If you’re just using ChatGPT Deep Research as an advanced form of search (or not using it at all), you’re missing out! Everyone is underestimating the power of this tool.
This week, I had a new slide deck to prepare. It needed to be cohesive, impactful, and tailored to very specific client goals.
The client agreed the deck could be a selection of my slides from past presentations to meet her needs, but I didn’t have a pre-made deck for this exact talk.
What I did have were almost 2.5 years of random past decks that probably had ideal slides somewhere in them, an agreement outlining the requirements, and notes on the talk’s objectives.
But I didn’t have time to sift through all those old decks to figure out which slides were best.
Enter Deep Research.
I uploaded 7 of my past decks that were most likely to have the best slides, into my ChatGPT Team account (note: I’d NEVER risk that much IP with data training on. And, to be clear, these were my own general talk decks - not custom ones for clients).
Alongside the decks, in my prompt I also gave my notes, and the language in the agreement outlining the length and expectations.
I asked Deep Research to 👇
📌Analyze the decks.Â
📌 Identify slides that aligned with the talk’s objectives.Â
📌Highlight gaps where no existing slide sufficed, and suggest new content by synthesizing my ideas from multiple slides.Â
📌 Generate slide copy and GPT-4o image generator prompts for any new slides needed.
In just 13 min, Deep Research combed through hundreds of slides and delivered a detailed slide plan for the talk.
It pinpointed relevant existing slides, and told me which slides to use from which deck.
More importantly, it ID'd and solved content gaps, and combined elements from multiple slides of MINE from different presentations, to outline new slides using my ideas.
For these new slides, it wrote prompts for GPT-4o to create the slides for me. The prompts included the design AND the exact text.
What amazed me most was how Deep Research didn’t just help me reuse content - it pushed me to improve it.
While I planned to rely on existing slides, it suggested combining specific insights from slides created years apart to create entirely new slides that form a cohesive narrative that met the goals EXACTLY.
Then, I had GPT-4.5 join the convo to refine the new prompts slightly.
Finally, I put each refined prompt into a new chat with GPT-4o to generate the slides.
The slides were so good they only required minor tweaks (if any at all).
Deep Research did all of this for me, without doing any search at all. I don't think it connected to the internet.
I ended up with a cohesive 9-slide deck that felt entirely mine (all the ideas WERE mine).
It is polished, strategic, and perfectly aligned with the client’s expectations.
The client loved it, and I completed the project in a fraction of the time it would’ve taken manually.
And yes, after they saw the deck, I told them exactly how I created it.
This week, I had a new slide deck to prepare. It needed to be cohesive, impactful, and tailored to very specific client goals.
The client agreed the deck could be a selection of my slides from past presentations to meet her needs, but I didn’t have a pre-made deck for this exact talk.
What I did have were almost 2.5 years of random past decks that probably had ideal slides somewhere in them, an agreement outlining the requirements, and notes on the talk’s objectives.
But I didn’t have time to sift through all those old decks to figure out which slides were best.
Enter Deep Research.
I uploaded 7 of my past decks that were most likely to have the best slides, into my ChatGPT Team account (note: I’d NEVER risk that much IP with data training on. And, to be clear, these were my own general talk decks - not custom ones for clients).
Alongside the decks, in my prompt I also gave my notes, and the language in the agreement outlining the length and expectations.
I asked Deep Research to 👇
📌Analyze the decks.Â
📌 Identify slides that aligned with the talk’s objectives.Â
📌Highlight gaps where no existing slide sufficed, and suggest new content by synthesizing my ideas from multiple slides.Â
📌 Generate slide copy and GPT-4o image generator prompts for any new slides needed.
In just 13 min, Deep Research combed through hundreds of slides and delivered a detailed slide plan for the talk.
It pinpointed relevant existing slides, and told me which slides to use from which deck.
More importantly, it ID'd and solved content gaps, and combined elements from multiple slides of MINE from different presentations, to outline new slides using my ideas.
For these new slides, it wrote prompts for GPT-4o to create the slides for me. The prompts included the design AND the exact text.
What amazed me most was how Deep Research didn’t just help me reuse content - it pushed me to improve it.
While I planned to rely on existing slides, it suggested combining specific insights from slides created years apart to create entirely new slides that form a cohesive narrative that met the goals EXACTLY.
Then, I had GPT-4.5 join the convo to refine the new prompts slightly.
Finally, I put each refined prompt into a new chat with GPT-4o to generate the slides.
The slides were so good they only required minor tweaks (if any at all).
Deep Research did all of this for me, without doing any search at all. I don't think it connected to the internet.
I ended up with a cohesive 9-slide deck that felt entirely mine (all the ideas WERE mine).
It is polished, strategic, and perfectly aligned with the client’s expectations.
The client loved it, and I completed the project in a fraction of the time it would’ve taken manually.
And yes, after they saw the deck, I told them exactly how I created it.
