WordPress AI visibility: measure the gaps, then fix the posts from Claude

The WordPress.com connector is the hands. The vendor's tools reference groups its tools into facades: content authoring for posts, pages, media, categories and comments, a site tool for settings, statistics and plugins, and a site editor context tool for theme data. Each takes an action of list, describe or execute, and every write needs your confirmation.

Linkeddit is the eyes. Answer Radar asks the five engines your buying questions and records who is named, who is recommended and which sources are cited. Search records weekly Google ranks with the AI Overview above them and reads your Search Console queries. Compete tracks what competitors change. All of it comes with one plan at $60 a month.

Claude does the analysis and the writing. Linkeddit supplies data and writes no advice, so the customer's own assistant decides what to change.

The connector is WordPress.com's, hosted at public-api.wordpress.com/wpcom/v2/mcp/v1. The vendor documentation says it works on every WordPress.com site, free or paid, and that self-hosted sites connected through Jetpack reach the same server with the same tool catalog, without a Jetpack plan. The documentation does not describe a route for a self-hosted site with no Jetpack connection.

That matters because viewers keep asking. On WordPress.com's official Claude video, which has 122,000+ views, a WordPress site owner asked whether it works on self-hosted sites, and the question had 16 likes. Another reported a permissions error that mentioned a Jetpack plan. The Agent Index listing also describes plan limits, so confirm what your account allows when you enable MCP.

Which tools you see also depends on your permissions and tool settings. Choosing Read only at sign-in turns every write tool off, which is a safe way to run an audit before allowing any edit.

Three guides are live. Getting cited by ChatGPT finds the buying questions that name a competitor and fixes the posts they point to. Refreshing old posts from Search Console finds queries sitting just off page one and edits the owning post block by block. Publishing from Claude takes a keyword through to a WordPress.com draft, with a paste-in handoff when write tools are off.

Each guide ends by re-running the same measurement in Linkeddit, so you judge the change on repeated samples rather than on one answer. The partner side of every guide is described from the vendor's documentation and was not run.

Linkeddit's compete_signals_get returns dated competitor moves with a category, a priority, why it matters and the source URL. search_competitors_get lists the competitor domains you track in Google results. Claude reads both, drafts the comparison, and the WordPress.com connector creates it with pages.create, which makes a draft by default.

Choose signals that change a buyer's decision. On Linkeddit's test project the two signals returned this week were both low priority: an Umami preview of its v3 release and a roadmap overview. A price change or a dropped feature would justify a comparison page. A roadmap post does not, so Claude should say so instead of writing one.

Keep the page factual. Cite each competitor claim to the source URL the signal carries, give the date, and show the draft to a person before it goes live. This is a section of the hub, not a separate guide, because the workflow is the same draft-and-approve path as publishing from Claude.