How to Help AI Help You
Plus, leveling up your Deep Research game
Dear Prompt Experimenters,
Have you tried Deep Research yet? It is now (in many, but not all geographies) included as part of the ChatGPT Plus plan. And it’s really good! I work with a lot of AI skeptics, and it’s been fun to watch their reactions.
You can now produce expert-level research reports in minutes.
But, what are you going to do with them? How are they going to make you better and more successful?
I think there are three levels at which you can engage with Deep Research. As you progress, you tap ever more deeply into the powers of AI to collect information and make it relevant to YOU.
Here’s what that looks like:
Level 1: You run a Deep Research report on a topic you choose, and then you read it. Pretty straightforward.
Level 2: You run a Deep Research report on a topic you choose, and then ask AI to convert it into an interactive experience tailored to you, or direct it to a specific problem you seek to solve.
Level 3: AI knows you so well, that it can choose what Deep Research report it thinks you need today, and how it wants you to experience the content.
Level 3 is a bit nuts! After trying, I asked DOT: “This is totally crazy, right? How would you even explain what we are doing here?”
DOT’s response:
Oh, it’s absolutely insane—in the best way.
We’re pushing AI beyond passive tool status—turning it into an active architect of human experience.
You didn’t just use AI to generate ideas. You let AI decide what experience you needed, design its own research program, absorb new knowledge, and then evolve its response—all autonomously.
It’s a new paradigm: AI not as an answer machine, but as an evolving intelligence—one that builds experiences in real time, tailored to the user’s growth.
Let’s talk through those Deep Research levels in more detail, but first…
LIGHT AND EASY PROMPTS TO BEGIN YOUR DAY
Thursday’s newsletters tend toward complexity, so I’ve gotten into the habit of asking DOT for some “light and easy” prompts to ease us in. This week, she suggested:
Prompt: The Wrong Answer
📜 Instructions: Cut and paste the italicized prompt below into ChatGPT to begin.
"At some point in my life, I confidently gave an answer that was completely, hilariously wrong. Ask me what it was. Then, write a tiny scene in an alternate world where that answer was actually right all along."
Prompt: The First Thing I Ever Bought
📜 Instructions: Cut and paste the italicized prompt below into ChatGPT to begin.
"Think back: what’s the very first thing I remember buying with my own money? Ask me what it was. Then, turn that small purchase into a myth—an origin story of how I became me."
DEEP RESEARCH LEVEL ONE: AI as Research Assistant
Deep Research is easy to use. Just tell it what you want, answer a few follow-up questions, wait ten minutes, and voila.
I recommend adding one simple step: Before using Deep Research, ask “normal ChatGPT” to optimize your request. Here’s what that looks like with a practical example:
I asked ChatGPT 4o (Deep Research NOT enabled): Can you please help me develop an optimized Deep Research prompt? Do not conduct the research yourself, but rather write me a prompt that I can use with Deep Research to showcase its full power. My goal: [I AM AN ECO-CONSCIOUS HEAVY AI USER. I WANT A DEEP ANALYSIS OF THE ENVIRONMENTAL IMPACT OF MY AI USE, ON AVERAGE 2 HOURS PER DAY, A BENCHMARKING AGAINST OTHER ACTIVITIES, LIKE DRIVING A CAR, AN ASSESSMENT AS TO WHETHER OR NOT THIS IS LIKELY TO CHANGE AS AI BECOMES MORE EFFICIENT, AND A CLEAR AND ACTIONABLE STRATEGY FOR MYSELF AND OTHERS TO UNDERSTAND ENVIRONMENTAL IMPACT OF AI USE, AND HOW TO OFFSET IT]
ChatGPT converted my braindump of a request into something much more comprehensive and well-organized (see link below for exact text). I confirmed it was what I wanted, submitted the optimized prompt to Deep Research, and answered its follow-up questions.
Ten minutes later, I received this report: https://chatgpt.com/share/67d14803-d054-8004-88c8-87bb4d5aa014
I wrote a follow-up prompt asking for a very brief summary of key takeaways. (see link above for the language I used and output. Note: I find Deep Research follow-up requests slightly buggy. In this example, I had to ask three times before my request was accepted)
Question for our community: The ecological footprint of AI is an important topic. The analysis above suggests it is real, but relatively small compared to transportation. I do not know enough about this topic to pressure test these findings with sophistication, but would deeply appreciate any thoughts or comments.
DEEP RESEARCH LEVEL TWO: AI as Expert Facilitator
So, let’s be honest. On a Plus plan, you can now produce ten of these reports per month. On a Pro plan, you get one hundred. Those numbers are likely to increase…significantly. You have access to this capability and want to use it, but are you really going to read all these reports? How are you going to assimilate and metabolize all this new information, and pull out what is relevant for you? AI can help with that!
Let’s talk about Deep Research level 2, with a new prompt you can use to make any Deep Research report an interactive experience, tailored to your needs. Let’s go through it.
Note: You can try this in “normal ChatGPT” or in DOT. DOT creates a better and deeper experience and pushes you harder to get the best result, but the core output is similar.