Declare Once, Work Everywhere

2 June 2026 · Tom Cranstoun · 5 min read

Every machine reading your content right now is guessing. It guesses what your pages sell, whether they're sensitive, how much to trust them, and if acting on them is worth the risk. It does that work fresh, on every page load, independently of every other machine doing the same thing on the same content. None of those guesses are yours to see or correct.

We change that.

The Problem

AI agents, browsers, search systems, and assistants are reading your content to serve their users. That isn't the problem. The problem is that they do it by inference, and that process is private. Chrome carries internal lists - built into the browser binary, updated without notice - that decide whether an agent may act on your pages. A shopping model chops each page into hundred-word chunks and judges if you trade. Assistants form their own read of your credibility and quote accordingly. None of it's disclosed, none of it's contestable, and none of it reflects what you have said about yourself.

The result is a web where the decisions about your content are taken without you. When they're wrong, there's no one to answer for it.

The Inflection

Two things happened in June 2026 that mark where this is heading.

Salesforce paid to acquire Contentful. The stated reason is agent infrastructure: content assembled by machines, delivered across channels, managed inside one platform's stack. Agent-readiness is now a feature you purchase and a register you join. Withdraw from the platform and the signal leaves with you.

In the same month, researchers reverse-engineered the classification data Chrome ships alongside itself. More than thirty thousand domains gated from AI assistant actions. Another thirty thousand selected for semantic indexing. A live model deciding page by page whether you count as commerce. All of it written by one vendor, for that company's purposes, with no route to read or contest your entry.

The market has confirmed that machines need structured signals about content. The question it hasn't answered is who writes those signals, and whether the site has any say.

What We Do

We make content that manages itself.

A COG is a carrier-neutral record that travels with the file: what it is, who produced it, when it changed, how machines should treat it. The format is open, governed by The Gathering, published under a public licence. Any system can read it. No vendor owns the schema.

REGINALD adds the trust layer. An Ed25519 signature over the canonicalised bytes lets any reader confirm the record is genuinely yours and hasn't been altered in transit. The signature covers provenance and integrity. It doesn't make claims about editorial truth - that judgement stays with you and your readers. It gives the record an author and a date, so a machine that reads it knows whose word it's taking.

The pair - COG plus REGINALD attestation - means your content introduces itself to every system that encounters it. Not through a vendor's API, not through membership in a list you can't see, but through a record embedded in the file.

What You Get

Content that's machine-readable on your terms, not inferred on someone else's. When an agent reads your page and misrepresents it, you have a signed, dated statement of what it actually says - and you have an accountable named party, which is more than the inference has.

Content that travels. Platform lock-in is the dependency on a vendor to explain your content for you. A COG removes that dependency. The record is in the file. Move the file and you move the record.

Evidence for governance. The AI Act, accessibility regulations, and digital-records law are moving in the same direction: disclosed, attributable, auditable decisions. A structured, signed declaration is what those regimes expect. We build it. We don't make you compliant - that's a legal duty - but we produce the documentation that compliance requires.

How to Start

The fastest entry is an MX audit. We assess your current content against the machine-readability standard, score it against what agents and browsers actually resolve, and tell you exactly what to fix and in what order. Most teams find the gap is narrower than expected: the structured content work their CMS already forces covers the bulk of the process. What's missing is the record that travels with the file.

From there we work with your team to implement COG metadata, connect REGINALD attestation for the records that carry regulatory weight, and build the content infrastructure that no acquisition can take away.

Write to us at info@cognovamx.com or see what we offer at our services page.