My name is Hunter Hill, founder of Akaeon. We work on creator-rights and provenance infrastructure for the AI era, and recently published ADP-1, an open specification that adds an authority-binding and publicly verifiable timestamp layer on top of existing reservation protocols (TDMRep, aipref, C2PA). I'm raising this as a design consideration, not a request to change the core format.
As llms.txt adoption grows and the conversation around standardization continues, one gap seems worth surfacing early: the file as specified carries no portable proof of who authored or curated it, or when. Anyone who can write to a site root can publish or alter an llms.txt, and a third party reading it later has no independent way to verify the asserted authority or establish a tamper-evident timestamp. For a file that is increasingly treated as a site's own statement of intent to LLMs, that authority-and-timestamp layer is the missing piece — the same gap we found across the candidate solutions reviewed in the EUIPO TDM study (robots.txt, TDMRep, C2PA assertions, Spawning's registry, and others).
This does not require changing llms.txt's Markdown structure. The reservation/curation signal and the proof of who-made-it-and-when can live as separate, composable layers, exactly as ADP-1 treats them for the TDM opt-out protocols. I'd be glad to share the ADP-1 spec or contribute a worked example if it's useful to the discussion. Either way, flagging it now so it can be considered before any path toward formalization hardens the format.
With thanks for the work on this proposal.
My name is Hunter Hill, founder of Akaeon. We work on creator-rights and provenance infrastructure for the AI era, and recently published ADP-1, an open specification that adds an authority-binding and publicly verifiable timestamp layer on top of existing reservation protocols (TDMRep, aipref, C2PA). I'm raising this as a design consideration, not a request to change the core format.
As llms.txt adoption grows and the conversation around standardization continues, one gap seems worth surfacing early: the file as specified carries no portable proof of who authored or curated it, or when. Anyone who can write to a site root can publish or alter an llms.txt, and a third party reading it later has no independent way to verify the asserted authority or establish a tamper-evident timestamp. For a file that is increasingly treated as a site's own statement of intent to LLMs, that authority-and-timestamp layer is the missing piece — the same gap we found across the candidate solutions reviewed in the EUIPO TDM study (robots.txt, TDMRep, C2PA assertions, Spawning's registry, and others).
This does not require changing llms.txt's Markdown structure. The reservation/curation signal and the proof of who-made-it-and-when can live as separate, composable layers, exactly as ADP-1 treats them for the TDM opt-out protocols. I'd be glad to share the ADP-1 spec or contribute a worked example if it's useful to the discussion. Either way, flagging it now so it can be considered before any path toward formalization hardens the format.
With thanks for the work on this proposal.