One of the easiest ways to get a bad answer from AI is to ask it a question that sounds simple but actually isn’t.
“What should I do?”
“What’s the best option?”
“Which one should I pick?”
Those feel natural. They are also where a lot of people quietly get burned.
I explored to try to get to the root of the issue.
AI is often much better at comparison than judgment.
That sounds backwards at first. These systems can write essays, summarize reports, and produce very confident recommendations. So why not just ask for the answer?
Because the moment you ask for “the best” option, the system tends to collapse all the tradeoffs too early. It stops exploring the shape of the problem and starts performing certainty.
That is where things get slippery.
A lot of real decisions do not have one clean right answer. They have tradeoffs.
The cheapest option may be slower.
The fastest option may be riskier.
The most polished response may not be the most honest one.
The technically correct choice may still be wrong for your priorities.
Humans know this instinctively, even when we pretend otherwise. We live inside tradeoffs all the time. But AI is very good at taking a messy field of tradeoffs and flattening it into a neat recommendation that sounds more settled than it really is.
That is why one of the most useful habits I know is this:
Don’t ask AI to decide first. Ask it to compare first.
That small shift changes the quality of the interaction immediately.
Instead of:
“Which laptop should I buy?”
Try:
“Compare these three laptops for battery life, portability, repairability, and long-term value. Show the tradeoffs. Tell me who each one is best for. Then tell me what extra fact would change the ranking.”
Instead of:
“Write the best reply to this email.”
Try:
“Give me three reply options: one warm, one direct, and one low-risk. Explain the tradeoff each one makes.”
Instead of:
“What should I do about this problem?”
Try:
“Give me the strongest case for three different approaches. What does each optimize for, and what does each sacrifice?”
That is where AI becomes genuinely useful.
Not as an oracle.
As a contrast machine.
When you ask for comparisons, the system has less room to bluff a single polished answer into sounding inevitable. It has to show its work a little more. It has to expose the shape of the decision instead of just handing you a verdict.
That matters because most bad AI use does not come from spectacular mistakes. It comes from subtle overcompression. The model gives one tidy answer. You act on it. Later you realize the answer was not exactly wrong — it just quietly optimized for the wrong thing.
Comparison helps prevent that.
It also does something else important: it keeps you in the loop.
A good comparison forces you to clarify your own priorities. Do you care more about speed or certainty? Simplicity or control? Cost or quality? Privacy or convenience? AI cannot answer those for you. It can only help you see them more clearly.
If you want one practical template to use this week, use this:
Compare these options using the criteria that matter most. Show the tradeoffs clearly. State the assumptions. Tell me what each option is best for, what it risks, and what new information would change the recommendation. Do not pick a winner until after the comparison.
That one prompt will get you farther than a lot of fancier tricks.
The mistake people keep making is using AI like a vending machine for certainty. Put in a question, get out the answer.
The better use is often much simpler.
Ask it to widen the picture before you let it narrow the choice.
That is where AI starts to become genuinely useful: not when it replaces your judgment, but when it helps you see the decision more clearly before you use it.