Visibility
Weekly visibility pulse. Needs read:projects and read:visibility.
Month-over-month trend. Needs read:projects and read:visibility.
Competitors
Competitor share of voice. Needs read:projects and read:visibility.
Who's winning the conversation. Needs read:projects and read:visibility.
Citations and sources
Which sources to pitch. Needs read:projects and read:visibility.
Community footprint. Needs read:projects and read:visibility.
Own-site citations. Needs read:projects and read:visibility.
ChatGPT shopping
Shopping carousel check. Needs read:projects and read:visibility, on a plan with Shopping.
One product up close. Needs read:projects and read:visibility, on a plan with Shopping.
Demand
Demand research. Needs read:projects and read:demand.
Prioritize prompts by demand. Needs read:projects, read:visibility and read:demand.
Agents and tables
Read a table. Needs read:projects and read:tables.
Run an agent. Needs read:projects, read:agents and run:agents. The run uses your workspace's credits, and most clients ask before starting it.
Writing better prompts
- Ask for the data before the opinion. “List the cited pages, then group them” gets more reliable answers than “what should I do about citations?”
- Ask for more than one page when you need it. Tools return 20 items by default and up to 100. Say “list all” or “the top 50” and your assistant follows
next_cursoror raiseslimit. - Stick to what the tools return. Visibility is per brand and per day, and citations are per brand. The MCP server doesn’t break visibility down by engine or citations down by prompt.