GEO Prospect Manager: A CRM for Selling AI Search Services
Tracks GEO agency leads through lead, qualified, proposal, won, and lost stages in a local JSON file, with audit scores, notes, and MRR forecasting
Twelve prospects, three audits, and a follow-up you meant to send on Tuesday, all living in a spreadsheet and your memory. This playbook keeps the pipeline in one JSON file Claude reads and writes, so 'show me the pipeline' returns committed MRR, stage counts, and the two deals that need moving today.
Who it's for: GEO and SEO agency owners, solo consultants, freelance marketers, agency sales leads, anyone selling AI search optimization retainers
Example
"Show me the pipeline" → A stage-by-stage summary: 2 leads, 1 qualified at €4,500/mo, 1 proposal awaiting signature, 3 won for €18,500 committed MRR, plus next actions naming which prospect to send a proposal to and which follow-up is 8 days overdue
New here? 3-minute setup guide → | Already set up? Copy the template below.
# GEO Prospect Manager
A CRM-lite for a GEO agency. Track leads through the sales pipeline, store audit history and notes, and forecast revenue without opening a CRM tool.
GEO means Generative Engine Optimization — getting a client's site cited by ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. The pitch works because the score is measurable and most sites score badly.
All data lives in `~/.geo-prospects/prospects.json` and persists across sessions.
## How to Ask
Speak plainly. Match the request to the action:
| Say | What you get |
|-----|--------------|
| "New prospect: electron-srl.com" | Create a prospect record |
| "Show my pipeline" / "List prospects" | Summary table of every prospect |
| "List qualified prospects" | Filtered by status |
| "Show me PRO-001" | Full detail with note history |
| "Audit electron-srl.com" | GEO scan, saved to the prospect record |
| "Note on Electron: they asked about pricing" | Timestamped interaction note |
| "Move acme.com to proposal" | Status change |
| "We won bigshop.it at 6000/month" | Mark won, set contract value |
| "We lost acme.com, budget freeze" | Mark lost with reason |
| "Pipeline summary" | Revenue forecast by stage |
## Data Structure
Each prospect is one JSON record:
```json
{
"id": "PRO-001",
"company": "Electron Srl",
"domain": "electron-srl.com",
"contact_email": "info@electron-srl.com",
"contact_name": "",
"industry": "Educational Equipment Manufacturing",
"country": "Italy",
"status": "qualified",
"geo_score": 32,
"audit_date": "2026-03-12",
"audit_file": "~/.geo-prospects/audits/electron-srl.com-2026-03-12.md",
"proposal_file": "~/.geo-prospects/proposals/electron-srl.com-proposal.md",
"monthly_value": 0,
"contract_start": null,
"contract_months": 0,
"notes": [
{
"date": "2026-03-12",
"text": "Initial GEO quick scan. Score 32/100 - Critical tier. Strong candidate for GEO services."
}
],
"created_at": "2026-03-12",
"updated_at": "2026-03-12"
}
```
The JSON file is the single source of truth. Read it before every operation, write it after every change.
## Storage Layout
```
~/.geo-prospects/
├── prospects.json # Main CRM database
├── audits/ # Audit snapshots
│ └── electron-srl.com-2026-03-12.md
└── proposals/ # Generated proposals
└── electron-srl.com-proposal.md
```
Create the directories if they do not exist:
```bash
mkdir -p ~/.geo-prospects/audits ~/.geo-prospects/proposals
```
## Operations
### New prospect
1. Create `~/.geo-prospects/prospects.json` as an empty array if it does not exist
2. Derive the company name from the domain (`electron-srl.com` → `Electron Srl`)
3. Assign the next sequential ID: `PRO-001`, `PRO-002`, and so on
4. Ask for contact name (optional), contact email, and a monthly value estimate (optional). Ask once, in one message.
5. Set status to `lead`
6. Save
7. Suggest the next step: run an audit to score the prospect
### List prospects
Read the JSON and render a table:
```
GEO Prospect Pipeline — March 2026
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
ID Domain Company Status Score Value
─────── ────────────────────── ──────────────── ────────── ───── ──────
PRO-001 electron-srl.com Electron Srl Qualified 32/100 €4.5K
PRO-002 acme.com ACME Corp Lead — —
PRO-003 bigshop.it BigShop Won 41/100 €6.0K
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Pipeline: 1 lead | 1 qualified | 0 proposals | 1 won | 0 lost
Committed MRR: €6,000 | Pipeline Value: €4,500
```
Accept a status filter and show only matching rows.
### Audit a prospect
1. Run a GEO scan of the domain and produce an overall score out of 100
2. Save `geo_score` and `audit_date` to the prospect record
3. Write the audit output to `~/.geo-prospects/audits/<domain>-<date>.md`
4. Store the path in `audit_file`
5. Add an automatic note: "Quick audit run. GEO Score: XX/100."
6. If the score is under 55, flag it: strong sales opportunity, generate a proposal next
A minimum viable scan checks robots.txt for AI crawler access (GPTBot, ClaudeBot, PerplexityBot and the rest), the presence of llms.txt, schema.org coverage on key pages, author and date signals on content, and brand presence on the platforms AI engines cite. The full scoring engine, with weighted categories and Python analyzers, lives in the `geo-seo-claude` repo. Without it, produce a directional score and label it as a snapshot rather than a full audit.
### Add a note
Find the prospect by ID or domain, append the note with the current ISO date, save, confirm.
### Change status
Valid statuses: `lead`, `qualified`, `proposal`, `won`, `lost`.
Update the field, add an automatic note recording the change, save.
Mark won with a monthly contract value. Mark lost with a reason, always. A lost deal without a logged reason teaches you nothing.
### Pipeline summary
```
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
GEO AGENCY PIPELINE SUMMARY — March 2026
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
STAGE COUNT POTENTIAL VALUE NOTES
───────────── ───── ─────────────── ─────────────────────
Lead 2 €8,000/mo New discoveries
Qualified 1 €4,500/mo Ready for proposal
Proposal Sent 1 €6,000/mo Awaiting signature
Won 3 €18,500/mo Active clients (MRR)
Lost 1 — Budget freeze
COMMITTED MRR: €18,500
PIPELINE (qualified+): €10,500
TOTAL POTENTIAL: €29,000/mo → €348,000/yr
Next actions:
→ PRO-003 (acme.com): Send proposal — score 38/100 (strong case)
→ PRO-007 (shop.it): Follow up — proposal sent 8 days ago
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
```
Committed MRR counts won deals only. Pipeline counts qualified and later. Do not add lead-stage guesses to a forecast.
End with concrete next actions: which prospect to move, and why.
## Pipeline Stage Definitions
| Status | Meaning | Typical next action |
|--------|---------|---------------------|
| `lead` | Discovered, not yet contacted | Run an audit, assess the opportunity |
| `qualified` | Audit done, pain points confirmed | Generate a proposal |
| `proposal` | Proposal sent, awaiting decision | Follow up, answer questions |
| `won` | Contract signed, active client | Run a full audit, start onboarding |
| `lost` | Closed lost | Log the reason for future reference |
## Qualification Signals
A domain is worth pursuing when the audit shows:
- Score under 55, which means visible, fixable problems
- Blocked or unconfigured AI crawlers in robots.txt, the fastest quick win to demo
- No schema markup on product, service, or location pages
- Content with no author, no dates, and no citable structure
- A competitor already showing up in AI answers for the same queries
That last one closes deals. Ask ChatGPT or Perplexity a buying question in the prospect's category, screenshot the answer, and point out who got named.
Skip prospects scoring above 75. The work is maintenance, the budget is small, and the pitch is weak.
## Rules
- Read the JSON before every operation. Never write from memory.
- Print confirmation and the current prospect state after every change.
- Never overwrite the notes array. Append.
- Do not create files outside `~/.geo-prospects/` unless asked.
- Store no data you would not want in a plain-text file on your laptop. This is a JSON file, not an encrypted CRM.
What This Does
Runs a lightweight CRM for a GEO agency out of a single JSON file. GEO means Generative Engine Optimization: getting a client's site cited by ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews.
You talk, Claude updates the file. Add a prospect, score it with an audit, log a call note, move it to proposal, mark it won with a contract value. Ask for the pipeline and you get stage counts, committed MRR against pipeline value, and named next actions.
The pitch works because the score is measurable and most sites score badly. This playbook keeps track of who you scored, what you found, and where each deal sits.
Quick Start
Step 1: Create a Project Folder
mkdir -p ~/Documents/GEOAgency
Step 2: Download the Template
Click Download above, then:
mv ~/Downloads/CLAUDE.md ~/Documents/GEOAgency/
Step 3: Add Your First Prospect
cd ~/Documents/GEOAgency
claude
Say "New prospect: electron-srl.com." Claude derives the company name from the domain, assigns PRO-001, asks once for contact details and a value estimate, and writes the record.
Data lives in ~/.geo-prospects/prospects.json, outside the project folder, so it survives across sessions and projects.
The Commands, in Plain English
| Say | What happens |
|---|---|
| "New prospect: acme.com" | Creates a record, status lead |
| "Show my pipeline" | Table of every prospect with status, score, value |
| "List qualified prospects" | Same table, filtered by stage |
| "Show me PRO-001" | Full detail with note history |
| "Audit electron-srl.com" | GEO scan, score saved to the record |
| "Note on Electron: asked about pricing" | Timestamped note appended |
| "Move acme.com to proposal" | Status change with an automatic note |
| "We won bigshop.it at 6000/month" | Marks won, sets contract value |
| "Pipeline summary" | Revenue forecast with next actions |
Pipeline Stages
| Status | Meaning | Next action |
|---|---|---|
lead |
Discovered, not contacted | Run an audit, assess the opportunity |
qualified |
Audit done, pain points confirmed | Generate a proposal |
proposal |
Sent, awaiting decision | Follow up, answer questions |
won |
Signed, active client | Full audit, start onboarding |
lost |
Closed lost | Log the reason |
Committed MRR counts won deals only. Pipeline value counts qualified and later. Lead-stage guesses stay out of the forecast.
Qualification Signals
The template tells Claude what makes a domain worth pursuing:
- Score under 55, meaning visible and fixable problems
- Blocked or unconfigured AI crawlers in robots.txt, the fastest quick win to demo
- No schema markup on product, service, or location pages
- Content with no author, no dates, and no citable structure
- A competitor already appearing in AI answers for the same buying queries
That last signal closes deals. Ask ChatGPT or Perplexity a buying question in the prospect's category, screenshot the answer, and show them who got named instead.
Prospects scoring above 75 get skipped. The work is maintenance, the budget is small, the pitch is weak.
Pairs With the Proposal Generator
Prospecting feeds proposals. Once a prospect hits qualified, the GEO Proposal Generator reads the same audit file and writes a priced, client-ready proposal, then flips the record's status to proposal and stores the file path. Drop both CLAUDE.md files in the same folder and the two halves connect through ~/.geo-prospects/.
Limitations
- The scoring here is a snapshot, not a full audit. A directional score comes from checking robots.txt, llms.txt, schema coverage, and author signals. The weighted 11-dimension engine with Python analyzers lives in the source repo.
- Storage is a plain JSON file on your disk. No encryption, no sync, no multi-user access. Keep contact data you would be comfortable leaving unencrypted on a laptop.
- No email integration, no reminders, no calendar. Follow-up timing comes from you asking for the pipeline.
- One user. Two people editing the same
prospects.jsonwill overwrite each other.