GEO Agent

The GEO Agent is an autonomous system that monitors your AI visibility, identifies opportunities, and executes GEO improvements based on a mission you define.

Agent vs automation rules

The GEO Agent is not a rule-based scheduler. It uses a planning cycle to reason about your visibility data, prioritize opportunities by impact, and decide which tasks to execute. Each run produces a plan with specific tasks grouped into phases.

Automation rules (playbooks) are simpler: they fire when a threshold is crossed (e.g., visibility drops below 40) and execute a single action (e.g., generate a brief). The Agent is more comprehensive — it runs a full Monitor → Research → Content → Outreach cycle.

Mission

A Mission is a plain-language goal that guides the Agent's planning. Example: "Increase Tier 1 mentions for comparison prompts to 60% win rate."

The Mission sets the direction. The Agent determines which actions to take to move toward it, using current visibility snapshots to track currentValue against the mission target.

How to set your first mission

1. Go to Settings and activate the GEO Agent
2. On the GEO Agent page, click New Mission
3. Type your goal in plain language
4. Set a target value (e.g., Visibility Score of 70)
5. Save — the Agent will begin its first planning cycle

Agent fleet

The coordinator runs five specialized agents in sequence:

Agent	Role
Coordinator	Orchestrates the full planning cycle, builds the final plan
Monitor	Scans visibility data, identifies source gaps
Research	Analyzes monitor output, identifies schema/citation/content opportunities
Content	Creates optimize actions and brief drafts from research opportunities
Outreach	Creates outreach drafts targeting citation gap sources

Automation levels

Level	Name	What the Agent can do
0	Monitor	Reads data, logs findings. No actions taken.
1	Drafts	Creates action and brief drafts for human review. All tasks require approval.
2	Semi-Auto	Executes schema fixes and brief creation automatically. Outreach requires approval.
3	Full	Executes all task types autonomously. Approval only for outreach at enterprise request.

Your plan determines the maximum level available (see table below).

Plan limits

	Starter	Pro	Growth	Enterprise
Agent enabled	No	Yes	Yes	Yes
Max level	—	1	2	3
Coordinator runs/month	0	4	30	Unlimited
Actions/month	0	20	Unlimited	Unlimited
Fix generations/month	0	10	50	Unlimited
Chat messages/month	0	30	100	Unlimited
Max playbooks	0	2	10	Unlimited

Reasoning feed

The reasoning feed shows the Agent's live thinkingLog during a coordinator run. Each entry includes a message (e.g., "Research: 4 opportunities identified") and a log level (info / success / warning). Use it to understand why the Agent chose specific tasks.

Approval queue

Tasks that require approval show in the Approval Queue before execution. Schema fixes and outreach drafts always require approval at Level 1. At Level 2, only outreach requires approval.

Agent memory

The Agent learns from scan history and action outcomes:

Memory type	Trigger	Confidence
pattern	Brand tier improved after schema addition within 30 days	0.75
failed	Action was dismissed or had no measured impact after 14 days	0.90

Memory is stored per workspace and used in future planning cycles to avoid repeating failed approaches.

Playbooks

Playbooks are trigger-based automation rules that run independently of the main coordinator cycle:

Trigger	Fires when
visibility_drop	Average score < threshold (24h lookback)
competitor_surge	Competitor mention % > threshold (24h lookback)
sentiment_decline	Negative response % > threshold (24h lookback)
new_citation_opportunity	Source is cited but your brand is absent

Each playbook has a 6-hour debounce to prevent repeated firing. The weekly_summary trigger is disabled in the current version.

Next steps

• Agent API — control the Agent via API
• Optimize Actions — review Agent-generated actions
• Briefs — Agent-created content briefs