How Deep Research will change investigative journalism

Both OpenAI and Google offer powerful AI research tools, and journalists are well positioned to benefit — as long as they know how to use them.

OpenAI Deep Research is here. The tool — which can perform extensive online research and give you a thorough report in a matter of minutes — has obvious implications for journalists and media researchers. Personally, I'm just glad the AI world has something else to talk about besides DeepSeek, at least for a minute.

Using Deep Research isn't just a simple matter of tossing it a prompt and grabbing a coffee. Making use of these tools (Google has one, too) requires a more methodical approach, and that’s the topic of today's column. Many of the fundamentals of that approach stem from the AI training sessions I offer to journalists, PR pros, and content creators, and, as a side note, I’m thrilled that AI for PR & Media Professionals begins today. If you’re interested in a last-minute spot, hit me up.

I'm also looking forward to speaking at State of Digital Publishing's WP Publisher Success Week on Feb. 28. I had a fantastic experience last year speaking about the opportunities for publishers. Things obviously look very different today, and I'm excited to join Matt Karolian from The Boston Globe to talk about some real-world strategies for using AI in newsrooms in 2025. To make sure you don't miss it, sign up via the link below 👇

---

The risks and rewards of AI-driven investigative journalism

Sam Altman promised that OpenAI would “pull up” some releases in the wake of DeepSeek, and they've appeared to do exactly that, launching Deep Research over the weekend. The tool, which can tackle projects that require thoroughly reading lots of source material, processing all the relevant parts, and coming up with a specific set of conclusions, has obvious and tremendous utility — especially for investigative journalists.

It's also a straight-up copy of what the identically named feature of Gemini does. This isn't shocking — tech companies always steal the successful ideas of their competitors (whatever you think of TikTok, you can thank it for all the vertical videos that are now everywhere).

Google Deep Research was an instant hit when it debuted in the fall, and its usefulness prompted me to resume my Gemini subscription (which I had let lapse due to its disappointing performance in Gmail and Google Workspace—more on that in a future column). Disappointingly, OpenAI Deep Research is only available if you shell out for the $200/month Pro subscription, though Altman promised it would come to Plus users “soon.”

I think these research tools represent a leap forward in our use of AI for journalism, possibly even more than multimodal models. However, to truly get the most out of them, users need to get two things right: structured prompting, and a human-driven (but AI-assisted) process for verification of the results.

Let's go over each of those and how journalists can make the most out of this tool.