Guide
How to do competitive research with AI
AI is good at reading a lot of public material quickly. It is less good at keeping five companies straight in one conversation. A sound process plays to the first and protects against the second.
Start from what you need to decide
Competitive research without a question becomes a pile of facts. Write down what the research is for: choosing a vendor, positioning a product, pricing a plan. Then list the criteria that matter. Put that document on the canvas first, because every later step will be measured against it.
Keep each competitor separate
When several companies are discussed in one thread, details migrate. A feature of one gets attributed to another, and by the tenth message nobody can say where a claim came from. Giving each competitor its own branch prevents this. Ask the same set of questions in each, with web search on so pricing and features are current and cited.
Go deep without derailing
Some questions need more than an overview: will this API limit hold at twice our volume, does this integration cover our case. Fork those from the relevant competitor’s branch. The deep dive is recorded where it belongs and the overview stays readable.
Merge, then verify
Merge the competitor branches into one chat and ask for a comparison against your criteria. Before you share it, open the sources behind any claim the decision rests on. Public pages go out of date and models can misread them.
When you send the result to colleagues, share the project instead of pasting the summary. They get the reasoning and the rejected options along with the conclusion.
