Why prompt engineering matters more than ever

Deep research tools from OpenAI, Perplexity, and Google have elevated the need to be thoughtful about what we say to AI.

With the rise of deep research tools, communicating with AI is becoming a more precise exercise. (Credit: Midjourney)

As AI tools continue to proliferate, predictions about how we'll use them are getting equally bold. One of the oft-repeated prophecies about the future of AI is that, sometime soon, prompt engineering will go away. Models will simply get so good at inferring what we want that spending brain power on crafting the perfect prompt — perhaps with one of the many templated frameworks — will become a thing of the past.

I'm now convinced this is incorrect. In fact, my experience is that we're going in the other direction: that the more advanced AI tools become, prompt engineering matters more than ever. The most powerful AI tools today are reasoning (a.k.a. thinking) models, which power the new deep research tools from Google, Perplexity, and OpenAI. To use these tools effectively, you need to be precise in what you ask for, and that requires choosing your words carefully.

I was jolted into this reality recently by OpenAI's version of deep research, which by most reports is the most thorough version of the tool. It's also the most expensive, briefly only accessible to the $200-a-month ChatGPT Pro plan, but now available to all paying users, including those on the Plus tier ($20 a month). However, OpenAI puts limits on every level — even Pro users get only 120 requests per month. Plus users max out at just ten.

The need for precision

If you only have 10 tries at something per month, you're going to want to make them count. Not to mention these research reports take time, which is perhaps an even greater investment. So when I asked ChatGPT to perform some deep research on how journalists are using AI and how their adoption compares with the workforce at large, I wanted the tool to get it right the first time.

---

🛑 LAST CHANCE to Save on AI for Media & PR—Next Cohort Starts Mar. 18!

AI is changing the PR playbook, making media monitoring, content creation, and crisis management faster and smarter. But here’s the catch: basic prompting isn’t enough. If you’re just asking ChatGPT to draft press releases, you’re barely scratching the surface.

In AI for PR & Media Professionals, you’ll go beyond the basics—learning how to build AI-powered workflows that streamline your work without sacrificing strategy or creativity. Over six weeks, you’ll develop hands-on skills to automate tasks, analyze trends, and create high-impact campaigns with AI.

📆 Next cohort starts March 18 | Live training + personalized coaching

💡 Enroll now and save, but hurry—the advance discount expires TONIGHT! Use code ADVANCE15 at checkout.

---

It's not like I just dashed off the first thing that popped into my head. I spent a few minutes thinking about what I wanted, and gathered what I thought was the proper context: a Pew Research report about current workplace trends in AI and the recent Reuters institute report on journalism tech trends for 2025, the latter being an example of the kind of research I was looking for.

Here's the exact prompt I used: