Need to level up on your competitors quickly? Ever wondered what competitors can learn about your product from ChatGPT?
I was recently inspired by Torsten Walbaum's post and his section on prompting AI Deep Research for competitive intelligence. The prompt below is more generic — over the last few months I've evolved it for different audiences, but this one is a great place to start.
After a few months of experimenting, my top tips for other PM teams:
- Torsten's step to "ask AI for its research plan" is a must, especially with Deep Research mode. There are times you want a rougher estimate and times you want AI to comb every paper ever published on a topic.
- Using lighter-weight models for the planning step above can help refine your plan before you feed it into Deep Research or other pro models.
- The market research colleagues I spoke with confirmed Deep Research answers were impressively accurate and useful, with the usual caveats — so always ask it to cite and check the sources.*
- Consistent with the article, GPT Deep Research returned the best findings for competitive intelligence work vs. others, at least as of this summer (2025).
*Even when AI returns source URLs, for important decisions you should hand-check as many as possible for broken links. I'm still curious whether those are hallucinations or just ghost links for content that moved but was present during the model's training. Regardless, you can't use those portions for any significant interactions with other teams. This step becomes the bottleneck pretty quickly, and only increases the importance of your planning phase.
Prompt Template
I am analyzing a competitor product. You will help me distinguish between marketing claims and actual functionality. Here is the competitor I want you to research: [Insert Competitor Name + Product/Module Here] Step 1: Review the company's official website, product documentation, and marketplace listings (if applicable). Step 2: Collect information from analyst reports (e.g., Gartner, Forrester, KLAS), customer reviews (G2, TrustRadius, Gartner Peer Insights), and credible third-party sources. Step 3: Populate a table with the following columns: - Product / Module - Marketing Claim (What they say) - Verified Functionality (What it actually does) - Evidence / Source (cite URLs or documents) - Adoption Evidence (reference customers, industries, scale) - Notes / Risks (limitations, unclear features, heavy customization required) Step 4: Highlight where features are vague, overstated, or lack supporting evidence. Clearly distinguish between out-of-the-box capabilities vs. features requiring add-ons, integrations, or roadmap promises.