Autopilot now logs every change — who flipped it on or off, what pace they set, and whether they did it from settings or onboarding. Admins can see that history per brand in the plan-and-usage panel, alongside whether the brand qualifies for the Pro guarantee.
The Pro guarantee baseline now has sharper data to work from. Tracked questions record how they entered the system — generated by Pendium, pulled from a plan, added manually, or edited — so only the right ones count toward eligibility. The admin panel shows each brand's eligible question count directly.
Free accounts now complete onboarding before anything else opens. After that, the report, Fact Sheets, the free page and post, and other sections show their locked state rather than staying accessible. The sidebar no longer uses lock icons on individual rows — the locked views handle that instead. Cold-email report links now route through homepage sign-up and agent setup rather than dropping users into a partial state, and the old visibility scan preview URL redirects to the homepage.
The /mcp page has a setup prompt again. Below the connection snippet, "Copy the setup prompt" puts a plain-English prompt on your clipboard that you paste into an agent — Claude Code, Cursor, Codex, or similar — and it installs the connector and walks through sign-in. "See the prompt" lets you read it first.
In Create, if a question or service you enter into a multi-idea run assumes something your knowledge base, Business Info, or facts sheet can't back up — a location, a price, a specific offer — the run continues, but the outline and draft are told not to claim it. The run list flags it clearly so you know what was left out and why. AI Agent Activity now counts requests for a brand's llms-full.txt across all surfaces and excludes it from the pages list, the same way llms.txt is handled.