Prompt tracking
Your buyer prompts, re-asked on every engine each run.
Find the prompts you're losing.
Asks your buyers' questions on ChatGPT, Claude, Gemini, Perplexity, Grok, DeepSeek and Llama, then shows which answers cite you, which cite a competitor, and what to change.
Free analysis · 5 credits · no card · first score in about 30 seconds
7 answer engines · every run · llms.txt + JSON-LD
coverage 4 of 25 cited pages listed
missing /pricing /earn /card /tax-reports
competitor coinbase.com lists 18 pages
impact +4 score · projected 3 new citations
sample run · demo data
01 / The problem it removes
Your docs answer the question. The engine cites someone else.
Buyers ask an answer engine what to use. The reply names a competitor, and no analytics tool shows you that conversation.
Without it
- Buyers ask ChatGPT, not Google, and you never see those answers.
- Competitors get cited for questions your docs answer better.
- llms.txt and JSON-LD rot quietly; nothing alerts you.
With the GEO agent
- Every tracked prompt runs on all 7 engines, every run.
- Each citation gap becomes one fix: llms.txt, JSON-LD, entity pages.
- The next run shows whether the engines picked it up.
02 / How it runs
A loop, not a task list.
- 01
Ask
Your tracked buyer prompts run on 7 answer engines.
- 02
Compare
Citations are scored against your competitors, per engine.
- 03
Draft
llms.txt rewrites, JSON-LD blocks and entity pages land as drafts.
- 04
Re-measure
The next run shows whether the engines picked it up.
03 / What it can do
What the agent does
Citation diffing
Who got cited instead of you, and from which source.
llms.txt drafts
Full-coverage rewrites, ready for review.
Structured data
Organization, FAQ and Product JSON-LD blocks.
Entity coverage
Thin pages flagged for merge or expansion.
GEO score
One 0 to 100 number, tracked run over run.
04 / In the product
A recommendation lands. You ship it in minutes.
Every GEO agent item arrives with evidence and a ready prompt. Copy it into Claude Code or Cursor, review the diff, approve. The next run measures whether it moved the score.
Review queue
In your queue
Rewrite llms.txt with full product coverage
↓ then paste it
claude "Rewrite llms.txt with full product coverage"
⏺ working… reading evidence from ai-cmo (MCP)
⏺ diff ready · 2 files changed
✓ shipped · re-measure queued for the next run
sample run · demo data
coverage 4 of 25 cited pages listed missing /pricing /earn /card /tax-reports competitor coinbase.com lists 18 pages impact +4 score · projected 3 new citations05 / Questions
Before you install it.
Seven: ChatGPT, Claude, Gemini, Perplexity, Grok, DeepSeek and Llama. The standard preset asks up to 150 buyer questions per run, spread across the engines, and records whether you are mentioned, where, and which sources get cited. Five answer with live retrieval, two from model memory.
A 0 to 100 number with five weighted parts: share of voice 35%, visibility 25%, average position 15%, modeled reach 15%, sentiment 10%. A part that cannot be calculated is dropped and the rest re-normalized, so a missing signal never counts as zero.
Yes. Free includes the GEO, SEO and X agents with 5 one-time credits that never expire, no card. A credit covers $0.20 of provider cost plus a fixed surplus. Work your balance cannot cover is refused up front, so there are no overage charges.
No. Each gap lands as a proposal you accept, decline or archive, and an item only completes once a review records the evidence. It runs monthly on Free, weekly on Lite, daily on Pro, or you switch it off.
Included in the free analysis
Start with the free analysis.
Free analysis, 5 credits that never expire, no credit card. The GEO agent briefs itself on your site during onboarding.