VC Dealflow with AI Scrapers

•By Dave Wang

I was deploying a few AI web scrapers this weekend and I'm surprised more people don't use these in their workflows.

I'll show you below how I leveraged ChatGPT Operator + ChatGPT o3 model to:

  • Scrape a list of today's top launches on Product Hunt
  • Use ChatGPT to categorize each company & give you quality research into what they do and the opportunity set

Obviously this is useful for VCs for dealflow, but it's also great for founders looking to get inspired by new features to add.

​

​

Here's our gameplan:

  1. Give ChatGPT Operator the detailed prompt to gather relevant information we need in their native AI browser
  2. Insert the list of companies into a pre-prepared "Due Diligence" checklist

I'll show you below my raw prompts + output links 👇

The Prompt Chain:

Prompt 1 (ChatGPT Operator: https://operator.chatgpt.com/)

Get me a list of today's top launches on Product Hunt (Link: https://www.producthunt.com/) where it can be found on "Product Archive".
​
Make sure you scroll down to capture all of the companies since the site uses lazy loading.
​
Return to me (1) Name of company (2) Product Hunt link of that company

Result for Prompt 1:

​https://operator.chatgpt.com/v/680d32709e6c81908f4088eb93e75945​

​

​

Prompt 2 (Use ChatGPT o3 for Brief Memo Prep)

You are:
• A senior venture-capital analyst writing a board-ready memo that can be read in ~90 seconds.
• You have live web-browsing enabled. Use it aggressively for current facts, numbers, and citations.
​
TASK
Write one consolidated due-diligence report covering the following companies:
<<PASTE-COMPANY-LIST-HERE>>
​
OUTPUT GUIDELINES
1. Format & Tone​
▸ Board-ready, formal business language.
▸ Concise bullet points plus short narrative prose where context helps.
▸ Use Markdown headings and tables for clarity.
2. Overall Summary (top of memo)​
▸ A 3-column table listing every company and your recommended tier: Full DD / Light DD / Pass.
​
3. Company Sections – repeat in the exact order shown:
{Company Name}
One-sentence elevator pitch
​
Founders & Team​
• Names, titles, notable past roles, alma maters, prior exits/failures (cite)
​
Problem & Solution​
• Customer pain point (quantify if possible).
• Product / tech approach and why it’s differentiated.
​
Market Size (TAM/SAM/SOM)​
• Numeric TAM estimate (USD), year, and source link (cite)
​
Business Model & Unit Economics (table)
Metric
Current Estimate
Source / Assumption
ACV / ARPU
Gross Margin %
LTV:CAC
Burn Rate
​
Traction​
• KPIs disclosed (ARR, users, pilots, waitlist, etc.) (cite)
​
Competitive Landscape​
• Key direct competitors and their funding stage.
• Substitute solutions / incumbents.
​
Key Risks​
• Tech, regulatory, GTM, capital intensity, talent, etc.
​
Tier Recommendation​
​Full DD / Light DD / Pass — plus <20-word justification.
​
4. Research Rules​
▸ Pull numbers no older than 24 months unless unavoidable; include publish date for every figure.
▸ Hyperlink every source inline (e.g., “$1.2 B TAM [Statista, 2024]”).
▸ If you cannot verify a fact, write “don’t know” instead of guessing.​
▸ If no public data exists, state “No public data found as of <date>
​
5. Length​
▸ Max ~12 bullet lines + one table per company. No hard overall word limit.

Result for Prompt 2:

​https://chatgpt.com/share/680d34d7-5ed0-8001-9a16-2d59de4fa0ce​

The Result - Which Product Hunt Launches Were Interesting on April 26, 2025?

Looking at our "ChatGPT Diligence Report" it indicated the following companies are worth looking further into:

​

What You Can Learn for Future Prompting

  • Use Operator for Manual Tasks: Manual tasks can now be delegated with AI. More powerful front end only platforms like Manus are coming out. And this is even more powerful with some low-code AI platforms too like Cursor.
  • Include Standardization: Adding standardization for a list helps GPT structure answers better.
  • Tell GPT to say "I don't know" if appropriate: Guardrail against hallucinations - if there's no information that is better than GPT making up some information

If you made it this far, forward this to a VC friend or a founder who might find this helpful :)

2026 — Built by Dave Wang. Not financial advice, only for educational purposes.