How journalists can use deep research to find sources, fast
An AI-powered method for uncovering real experts and insights—without getting lost in search clutter.
Finding credible experts online shouldn’t feel like a full-time job—but it often does. Even when you step just a little outside your usual network, the hunt can eat up hours of scrolling, verifying, and cross-checking. Directories help, but only so much. That’s why deep research tools are such a breakthrough: they help media professionals cut through the noise, find the right people fast, and give you back the time you actually need to think.
That’s why, in our most recent AI for Journalists class, we teach a detailed, repeatable way to use these tools to create source lists that reporters can vet easily, saving considerable time, and concentrating human attention where it’s really needed.
Before we get into the workflow, a quick heads-up: this comes straight from our live courses, AI for PR & Communications Professionals and AI for Journalists, which both kick off the first week of November. There’s still time to grab the 25% early enrollment discount (ends after Oct. 6) with the codes AIPR25OFF-NOV (PR/comms) and AIJO25OFF-NOV (journalists).
You can learn more about the courses via the links below, then let’s take a closer look at the precise prompting to get the most out of your research query.
The Copilot uses generative AI to help Media Copilot editors create drafts of articles, allowing us to cover more stories at the intersection of AI and media. Media Copilot editors carefully edit every Copilot story for accuracy, and often add relevant details, context, and links.
What “deep research” actually is
Think of deep research as not just turbocharged search, but actually having an assistant who will thoroughly read all the sites you want to check out on a topic in record time. It’s a new class of search tools—including ChatGPT’s Deep Research, Perplexity Research mode, and Gemini Deep Research—that use reasoning models to build a research plan, ask clarifying questions, and deliver source-linked results.
Instead of surfacing whatever ranks highest, deep research evaluates credibility, synthesizes context, and shows its work. It’s slower than a regular query, but that’s because it’s prioritizing thoroughness over speed.
When you’re tracking a niche issue, vetting experts, or preparing a briefing, Deep Research gives you what Google can’t: context, credibility, and depth.
🔬 The Deep Research Drill
Here’s the workflow we teach in class. We recommend you try it yourself in Perplexity Research mode since it’s the fastest, and Perplexity is quite generous with the number of queries it offers on the free plan, though Gemini or ChatGPT work, too.
Step 1: Frame your goal, not your query
Forget keywords. Start with your intent, with the proper context of what you’re doing—in this case, a journalist building a source list. Before you come up with your deep research prompt, frame the broad outlines of your goal. Here’s an example of where to start:
“I’m an investigative journalist building a credible source list for a story on AI misinformation. I need 10 qualified experts who’ve published or appeared publicly since 2022.”
This isn’t the full prompt, but it immediately gives the AI a persona and an objective. It sets expectations for the output—not just what to search for, but why.
Step 2: Establish your parameters
From here, you can build the rest of the prompt by specifying exactly how you want the information presented. Here, it’s good to steer the AI to something specific, or it will probably default to a research report. In this case, you want something more scannable and organized so it’s easy to vet. A table makes the most sense:
“Create a a structured table with the experts in alphabetical order by last name. Include name, title/affiliation, sector, notable recent work, and link. Exclude influencers or anonymous accounts.”
Here’s a little something we like to add to outputs like this:
“For any missing field, return
<NA>.”
This provides a little extra insurance that the AI won’t hallucinate and make something up entirely when it can’t find the information.
Step 3: Prompt and audit
Feed your prompt into the deep research tool, answer any clarifying questions, and await your results.
Now you do what machines can’t: judge.
Are these real experts?
Are the sources credible?
Are the affiliations current?
Flag weak entries, then follow up:
“Show me recent quotes or interviews from these experts.”
Now you’ve got usable, on-record material ready for pitching, fact-checking, or story framing.
Step 4: Iterate
Once you’ve got a solid list, shift the angle.
“Identify five experts outside the usual media circles.”
“Add voices from different regions or specialties to broaden the perspective.”
In minutes, you’ll have a well-rounded roster you can hand to an editor or client, diverse in thought and geography, not just title.
Step 5: Save your findings
Copy the table to Google Sheets or Notion, tag by topic, and save it to your archive. In class, we pair this with NotebookLM for fueling more material like creating interview questions, cross-referencing areas of expertise, and authoring outreach emails.
Why this matters
Every journalist and comms pro has their usual go-to sources. Nothing wrong with that, but when you’re on deadline, they can sometimes keep you in your comfort zone when the story would be better served by a deeper bench. Deep research helps break that pattern. It’s like having an assistant who can surface the right people on the edge of your beat—the policy analyst in Nairobi, the academic in Helsinki, the niche expert you didn’t know existed. It’s a smarter, faster way to widen your net and strengthen every story or campaign.
Learn the full workflow
This tutorial is excerpted from our live six-week programs, AI for PR & Communications Professionals and AI for Journalists. Both courses start in the first week of November, and they are loaded with use cases like this. Also, you get two 1-on-1 coaching sessions that lead into a capstone project, where you’ll apply what you’ve learned to build your own tools and workflows. We’re focused on transforming the way you work with on-call intelligence, not just showing fancy demos (although the demos are fancy).
Important: Our early-bird discount ends after Oct. 6. Enroll now to save 25%. Grab your spot now right here:
This article is AI-assisted. Media Copilot editors used generative AI to help create a draft, then carefully edited the article for accuracy while adding relevant details, context, and links.