I added a single line to my ChatGPT prompts, and the difference was immediate

Published Aug 9, 2026, 3:30 PM EDT Mahnoor Faisal is a tech journalist covering AI and productivity tools with bylines at XDA, SlashGear, MakeUseOf, Laptop Mag, and Android Police. She's been writing professionally since she was sixteen, and has since penned hundreds of articles. This includes in-depth coverage of AI tools like NotebookLM to breaking news across the AI space. Her passion for technology started when she received her first iPod Touch (4th generation) on her 8th birthday, and she's been deep in the tech world ever since. Currently pursuing a degree in computer science, Mahnoor brings both a journalist's eye and a technical foundation to her coverage of how AI is reshaping the way we work and learn. I might be slightly exaggerating, but I think we're at the stage where we write more prompts in a day than we write actual messages to real people. It's depressing, dystopian, and disturbingly accurate. I write prompts from both the perspective of a journalist whose primary beat is AI and a subscriber to nearly every AI tool under the sun. I'm always looking for new ways to squeeze better results out of these very tools I'm paying for. Every so often, one actually delivers, and a single sentence ends up earning a permanent spot at the end of my ChatGPT prompts. I was tired of ChatGPT filling in the blanks Confidently wrong is still wrong Let's leave LLMs aside for just a moment. Imagine if your friend asked you to pick a restaurant for dinner, and they gave you nothing else to work with. No budget, no cuisine, no neighborhood, no mention of whether you're looking for somewhere casual or somewhere you'd actually bother dressing up for. While you can certainly give them an answer, you would probably just start filling in the blanks yourself. You'd begin making assumptions about what they like, how much they're willing to spend, and how far they're willing to travel. Now, you might get lucky and land on exactly the kind of restaurant they had in mind. However, there's just as good a chance you'll suggest something that completely misses the mark. The reason for this isn't necessarily because you have bad taste (though...you might). It's because you were never given enough information to make a genuinely good recommendation in the first place. ChatGPT does the same thing constantly. Give it a vague enough prompt, and it'll happily make dozens of little decisions on your behalf just to get an answer on the screen. Sometimes those assumptions are harmless. Other times, they're the entire reason the response feels generic, irrelevant, or just slightly off. What's even more frustrating is that ChatGPT typically doesn't make it obvious that it's guessing and filling the gaps itself. It won't necessarily stop and tell you that your prompt is missing a budget, a location, a goal, or whatever other detail could dramatically change its answer. It'll simply choose something plausible and keep going. Since every answer it gives arrives neatly formatted and has an LLM's classic tone of unwavering confidence, it's surprisingly easy to mistake those assumptions for conclusions it had enough information to reach. The trick is getting ChatGPT to admit when it needs more information Make it admit what it doesn't know The problem I explained above is fairly easy to solve. Instead of trying to cram every possible detail into my initial prompt (which I realistically can't), I started explicitly telling ChatGPT what to do when I haven't given it enough information. This is a prompt I got courtesy of a Reddit post: Before you answer, assess the uncertainty of your response. If it’s greater than 0.1, ask me clarifying questions until the uncertainty is 0.1 or lower. It sounds a little more scientific than it really is. ChatGPT isn't consulting some perfectly calibrated internal uncertainty meter and determining that it is exactly 0.47 unsure about your request. The number is less important than what the instruction is actually forcing it to do: stop, consider how much information is missing, and decide whether it should ask me questions before answering. To test it, I gave ChatGPT an intentionally vague request: “Plan a 5-day vacation to Japan for me.” On its own, that's an incredibly easy prompt for an LLM to bluff its way through. Tokyo, Kyoto, maybe Osaka, a few shrines, some ramen, and suddenly you've got what looks like a perfectly respectable five-day itinerary. With the extra line attached, ChatGPT didn't try to fill any of that in for me. Instead, it estimated its uncertainty at around 0.5 and asked seven questions before it would build the itinerary. It wanted to know my travel dates, departure city, whether it was my first visit, what I cared about most, my preferred pace, my budget, and whether I had any food requirements or other constraints. That was exactly the behavior I wanted. Every one of those details could materially change the trip. A five-day itinerary for someone obsessed with anime and shopping shouldn't look anything like one for someone who mainly wants temples, nature, and quiet neighborhoods. Even the difference between five full days in Japan and five days including two travel days could completely reshape the plan. Instead of confidently deciding all of that for me, ChatGPT finally admitted that it didn't know enough yet. The prompt you start with can make just as much of a difference as the AI assistant you choose. While you don’t need to overengineer every request, giving ChatGPT clear constraints, context, and instructions about how you want it to handle uncertainty can lead to a much more useful answer.