Let AI help me judge: is this advice for me?
When AI says it's a deal, that's usually a deal for the most ideal kind of person. Spell out your hours, the commute, and your hard line — then it starts doing the math for you.
Chibi Alice's yearly-pass flop: AI's “only ¥224 a month” becomes “¥672 per visit” once overtime and distance show up
The gym front desk slides the calculator toward you: a yearly pass for ¥2688, which works out to only ¥224 a month, plus two extra months if you pay tonight; monthly is ¥399. The salesperson's payment code is already standing on the corner of the desk, and they keep reminding you the deal ends at 10 tonight.
You can't decide, so you message AI: “Is this gym yearly pass worth it?” It fires back three reasons: cheaper per month, machines and group classes whenever you want, and paying up front makes it easier to stick with it. You can't poke a hole in any sentence.
You feel settled, and you scan to pay on the spot. Three months later, the door log shows you went only 4 times. Do the math, and each visit actually cost more than ¥600.
Pick one of the three common ways to ask, and see how Alice answers.
When you're making a call, you need at least three kinds of facts that can flip the conclusion: real habits, follow-through friction, a stop-loss line. Wishes, sales talk, and “I'll definitely stick with it” look useful. They don't prove you can do it.
Same yearly-pass price, same AI model. Tweak how you ask, and the rundown forks in the opposite direction:
First way to ask (price and a wish, nothing else): you only write “Yearly pass ¥2688. Is it worth it?” Faced with a blank, AI fills in an ideal stand-in who lives nearby, never works late, and goes three times a week. That stand-in goes 120 times a year — ¥22 a visit. The math has no holes. The conclusion only belongs to that stand-in who never gets tired.
Second way to ask (hours, commute, and real habits filled in): you spell out “I often leave after 8:30, it's 35 minutes one way, I only trained 3 times in two months, and the yearly pass can't be refunded.” AI has no room to call in the stand-in. It has to face the physical drag: leave at 8:30, arrive after 9, only a few dozen minutes before close; past attendance already proves the drag is huge. Recalculate, and the conclusion forks at once: do not buy the yearly pass. Pay per visit, or try one month.
Leave a blank in the question, and AI fills it with the most ideal stand-in. It didn't get the math wrong. The blank in the question misled it.
To get AI actually doing the math from your side of the table, hand it three things first:
- Real habits, not a nice wish: when you can actually leave your desk, and how many times you really put on shoes and worked out in the last two months.
- Follow-through friction — count the in-transit cost: how long one way takes, and what time the class you want starts taking attendance.
- A stop-loss line — name the worst case: whether you can pause or get a refund, and whether it would eat the sleep you won't give up at night.
Those three stakes draw a line the rundown is not allowed to cross, so it doesn't do the math in a vacuum.
The first block goes at the start of the question — it sets the rules. The second goes after the answer — it squeezes out the generic boilerplate hiding inside. For any big purchase or choice, swap in the specifics and it still works.
- When you ask, shut down generic boilerplate
I'm deciding whether to 【buy this ¥2688 yearly pass】. Here's my actual situation: I often leave work after 8:30, it's 35 minutes one way, and I only worked out 3 times in the last two months. Hard line: I can't let this eat into sleep after 11:30, and there's no refund after I pay. Don't give me generic pros and cons. Check my time, distance, and track record line by line, and estimate the real chance this money actually pays off for me. If you're missing information, ask me first.
- After the answer, squeeze out the water
Look back at the advice you just gave. Which conclusions only hold in an “ideal state where I show up on a steady schedule”? Flag those generic assumptions on their own, then advise again using only the real constraints I gave you.
In the first sentence, “check my conditions line by line, and estimate the real chance this actually lands” is the core. With that line in, it can't fob you off with “machines whenever you want.” It has to face the 35-minute one-way trip.
“If you're missing information, ask me first” matters too. Real life is messy; the first pass rarely gets every limit down. Let it ask first, and hidden costs — cancellation rules, the personal-trainer pitch — get turned over one by one before you pay.