activation.
• Identity Layer Semantic Anchor (my proposal): protocol level identity binding for llms.txt using a canonical JSON LD node.
Because the scopes are orthogonal, both concepts can coexist without collision. If useful for documentation purposes, I can provide:
• a short comparison outlining the layer boundaries
• interoperability notes describing how conceptual anchors and identity anchors can be referenced independently
• a minimal example showing how an llms.txt file can include an identity anchor while still referencing conceptual anchors at the content layer
This is optional, but available if it supports clarity in the ecosystem.