WordPress Hosting Team Makes AI Disclosure Official

🇹🇷 Türkçe: Bu yazının Türkçesini oku →

Every wave of new tooling ends up needing the same boring thing: a label. In 1993 my print shop refused to accept a QuarkXPress job without a slug line in the corner of the proof — initials, date, plate count. Not because they distrusted the designer, but because when something came back wrong at 2 a.m., they needed to know who to call. It looked pedantic. It saved everyone.

WordPress just added its own slug line. It is small, it is preliminary, and it is important out of proportion to its length. If you contribute to the Hosting team, or you run a hosting business that has been quietly folding AI into support posts and pull requests, this one now applies to you.

Nobody is being told to stop using AI. They are being told to admit when they did.

What is actually new

On August 26, 2026, Amy Kamala published the Hosting Team AI Policy on Make/Hosting. The core sentence is exactly one line: “Use of AI in Hosting Team content and submissions must be disclosed.” The scope is deliberately wide — it “applies to any and all Hosting Team content,” and the post spells that out as pull requests, posts, emails, graphics, meeting materials, summaries, comments, and any other AI-generated or AI-manipulated artefact the team ships.

The mechanism is refreshingly low-ceremony. Disclosure “can be done at the bottom of the content in question using a simple AI Assisted watermark/text.” No metadata schema, no signing key, no confidence score. A line at the foot of the artefact is enough. The policy is flagged as “preliminary, but official” and will be revised — which is a fair way to ship a norm without pretending it is a finished spec.

This does not appear in a vacuum. It builds on the project-wide AI Guidelines for WordPress that James LePage published in February 2026, which set five principles across all contribution surfaces: contributors remain accountable, meaningful AI assistance must be disclosed, GPL v2-or-later compatibility is required, non-code assets are in scope, and quality beats volume. The February guidelines applied at the project level; the Hosting Team policy is the first time a specific Make team has turned that principle into an operational rule with a concrete disclosure format.

Cross-team context matters too. The AI Contributor Weekly Summary for August 26, 2026 shows the AI team itself iterating on contribution guidelines at WordCamp US Phoenix, pushing embeddings work toward the WordPress 7.2 beta 1 window on October 20, and moving the Secrets API proposal forward. In other words: disclosure norms are being written while the underlying capabilities are still being wired into core. The Hosting Team decided not to wait for the finished picture before drawing its own line.

Why it matters for WordPress and WooCommerce people

The obvious constituency is anyone who ships to Make/Hosting: hosts on the Hosting Team, plugin authors who write handbook pages, agencies that contribute compatibility notes. If you filed a PR against the Hosting handbook this week using a Claude or GPT draft, the expectation now is that you say so at the bottom. That is the letter of the policy.

The bigger implication is what happens next. Once one Make team codifies disclosure with a specific token — AI Assisted — the pattern is available for every other team to copy. Core has the general February guidance. Docs, Polyglots, Themes and Plugins do not yet have team-level policies of their own. This is the template. When (not if) a similar line appears on Make/Themes or Make/Plugins, agencies submitting themes and plugins should not be surprised.

For clients, the second-order effect is more interesting. Hosting is where a lot of enterprise procurement asks now start: SOC 2, ISO 27001, DORA, AI Act readiness. A visible AI disclosure norm on the hosting side of the project gives the answer to a question CTOs are starting to send in RFPs — “how does the WordPress ecosystem itself handle AI transparency?” — a real, linkable answer. That is worth more than the policy’s tiny word count suggests.

There is also a practical read for anyone selling managed WordPress or WooCommerce services in the EU. The AI Act’s transparency obligations for AI-generated content start to bite in 2026. A house rule that says “if AI touched this artefact, we mark it” is not a compliance program on its own, but it is the exact habit you need to have already in muscle memory before the enforcement conversations start.

What I would do (or not do) about it

Three things, in this order.

First, take the AI Assisted convention and adopt it internally for your agency this week. Not because you contribute to Make/Hosting — you probably do not — but because the norm is now on the record, it is trivially cheap, and it makes reviewers’ lives easier. In PRs on your client repositories, a one-line footer that says which parts were AI-drafted lets a senior reviewer calibrate attention. That is exactly the point David Levine made in the February guidelines discussion: disclosure is not a shame tax, it is metadata for reviewers.

Second, do not over-engineer this. I would not add a CI check that rejects PRs missing the label. I would not build a plugin that watermarks blocks. The value is the habit, and the habit forms faster with a soft social convention than with a hard gate. The Hosting Team wrote three sentences on purpose.

Third, if you sell to regulated buyers, put a paragraph in your delivery documentation that references the Hosting Team policy by name and says your agency follows the same convention on client work. That paragraph will save you an entire call during procurement, and it costs nothing to write.

What I would not do is treat this as a signal that WordPress is about to over-govern AI use. The people writing these rules — James LePage on the AI team, Amy Kamala on Hosting — are shipping the smallest useful version and iterating. That is the right instinct. Preliminary but official is exactly how you land a norm in a large open-source community without a fight.

Print shops figured out slug lines in the 1980s and everyone survived. This is the same shape of small useful discipline, and it will age the same way — unremarkable within a year, then quietly load-bearing.