The hardest part about working with ChatGPT, Claude, and Gemini is getting the prompt just right. If you’re too specific, the AI may give you a narrow answer that misses the big picture. If you’re too vague, you’ll end up with a wishy-washy reply. Or maybe you’re asking the model to solve a problem that doesn’t actually need fixing.
Of course, there are AI experts who devote their careers to crafting the best prompts to make an AI truly “sing,” delivering just the response you were looking for. But most of us aren’t prompt engineers, nor do we have the time for extended trial-and-error sessions. We just need help right now.
So, if you don’t have the wherewithal to create the perfect AI prompt, here’s a handy shortcut: asking the AI for help.
Give this prompt a shot, placing it just after your initial prompt:
Before you answer my request, critique the question itself. Is this the right question to be asking to achieve [my goal]? If not, suggest a better question and explain why.
This “right-question meta-hack” does something that sets it apart from the usual “help me craft a prompt that does ‘x’” request: it breaks the AI out of its “I must be helpful at all costs” rut, which drives the model to do your bidding even if your prompt is unfocused or attempts to tackle a problem from the wrong angle.
I tried this “right question” meta-prompt (here’s another version, and another) using ChatGPT, with this relatively low-stakes (and admittedly nerdy) question:
What’s the current state of Diablo 4? Give me details on player count and happiness, current season (when did it start/when will it end), latest expansion, reviews of expansion, and so on.
When I asked without the “right question” meta prompt, here’s the beginning of the lengthy, meandering answer I got:
The short version: Diablo IV is in a much healthier place than it was at launch, but it’s still a somewhat polarizing game depending on what kind of Diablo player you are.
The overall vibe in 2026 is probably best described as: “Blizzard finally found the treadmill.” The game now has a much stronger endgame loop, more build variety, better loot progression, and more reasons to keep playing season-to-season. At the same time, some longtime fans still think it leans too heavily into fast-paced, Diablo III-style scaling and seasonal power creep.
And here’s what I got after adding the meta prompt:
Your question is good, but it mixes together several different goals:
“Is Diablo 4 alive?”
“Is Diablo 4 good now?”
“Is Diablo 4 worth returning to?”
“What’s the actual community mood versus the online outrage cycle?”
Those are related, but not identical.
If your real goal is deciding whether to jump back in (or start fresh), a sharper question would be something like:
“Is Diablo 4 in a healthy long-term state in 2026, and is it worth investing time in right now compared to other ARPGs like Path of Exile 2 or Last Epoch?”
Why that’s a better framing:
Raw player count alone is misleading for Blizzard games because most players are on Battle.net, not Steam.
“Happiness” is vague unless you separate launch-era complaints from the current state.
Seasonal ARPGs naturally spike and dip, so a temporary lull doesn’t necessarily mean the game is dying.
The more important question is whether Blizzard solved the core problems people complained about at launch: itemization, endgame, build variety, pacing, and progression.
That second answer yielded a much more focused answer to the question I was really asking: Is Diablo 4 worth playing right now?
The “right question” meta prompt did a few interesting things here: It parsed my question, challenged my choice of metrics about the health of Diablo 4, and then it reframed my prompt, resulting in (to my mind) much better reply.
The prompt works best with broad, wide-ranging requests, and it’s particularly helpful for honing “deep research” prompts. On the flip side, the “right question” prompt is a waste of time (and tokens) if you’ve already formulated a narrow, focused prompt for a specific task.
Also, don’t be afraid to repeat the “right question” prompt, asking “is this the right question?” again after getting an initial refinement.
