WordPress 7.1 Roadmap: AI, Notes, and an Honest RTC Note

Every release cycle, two documents matter to the people who actually run WordPress sites for a living. The first is the release squad post — who is steering the ship. The second is the roadmap — where the ship is pointed. We covered the squad on Wednesday. The roadmap landed at midnight today, and it is the most honest one I have read in a few cycles.

What I appreciate about this one is the caveat sitting halfway down the page: “what’s shared here is being actively pursued, but doesn’t necessarily mean each will make it into the final release of WordPress 7.1.” After 7.0 had Real Time Collaboration pulled at the last minute and the React 19 upgrade got reverted inside a point release, that kind of language is welcome. It is the difference between marketing and engineering.

So let me walk through what is actually on the table for August 19, what it means for agency work, and where I would be cautious.

What is actually new

Anne McCarthy, the 7.1 release lead, published the Roadmap to 7.1 on Make/Core today. The release is still aimed at August 19, 2026, aligned with WordCamp US in Phoenix, and the focus splits across collaboration, AI, customization, and a few quietly meaningful performance and developer changes.

On the AI track, the roadmap commits to streaming generation and embeddings support in the AI Client, expanded authentication for Connectors beyond plain API keys (username plus application passwords are explicitly mentioned), and a new Guidelines primitive — “a persistent, structured way to encode editorial rules into WordPress.” That last one is the interesting one for editorial teams: it is the closest thing core has proposed to a built-in house style layer. This builds on top of the client-side Abilities API that shipped in 7.0, which itself extended the server-side Abilities API from 6.9. The 7.1 work pushes Abilities toward queryable capabilities — useful when an AI agent or another plugin needs to ask “what can this site do?” without scraping a thousand hook references.

On collaboration, Notes pick up rich text, emoji reactions, suggestion-application, and cross-block threads. Real Time Collaboration itself stays in the “strategic decision pending” bucket — the post is honest that the team has not finalized scope or storage, which lines up with the outreach effort opened on June 3 and the fact that RTC was removed from 7.0 in May. Treat RTC in 7.1 as opt-in plumbing, not a finished feature.

Three new blocks are planned for core: Playlist (with waveform), Table of Contents, and Tabs. Styling gets interactive states (hover, focus, active) without writing CSS, responsive controls without writing CSS, customizable viewport breakpoints, and a sidebar panel that finally shows inherited styles. The admin gets a persistent omnibar across editors, the Site Editor respects admin color schemes, and a new Identity section consolidates site settings that have been scattered for years.

Media handling moves further client-side: HEIC, Ultra HDR, and GIF-to-video conversions in the browser, plus a freeform cropper in the media modal. This is consistent with the broader performance team’s push toward off-server image work, and complements the separate libvips backend proposal we covered Wednesday.

Two items that will not get press but should: extended Unicode support in email addresses, usernames, and slugs (already in trunk), and the speculative loading default moving from conservative to moderate eagerness when caching is detected. The 6.8 speculative loading post explained why core started conservative; this is the first time the default is being tightened, and it changes how prerendering behaves on real traffic. Finally, the React 19 upgrade — reverted in Gutenberg 23.3.2 on June 5 — is back on the 7.1 list with an experimental feature flag and a compat layer for already-released plugins.

Why it matters for WordPress and WooCommerce people

For agencies and in-house teams, a roadmap is only useful if you can plan against it. Here is what I read between the lines.

The AI work is no longer about demos. Streaming and embeddings in the AI Client mean you can build editorial assistants that feel like the ones content teams use in dedicated tools, but inside the editor where the content actually lives. The Guidelines feature is the one I would watch closest — if it ships, it gives every site a structured place to put house style, brand voice, and forbidden terms, and any AI integration on top of WordPress can read from it instead of being prompted from scratch. That changes how multi-author and multi-language teams operate.

Real Time Collaboration in 7.1 will not be the Google Docs experience anyone is imagining. The post is careful to say “strategic decisions pending,” and the outreach effort is still gathering signal. If you have a client who has been promised collaborative editing in core, push the expectation to “Notes plus suggestions, with RTC arriving incrementally.” That is honest, and it is what the project is actually shipping.

The speculative loading default change matters more than it sounds. Moderate eagerness prerenders on hover instead of just prefetching, and that means real backend load on pages users never visit. If your WooCommerce store runs custom add-to-cart logic, session bootstrapping, or analytics in PHP at page render, moderate prerendering will execute that code more often. The opt-in caching detection in 7.1 should soften that, but it deserves a staging pass before you let 7.1 hit production.

For developers, the React 19 upgrade returning with a compat layer is the right call. The core committers’ WCEU meeting recap already flagged how disruptive an unflagged React major bump would be for the plugin ecosystem. Doing it behind a feature flag with a compat layer means your blocks keep working while you migrate at your own pace. That is the kind of governance choice that WordPress, at its best, gets right.

What I would do (or not do) about it

If I were planning August deployments right now, I would do three things this week and skip a fourth.

First, install the WordPress Beta Tester plugin on a staging environment that mirrors production, switch the channel to Bleeding edge nightlies, and start running the existing acceptance test suite against trunk. Beta 1 is targeted for July 15, which gives you roughly four weeks of real testing before the freeze. That is the window where blocking issues get fixed; after RC there is far less appetite for changes.

Second, audit any custom blocks for React 19 readiness now, while the feature flag is still off by default. The compat layer will catch most cases, but legacy class components, deprecated lifecycle methods, and direct manipulation of findDOMNode will surface during the upgrade. Catch them on staging in July, not in production in September.

Third, for WooCommerce stores, pin a clear story with your client about speculative loading. If you run any cart or session logic on hover-prerenderable pages, plan to either gate it behind a real-user signal or add the "no-vary-search" header your endpoints expect. The 6.8 documentation explains the trade-off well; the 7.1 default change just makes it more pressing.

The thing I would not do is start selling collaborative editing as a 7.1 feature. The roadmap is openly undecided on RTC, and selling something the project itself describes as “strategic decisions pending” is the fastest way to set a client up for disappointment. Notes plus suggestions is a real, shippable improvement; sell that.

A roadmap with honest caveats and dated commitments is the kind of artifact a mature platform produces. The 7.1 one is exactly that. Now we test.