Chat
The assistant behind the Browse/Chat toggle. Ask reads your data and answers in plain English; Act proposes changes that run after you approve them.
Chat is the second surface in RankX AI. The Browse/Chat toggle under the website switcher swaps the left rail between the dashboard sections and your conversations, and the chat itself answers from your own tracked data: your prompts, rankings, audits, traffic, catalogue and tasks.
It has two modes, and the difference between them is the entire design:
| Mode | What it may do |
|---|---|
| Ask | Read your data and answer. It cannot change anything, and it cannot spend credits beyond the message itself |
| Act | Everything Ask does, plus propose changes. Nothing is applied until you approve the exact change on a card |
Ask and Act covers the mechanics; what it can and cannot do is the complete capability list, including the permanent limits.
A conversation belongs to one website
Every conversation is bound to the website selected when it started, and Chat will not switch websites mid-conversation. To ask about another site, start a new chat with that site selected. Conversations do not carry memory between each other: what you told one is not known to the next.
What a message costs
A message costs the same flat rate in both modes, however many steps the answer takes, and the rate is published in credit costs. Approving or declining a proposal, giving feedback and slash commands are free. When an approved action runs something that has its own price, a fresh rank check or an article draft, that action's own rate applies and the approval card quotes it before anything runs.
Who has it
Chat is on every plan, for you and your team, with no separate subscription and no per-seat fee. Agency Client portal logins do not see Chat in this release. A per-workspace Chat permission already exists so an agency can decide in advance of any future client access; see what clients can see.
Where the answers come from
Chat reads the same data the dashboards render, through the read tools the MCP server also exposes, with the same rule about honesty of denominators: an answer RankX AI could not analyse is unknown, never a miss. The two surfaces are not equivalent, and Chat is the narrower of them: MCP publishes write tools that Chat is never given, which is why what it can and cannot do is a shorter list than the tool reference. If the data does not support an answer, Chat says so rather than inventing one.
Last updated
The Task Creation Assistant
Turning an audit finding into a properly written task, with steps and a definition of done. What it costs, when to use it, and when not to.
Ask and Act
The two chat modes, the approval card every change goes through, the undo that survives it, and how conversation sharing works.