I use AI every day — but writing prompts still slow me down. Even with a background in prompt engineering, it’s hard to keep up. Just when I feel like I’ve figured it out, the models update, the responses shift and suddenly the same prompts don’t hit the same way.
It’s not that I don’t know how to prompt — I do. But constantly trying to phrase things “the right way” adds bottlenecks to productivity fast. Instead of getting answers from ChatGPT, Claude or Gemini, I’m tweaking wording, adding context and second-guessing myself.
So instead of trying new prompts, I tried something different. I tested a few simple ways to make prompting easier — and what happened surprised me.
1. Start messy (and let AI clean it up)
This is perhaps the biggest shift but also the easiest. Instead of trying to write the “perfect” prompt, I started typing exactly what I was thinking, even if it was vague or incomplete. Then I asked the AI to improve it.
With tools like ChatGPT and Google Gemini, this works surprisingly well. For example, I’d type: “Help me make this better” Then let the AI rewrite or expand it.
Result: I saved time by dumping everything on the AI and letting it figured out where I wanted to go with it. When I stopped overthinking, I got better answers faster. One thing to note, if you try this with Claude, the AI will ask follow up questions before generating a response. This gets the guesswork out of the way first, rather than re-prompting based on what the AI could decipher from the notes.
2. Use follow-ups instead of rewriting
You might find that this is where everything starts to click. Instead of starting over every time, I started building on responses. This is similar to my 3-rule prompt, but less structured. For example, if something wasn’t quite right, I’d say:
“Make this shorter”
“Add more detail”
“Rewrite this in a friendlier tone”
This kept the conversation moving instead of resetting it while avoiding generic responses.
Result: I notcied that just starting and reprompting can help in some situations. You'll get faster improvements without losing context and this works with just about any chatbot.
3. Let the AI ask you questions
This might be the most underrated trick. While ChatGPT's endless follow up questions can make even the most patient user's head spin, this is a bit different. Instead of guessing what details to include, I started asking the AI:
“What do you need to know to give a better answer?”
Both ChatGPT and Gemini are surprisingly good at guiding you from there, asking clarifying questions that improve the result. Claude often asks questions before even getting started to gain clarity from the start.
Result: You'll get better outputs by asking the AI questions first. Sometimes, even before asking for a summary or using the "reverse brief," I will ask the AI if it can see the document. By checking in with the AI, it helps eliminate hallucinations or false information right from the start.
4. Use built-in suggestions and templates
The good news is, AI tools are quietly getting better at helping you get started. ChatGPT often suggests follow-up prompts, while Gemini adapts well to simple, conversational input. You don’t always need to start from scratch. Even Claude offers multiple choices to help you get started. Try ChatGPT apps, Gemini Gems and Claude Skills to make the most of your projects.
Result: You'll probably discover you have less “blank page” anxiety and get faster at simply beginning your project.
The takeaway
If you're struggling with knowing which AI tool or chatbot to use. Know that it might not be the tool at all. Changing how you use them and making the most of prompts can mean better results in less time.
There’s a misconception that you need to master prompt engineering to get good results from AI. But in many ways, prompt engineering is a fading skill becasue the models are updating faster than engineers can keep up.
The key is to start simple, iterate naturally and let the AI refine things along the way.
More from Tom's Guide
- Google’s ‘Personal Intelligence’ connects your emails, photos and purchases — and it could change how you use AI forever
- I tried the 'viral' ChatGPT prompts everyone is sharing — here's what actually works
- I wore Meta Ray-Ban smart glasses while shopping at Target — 7 things that were surprisingly useful
