Growth Agent
Ask "why did my AI visibility drop?" and get a data-grounded diagnosis with ranked recommendations.
The AI Growth Agent answers questions about your project's growth with your project's data — GEO runs, citations, gaps, rankings, audit issues — not generic SEO advice.
Chat
Open AI Growth Agent and ask in plain language:
- "Why did our AI visibility drop this week?"
- "Which competitor is gaining on us, and on what?"
- "What should we publish next?"
- "Is our site technically holding us back?"
The agent's chat works on a data snapshot: before answering, it gathers the project's current numbers (rankings, AI visibility, citations, content gaps, competitor benchmarks, audit issues) into a grounded context block and answers from that — the same underlying queries also power parts of MCP, but chat itself is not a tool-calling loop; the snapshot is assembled once per message. It cites the numbers it used in the answer, so you can verify every claim against the dashboard page it came from.
Conversations are saved per project; reopen them from the history list to read past answers or continue one.
Analysis
Analyze runs a full diagnostic — visibility trend, top movers, gap summary, critical audit issues — and files a structured report with ranked recommendations. It runs when you start it; nothing runs on a schedule. The weekly digest email (Tuesdays) summarizes data you already have and costs nothing.
Recommendations land on the project's Recommendations list with impact
estimates and statuses (open → in_progress → done / dismissed), so the
agent's output becomes a tracked backlog rather than a chat transcript.
Cost
Agent actions spend real provider tokens. Each one holds its maximum cost first and settles at the actual token cost; a provider call that fails before it is billed costs nothing.
| Action | About |
|---|---|
| Chat message | $0.018 |
| Analysis | $0.015 |
See Usage & balance.
What it's good for
The agent is a diagnosis layer: it reads every data source the dashboard offers and ranks what matters. It does not fabricate data — when an answer says "mentions dropped 12% in ChatGPT", that number came from a stored run. Use it for triage ("what changed?"), planning ("what next?"), and reporting ("summarize this month") — then execute through the gap analyses and briefs it points at.