How to Find New Clients with ChatGPT’s Deep Research

The latest thinking on AI integration for established experts — direct to your inbox.
Mark and Andy - Founders

Your Next Client Might Already Be Findable — You Just Haven’t Run the Right Search Yet

When work slows down, most consultants and advisors either wait for a referral or spend an hour on LinkedIn getting nowhere. Neither is a strategy. Both leave you dependent on timing you can’t control.

ChatGPT’s Deep Research changes this. It can take your ideal client profile and return a ranked list of real companies that match it — with scoring criteria it derives itself, trigger events for why they might engage now, and specific guidance on who to speak to and how to open the conversation.

If you have a ChatGPT Plus or Teams subscription, you already have access. No add-ons required. Here’s how to use it properly.

What You Need Before You Start

The foundation of this process is your ideal client profile (ICP). This is the document that describes exactly who you want to work with — the type of organisation, the problems they face, the triggers that make them ready to engage, and what good looks like for both sides.

If you don’t have one, it’s worth building before running this process. The quality of the output depends entirely on the quality of the ICP you feed in. A vague description gets vague results. A specific, detailed profile gets something genuinely useful.

How the Deep Research Prompt Works

Once you have your ICP, open ChatGPT and select Deep Research from the left-hand panel. Upload your ICP document, then paste in the prompt (linked at the bottom of this post).

The prompt does three things:

  • It tells the agent to derive its own scoring framework based on your ICP — not generic criteria, but criteria grounded in what you’ve described
  • It asks for 10 to 15 real companies, researched and ranked against that framework
  • It specifies the output format: a ranked table, detailed analysis per company, emerging patterns, and ICP refinement suggestions

Once you hit start, leave it running. This process takes between 40 minutes and two hours. It’s researching several hundred websites during that time. You don’t need to watch it — go do something else and come back.

What the Output Looks Like

The report opens with the scoring framework the agent built from your ICP. For a leadership communication consultancy, for example, it might surface criteria like change pressure with operational consequences, manager capability strain, and stakeholder complexity. These aren’t categories you gave it — it derived them from your description of who you work best with.

Below that is the ranked lead table: 10 to 15 real companies, each with an overall fit score, a brief explanation of why they match, and a key trigger event — a reason they might be looking to engage with someone like you right now.

Scroll further and you get detailed analysis for each company: what they do, why they fit the ICP, who the most likely buyer is, and a suggested conversation hook — the kind of language likely to land with that particular organisation based on what the research found.

The Sections Most People Skip

At the bottom of the report are two sections worth paying attention to.

The first covers emerging patterns and market timing signals — broad trends affecting the companies on the list, including mergers, organisational changes, and shifts in how they’re working. This is context for when and how to reach out, not just who to reach out to.

The second is ICP refinement suggestions. The agent will flag areas where your profile could be more specific, or where the research surfaced patterns you hadn’t anticipated. This is how the process improves over time — each run gives you better input for the next one.

A Note on Transparency

The prompt specifically asks the agent to separate observed facts from informed interpretation. Where it’s drawing conclusions rather than reporting what it found directly, it says so. That’s worth knowing when you’re deciding how much weight to put on any particular insight.

The research isn’t infallible. Companies change. Information goes out of date. But as a starting point for a targeted outreach list — one that’s grounded in your actual ICP rather than a generic keyword search — it’s significantly better than what most experts are doing.

← How to Stop One LLM From Becoming a Single Point of Failure How to create a Custom GPT in ChatGPT — and what to actually use it for →