Every day, freelancers and small agencies burn hours on the same routine: searching Google Maps, opening business after business, checking websites, copying phone numbers into a sheet. It works, but it eats your day. What if 80% of that client hunting could run on autopilot, for free? That is exactly what this Muse AI prompt does.
Give the prompt to Muse AI, answer three quick questions, and Muse does the heavy lifting: it finds businesses in low competition cities, pulls their WhatsApp numbers and social profiles, audits their websites, scores every lead, and writes a personalized outreach message for each one. Everything lands in one clean sheet. Your only job is the last 20%: sending the messages.
What does this client hunting automation prompt actually do?
Think of it as a ready-made workflow that Muse follows step by step. It searches Google Maps for your niche in a city you choose, collects the listings, then checks each business properly: Is the website on WordPress? What is missing on it? How many Google reviews does the business have? Which phone numbers look like mobiles, the ones most likely on WhatsApp? The result is a prioritized lead sheet, HOT leads first, each with its own personalized message based on a real finding, never a generic blast.
It is the same automation mindset we used for autonomous link building with Muse: you set the rules once, and Muse runs the boring, repetitive work for you. And if you are wondering why businesses even show up, or do not show up, in these searches, our guide on why local SEO is important for small businesses explains the basics.
What is the master prompt, and how do you copy it?
Here it is, the complete master prompt. Copy it, paste it into a new chat with Muse AI, and follow the short guide below it. It is free, and you do not need any plugin or extension.
Copy the prompt below into a new chat with Muse. Fill in any [PLACEHOLDERS] when you run it, then Muse asks you 3 quick questions and runs the whole hunt itself.
CLIENT HUNTING AUTOMATION - Master Prompt
Give this prompt to a fresh Muse (new chat). It describes a complete client-hunting workflow for a local SEO agency. The new Muse must follow it end to end.
CUSTOMIZE (2 minutes, do once)
Your name / agency: outreach messages contain no signature by default. Tell Muse your name if you want one added.
Chat language: default is Roman Urdu. Tell Muse your preferred chat language and it will switch.
Value line: messages say "I help local service businesses show up higher on Google." Tell Muse your actual offer to replace it.
Niche / country defaults: none - Muse asks you every run (Section 0).
0. FIRST INTERACTION - ASK BEFORE DOING ANYTHING
Before any work, ask the user these THREE questions in one message, each with tap-options (use muse.create_options, max 3 options per question, plus they can always type their own answer):
Niche? Options: Plumber, Electrician, Roofer (he can type any other trade, e.g. HVAC, cleaner, painter).
Micro-level city? Offer 3 mid-size, low-competition cities across English-speaking countries (USA, UK, Canada, Australia - never fixed to one country), e.g. Norwich (UK), Newcastle (Australia), Hamilton (Canada). NEVER suggest London, New York, Los Angeles, Toronto, or Sydney (too competitive) unless he asks. He can type any other city and country.
How many leads? Options: 10, 20, 30.
Chat with the user in Roman Urdu (or the language set in CUSTOMIZE). All outreach content in English.
If the goal workspace ~/workspace/goals/client-hunting-automation/ exists on this VM, read its GOAL.md first for extra context.
1. PROCESS OVERVIEW
You automate everything UP TO the outreach message. The user sends the WhatsApp messages themselves - never automate sending (ban risk).
Pipeline: niche + city → Google Maps listings → WordPress check → website mini-audit → social profiles → extra/mobile phone numbers → lead scoring → personalized 5-part outreach message → Google Sheet (CSV). Deliver summary + mobile numbers + sheet.
2. STEP-BY-STEP EXECUTION
Step 1 - Extract listings from Google Maps (browser, read-only)
Use browser.browse with task_id: "new". One browser task at a time - never run multiple browser tasks in parallel (shared Chromium profile hangs).
Search Google Maps for "<niche> in <city>" (e.g. "plumber in Norwich, UK").
Extract the first N listings (N = lead count): business name, phone number (as shown), website URL, star rating, review count, full address.
Read-only: no logins, no form submissions, do not click into business websites.
If a CAPTCHA appears, the browser solves it itself. If a hard IP block page appears (e.g. "Access Denied" that does not clear), STOP and report it - do not retry in loops.
Step 2 - Build the input JSON
Create ~/workspace/goals/client-hunting-automation/hidden_files/real_input_<city>.json - a JSON array, one object per business:
[{"name": "Example Plumbing Ltd", "phone": "+44 1603 123456", "website": "exampleplumbing.co.uk",
"rating": 4.9, "review_count": 87, "address": "1 High St, Norwich NR1 1AA",
"city": "Norwich", "country": "UK", "niche": "plumber"}]
The country field is optional but recommended (helps disambiguate phone formats): use UK, USA, Australia, or Canada.
Use the REAL data from Step 1. Never invent or placeholder phone numbers in a real run.
Step 3 - Run the pipeline
A tested pipeline script already exists on this VM:
~/workspace/goals/client-hunting-automation/hidden_files/pipeline.py
Run it:
cd ~/workspace/goals/client-hunting-automation/hidden_files
python3 pipeline.py real_input_<city>.json --out <city>
It outputs leads.csv + leads.json in the out directory. What it does per business:
WordPress detection (homepage markers). Non-WordPress sites are dropped from outreach entirely (the user's hard rule - confirm it still applies). If the drop rate looks suspiciously high (>70%), spot-verify 2 dropped sites with curl before trusting it.
Mini site audit (WP sites only): MISSING_TITLE, MISSING_META_DESC, NO_SSL, SLOW_TTFB, NO_VIEWPORT, NO_SCHEMA, NO_H1.
Social extraction: facebook.com / instagram.com / linkedin.com profile links from the site (skip share-widget URLs).
Extra phone numbers: from homepage + contact pages (tel: links first). Validated against fake patterns.
Mobile detection: UK numbers starting 07 (→ +447...) and AU numbers starting 04 (→ +614...) are flagged as likely mobile = likely WhatsApp-able. US numbers cannot be classified as mobile/landline from the number alone. The pipeline supports US/CA (+1), UK (+44), and AU (+61) numbers. Output columns include mobile_phones (the WhatsApp-ready subset).
Scoring 0-100: +25 if reviews < 150, +10 if < 500, +5 per issue (max +20), +5 if rating < 4.5 with 20+ reviews, −15 if reviews > 2000. Tiers: HOT ≥ 70, WARM ≥ 45, else COLD. Sorted HOT-first.
Personalized message per the 5-part framework (Section 3).
Step 4 - QA the output (mandatory)
Open leads.csv and check: no fictional/placeholder numbers, no 555 numbers, no garbage digit strings in phone columns.
Read 2-3 sample messages: every claim in the message must match a REAL finding (issue label or review count). No invented gaps.
Confirm non-WordPress rows are absent from the CSV.
Step 5 - Deliver
Reply to the user in Roman Urdu (or their preferred language) with:
Short result summary: counts (HOT/WARM/COLD), how many dropped as non-WP, how many mobile numbers found.
The mobile numbers only, one per line, no serial numbers, no names - he copies these straight into WhatsApp:
+447946445535
+447826745894
The sheet attached: 
1-2 sample messages for his review.
Note anything unusual (e.g. a very-low-review lead dropped as non-WP).
3. THE 5-PART MESSAGE FRAMEWORK (standing rule)
Every outreach message follows this exact order:
HOOK - earn attention: "Hi there! I came across {business} while looking at {niche}s in {city}."
RELEVANCE - one VERIFIED detail (proves it's personal). Priority: real audit issue → low review count (<150) → strong rating (factual, no praise) → fallback (hook context only, never invent a detail).
GAP - one likely need tied to that detail (e.g. slow site → "Slow sites usually rank lower on Google, so competitors quietly take those calls.").
VALUE - one credible benefit, no inflated claims: "I help local service businesses show up higher on Google."
QUESTION - one easy EITHER/OR question, never yes/no (e.g. "Is growing your Google presence a priority right now, or are you fully booked already?").
Pre-send checklist: verified detail, one relevant point, credible benefit, one easy question.
NEVER: fake familiarity, inflated claims, generic praise, multiple asks.
Message rules: English only. WhatsApp-short (4-6 lines). Never use the em dash character (use comma/hyphen). Max 1 emoji, usually none. No links. No signature line (no "- Your Name, Your Agency" - see CUSTOMIZE).
Example (HOT lead, low reviews):
Hi! I found Hyde Park Plumbing while looking through plumbers in Boise. I saw you have only 70 Google reviews. With fewer reviews than competitors, new customers often pick someone else first. I help local service businesses show up higher on Google. Have you been actively collecting reviews, or has it been on the back burner?
4. STANDING RULES (learned the hard way - do not break)
Browser-only, zero paid APIs. No DataForSEO, no SerpApi, no paid services. Everything runs on this VM for free.
Sending is manual. You prepare; the user sends. Never automate WhatsApp/IG/FB/LinkedIn sending.
WhatsApp availability cannot be auto-verified. The wa.me landing page does not reliably distinguish (tested). Do NOT add a fake whatsapp_available YES/NO column. Ground truth = the user opening the chat (profile visible = on WhatsApp).
Test data vs real data: never let fictional/test numbers into a delivered sheet. Test on copies; deliver only real runs.
Country reality (guidance for city suggestions, not a restriction): any English-speaking country works (USA, UK, Canada, Australia). For WhatsApp-first outreach, UK is best (07 mobiles) and Australia is good (04 mobiles); USA/Canada are weaker (landline culture, low WhatsApp hit rate). Prefer suggesting UK/Australian cities when the user wants maximum WhatsApp replies, but always accept their pick.
Small independents > big companies: low-review businesses usually list personal mobiles = WhatsApp-able. The scoring already favors them.
Live-site etiquette: read-only fetching. No form submissions, no logins, no changes anywhere.
5. SHEET COLUMNS
business_name | phone | whatsapp_number | extra_phones | mobile_phones | website | is_wordpress | review_count | rating | issues_found | social_facebook | social_instagram | social_linkedin | priority_score | priority_tier | personalized_message
whatsapp_number: primary number in international format (+44/+1...).
mobile_phones: pipe-separated likely-mobile numbers - the user's WhatsApp hit list.
Non-WordPress businesses: excluded from the sheet, counted in the summary.
6. GOAL LOGGING
A user goal named "Client hunting automation" exists. Log each completed run with user_goal.create_entry (title, one-line current_state, one-sentence description).
How do you use the prompt? A step by step guide
Here is the whole process in simple steps. Once you have done it one time, every next run takes you only a few minutes.
Step 1: Give the prompt to Muse AI
Copy the master prompt from the box above and paste it into a new chat with Muse AI. That is the entire setup. Muse reads the instructions and gets ready to run the workflow for you. No plugin to install, no extension, no account to create anywhere.
Step 2: Answer Muse’s 3 questions
Muse will ask you three things, each with options you can tap. First, your niche: plumber, electrician, roofer, or any trade you type in, like HVAC, cleaner or painter. Second, a city: Muse suggests mid-size, low competition cities across the USA, UK, Canada and Australia. Small cities are the point here, they are far easier to win than London, New York or Sydney. Third, how many leads you want: 10, 20 or 30. For your first run, 10 is perfect. Answer, and the automation starts on its own.
Step 3: Muse collects the data from Google Maps
This is where the real work disappears. Muse searches Google Maps for your niche in your city and collects what actually matters for outreach: the business name, phone and WhatsApp-ready mobile numbers, website URL, star rating, review count, full address, and social profiles on Facebook, Instagram and LinkedIn. It then audits each website: is it on WordPress, what is missing on it (title tag, meta description, SSL, speed), and scores every lead so the hottest ones come first.
Step 4: Everything arrives in one sheet
When the run finishes, you get a clean sheet with every qualified lead sorted into HOT, WARM and COLD tiers. Next to each lead sits a personalized outreach message, written around a real detail from that business, like having only 70 reviews while competitors have hundreds, or a website missing its title tag. Nothing generic, nothing invented.
Step 5: Send the messages on WhatsApp
Now your 20% begins. Open the mobile numbers from the sheet, check that the business is on WhatsApp (most small independents are, especially in the UK and Australia), copy the personalized message, and send it yourself. Sending stays manual on purpose: it keeps your number safe and the message reads like a person wrote it, because the research behind it is real.
Step 6: No WhatsApp? Use their social profiles
If a lead is not on WhatsApp, do not drop them. The sheet already has their Facebook, Instagram and LinkedIn profiles, and business owners are active there too. Send the same personalized message as a DM instead of a WhatsApp text.
And that is the whole system. No cheap plugin, no paid extension, no scraping subscription. The prompt plus Muse AI does the hunting, you do the talking. If you ever lose this page, the prompt is also pinned in the video comments, and on this website under the Resources tab, in the prompts section.
What are the most common questions about client hunting automation?
Is this client hunting automation really free?
Yes. The workflow runs on Muse AI and a normal browser. There are no paid APIs, no DataForSEO or SerpApi subscriptions, no plugins and no extensions to buy. The only cost is your own time for sending the messages.
Will WhatsApp ban my number for sending these messages?
The risky part of WhatsApp outreach is bulk automated blasting, and this workflow deliberately avoids that. You send a small number of personalized messages by hand, the same way you would message any business contact. Keep it human, respect replies, and never spam the same business twice.
Which niches and cities work best?
Local service trades work best: plumbers, electricians, roofers, HVAC, cleaners and painters. For cities, mid-size and low competition beats famous ones every time. UK and Australian businesses are especially good for WhatsApp outreach, because their listed numbers are usually personal mobiles.
What exactly do I get in the sheet?
For every lead: business name, phone, WhatsApp-ready mobile numbers, website, whether it runs on WordPress, review count, rating, the website issues found, Facebook, Instagram and LinkedIn profiles, a priority score with a HOT, WARM or COLD tier, and a personalized outreach message.
Do I need any technical skills to run this?
No. If you can copy a prompt, answer three questions and send a WhatsApp message, you can run the whole system. Muse handles the technical parts itself.
Want Clients Without Doing the Hunting Yourself?
This prompt gives you the system. But if you would rather skip the process and have a local SEO team bring your business more calls from Google, Local SEO Rankers can do it for you.
Chat on WhatsApp