The Exact System I Built to Research 5 Competitors While I Sleep (Claude Computer Use + 2 Hours)
Alright, you asked for the breakdown. Here it is.
Last week I built a system using Claude Computer Use that automatically researches my competitors every single morning. No manual clicking. No note-taking. Just a full competitive intel report in my inbox by 8 AM.
I'll walk you through the exact setup so you can build your own.
What You Need
  • Anthropic API key (Claude with computer use access)
  • A list of 5-10 competitor URLs
  • A simple Python script (I'll give you the structure below)
  • A scheduling tool (I use cron, but Windows Task Scheduler works too)
The Architecture
The system runs in 3 phases:
Phase 1: Screenshot Collection
Claude opens a headless browser. Navigates to each competitor URL. Takes full-page screenshots of:
  • Homepage / landing page
  • Pricing page
  • Any new blog posts or announcements
It saves each screenshot with a timestamp. This matters because Phase 2 compares against last week's screenshots.
Phase 2: Analysis
This is where it gets interesting.
Claude receives two images for each competitor. Last week's screenshot and this week's screenshot. Then it analyzes:
  • Messaging changes (did they update their headline or value prop?)
  • Pricing updates (new tiers, price changes, feature additions)
  • Positioning shifts (are they going upmarket? Pivoting to a new audience?)
  • New campaigns (banner ads, promo bars, new CTAs)
  • Feature launches (new product pages, updated feature lists)
The prompt I use is specific. I tell Claude to write the analysis in my voice and flag anything that requires immediate action vs. things that are just worth monitoring.
Phase 3: Report Generation
Claude compiles everything into a structured report:
  • Executive summary (3 bullet points, what matters most this week)
  • Competitor-by-competitor breakdown
  • Action items ranked by urgency
  • Raw screenshots attached for reference
The report gets emailed to me automatically. I wake up, skim it with coffee, and know exactly what my market is doing.
The Prompt Structure
Here's the skeleton of how I prompt Claude for the analysis step:
--
You are a competitive intelligence analyst for [your company].
Compare these two screenshots of [competitor name].
Image 1: Last week's capture
Image 2: This week's capture
Identify:
1. Any messaging or positioning changes
2. Pricing modifications
3. New features or product updates
4. Campaign or promotional changes
5. Design/UX shifts that signal strategic direction
Rate each finding:
- ACTION REQUIRED: Something we need to respond to
- MONITOR: Worth tracking but no immediate action
- FYI: Minor change, logged for context
Write your analysis in a direct, concise style. No fluff. Lead with the most important finding.
--
What I Learned After Week 1
The system caught 3 things I would have completely missed:
  1. A competitor quietly raised their starter plan price by 20%. No announcement. Just changed the number.
  2. Another competitor added "AI-powered" to their homepage headline. They're repositioning into our space.
  3. A third competitor removed their free tier page entirely. They're going upmarket.
All three of these required strategic responses. And I would have found out weeks later scrolling X or hearing it secondhand.
Common Questions
How long did this take to build?
About 2 hours on a Sunday. Most of that was getting the screenshot timing right.
How much does it cost to run?
Roughly $2-3 per day in API costs for 5 competitors. That's less than $100/month for daily competitive intelligence.
Can I use this for more than competitors?
Yes. I'm building a version that monitors industry blogs, product hunt launches, and subreddit discussions too.
Your Turn
If you want the actual Python script template, drop a comment below and I'll share it.
This is the kind of system that takes 2 hours to build and saves you 10+ hours every single week. The ROI is insane.
The best part is it gets better over time. The more weeks of screenshots you have, the better Claude gets at spotting patterns and trends.
Stop manually researching. Build a system that does it for you.
0
0 comments
Jordan Hong Tai
1
The Exact System I Built to Research 5 Competitors While I Sleep (Claude Computer Use + 2 Hours)
powered by
AI Marketing with Jordan
skool.com/ai-marketing-with-jordan-1336
Learn how to grow crypto, AI, and SaaS products using AI-powered marketing systems. Built by a marketer who codes.
Build your own community
Bring people together around your passion and get paid.
Powered by