Effective Prompting Techniques for AI Deep Research

deep research is the most underused tool in AI right now... not because it doesn't work, because people can't prompt it here's how to get better outputs: > context first: who's reading this? how is it going to be implemented? this is the most important step as it'll you get you completely different results based on the situation > source control: left on its own it'll scrape surface-level content all day, but if you point it at specific and unusual sources... this is when you get alpha insights > break it into phases: split your research into 3-4 stages, run each one separately, let each phase build on the last... this is where the depth comes from >define your output: tell it exactly how to structure the deliverable because if you don't, you'll get a 4,000 of barely readable slop nail these 4 components and you'll find more than you could on your own