prompt teardown.
HomeBlogArchive Subscribe
Home Blog Archive Subscribe
RSS feed Reddit community
guide 10 min read
Home › Blog › 3 AI prompts you'll wish you saved before you need them

3 AI prompts you'll wish you saved before you need them

NewsletterProductivityFinanceResearch

The Take

Nobody goes looking for a doctor prompt until they’re already in the waiting room.

That’s true of all 3 this week. You probably don’t need them today. But one of them will come up, and you won’t have time to look then.

They give you what you forgot to ask about. Then they wait.

So save this one. A lease comes up. A test result comes back. Someone gives you advice and sounds certain.

Seen This Week

The prompt that writes your questions for the doctor

Weeks to get the appointment. Now you’re 15 minutes from the door.

You already know not to ask AI what’s wrong with you. Wrong diagnosis, wrong prognosis, and you carry it into the room with you.

So don’t. Give it the mess and let it find the questions you’d forget to ask.

I'm preparing for a medical appointment, either mine or one I'm taking someone to. Do not diagnose, and do not guess at what's wrong.

Here's what's going on: [symptoms, how long, what's been tried, any test results]

Organize it into:
1. A short timeline, in order. If this is a follow-up, focus on what's changed since the last visit.
2. Context worth mentioning: medications, family history, recent changes
3. Questions to ask, most important first
4. What to bring: records, test results, a list of medications
5. Signs that would mean not waiting for this appointment

For 5, give me specific changes to watch for, not possible conditions. If what I've described already sounds like it needs care now, say that first.

What each line is doing

  • “Do not diagnose, and do not guess at what’s wrong.” Take this out and the answer opens with a list of conditions. Then you spend your 15 minutes asking about the wrong one.
  • “Either mine or one I’m taking someone to.” A lot of appointment prep is for a parent, a kid, or a partner who won’t ask. This one clause makes the whole prompt work for the person doing the asking.
  • “If this is a follow-up, focus on what’s changed.” A timeline built from scratch is the wrong thing to hand a doctor on the third visit about the same problem. What they need then is the difference.
  • Item 5, and the line under it. Ask for red flags and you get a list of diseases. Ask for changes to watch for and you get a threshold you can act on.
  • “Say that first.” Urgency moves to the top of the answer instead of sitting at number 5 where you might skim past it.

Why it works: You can describe what’s happening to you. Turning that into the right questions is the hard part. It builds the list you’d only think of on the drive home. Item 5 does the other job. It tells you when not to wait for the appointment at all.

One thing: It writes you 12 questions. You’ll get through 3.

Simple fix: Tell it how long the appointment is and ask it to cut the list to fit.

The prompt that finds the hidden costs of moving

The new place has an extra room and costs $50 less. You ran the numbers twice on your phone.

Both times you left out first and last, the deposit, the truck, and the fee the internet company charges just to show up.

Those wipe out a year of the savings.

Swap in a job offer, a phone plan, or switching insurance. Anything where the price you see leaves things out.

Turn on web search before you run this. It needs to look up what these things cost where you live.

I'm deciding whether to switch or stay put, and I want the costs I'm not thinking about.

Where I am: [city, or zip code]
Staying: [what you pay now, what you get]
Switching: [what you'd pay, what you'd get]

Search the web for current costs in my area before you answer. Local fees are the point, don't estimate them from memory.

1. Find the one-time costs of switching that people forget, for my location. Deposits, utility activation, internet or cable install, building or HOA fees, agent and broker fees.

2. Ask what I'm avoiding by switching. Some costs disappear, like an agent fee I don't have to pay, and those count against the total.

3. List the ongoing costs that change, including ones that never show up on a bill. If my commute changes, ask for the distance and look up current fuel prices near me to price it. If I'm moving somewhere smaller, ask what won't fit and price storage.

4. Ask how long I plan to stay. If I don't know, use 2 years.

Then give me:
- The real first-year cost of each, one-time costs and savings included
- Break-even: how long the monthly difference takes to cover the one-time costs
- What I gain or give up that isn't money, and what that's worth against the number

Use my numbers where I gave them, searched numbers where you found them, and label anything you estimated. Show your math.

What each line is doing

  • “Where I am.” The whole prompt runs on this. A deposit rule in one state and a transfer fee in another are different numbers, and an answer without your location is an average of places you don’t live.
  • “Search the web, don’t estimate from memory.” Without this it answers from training data. Fees it learned 2 years ago come back sounding current.
  • The named list in step 1. Naming deposits, activation, install, HOA and agent fees stops it returning “consider moving expenses” and makes it go find each one.
  • “Ask what I’m avoiding by switching.” The savings side. Skipping an agent fee, dodging a penalty, dropping a service you were paying for. These come off the total, and almost nobody counts them.
  • The storage line. For anyone moving somewhere smaller. You save $200 a month, then rent a unit for $150 to hold what doesn’t fit.
  • “If I don’t know, use 2 years.” A break-even means nothing without a time frame. This gives it one so the math finishes.
  • “Label anything you estimated.” Your real rent and a guessed HOA fee look the same in a table. This separates what you know from what it filled in.

Why it works: It asks where you live, then looks up real fees instead of guessing at them. It also counts what you save. Not paying an agent can be worth a month’s rent, and a side-by-side of the 2 rents never shows that.

One thing: A fee it finds online might be for the wrong city.

Simple fix: Ask it to name the city and the source for each number it looked up.

The prompt that gets a second opinion on any advice

Someone just told you what to do. A contractor, a video, a guy in a forum, the AI you asked.

It sounded right. It usually does.

Before you act, make it hold up against someone who would actually know.

I'm about to act on this recommendation:

[paste it]


It came from: [a person, a video, an AI, a company selling something]

My situation: [1 or 2 lines]


Pressure-test it before I commit.

1. The hidden assumptions, and what this source gains if I say yes.

2. How it most likely fails for someone in my situation, including who tried this and isn't around to talk about it.

3. The strongest honest case for not doing this, argued by whoever would actually know. Someone who has done it, someone who has been in my position, or someone qualified in that field.

4. What evidence would change your answer.

5. The smallest low-risk test I can run today.


If the advice is sound, say so plainly instead of inventing problems to find.

What each line is doing

  • “It came from.” Advice from a company selling you something breaks in a different place than advice from a friend who tried it once. Naming the source changes the whole answer.
  • “What this source gains if I say yes.” The fastest check there is. A contractor recommending more work, a channel selling a course, a friend who wants company in their own decision.
  • “Who tried this and isn’t around to talk about it.” Advice spreads because the people it worked for repeat it. The ones it sank went quiet. This asks for them.
  • “Argued by whoever would actually know.” Without this you get a generic contrarian flipping your own words back at you. With it, the objection comes from someone who has stood where you’re standing.
  • “What evidence would change your answer.” Makes it name the thing that would prove it wrong, which is the difference between an opinion and a position.
  • “The smallest low-risk test I can run today.” Ends the whole thing in something you do rather than something you worry about.
  • “If the advice is sound, say so plainly.” Ask for hidden problems and it will find 4 whether or not they exist. This gives it permission to tell you the advice is fine.

Why it works: Telling it where the advice came from changes the answer. A contractor who profits from the job and a friend who has done it twice fail in different ways. Item 3 argues back as someone who would actually know, so the objection carries weight.

One thing: It only sees what you paste. Give it your summary of the advice and it will pressure-test your summary.

Simple fix: Paste the advice word for word, even if it runs long.

Related posts

  • Newsletter Plan your money, gamify a book, test on a stranger
  • Newsletter The AI prompt that wins a yes, and one that cites sources
  • Newsletter Rewrite a dead email & plan a week you can finish

Contents

  • The Take
  • Seen This Week
  • The prompt that writes your questions for the doctor
  • The prompt that finds the hidden costs of moving
  • The prompt that gets a second opinion on any advice

Get the next teardown

Free weekly prompt breakdowns.

In this article

  • The Take
  • Seen This Week
  • The prompt that writes your questions for the doctor
  • The prompt that finds the hidden costs of moving
  • The prompt that gets a second opinion on any advice

Frequently asked questions

Why does the doctor prompt tell AI not to diagnose?

Because without that line the answer opens with a list of possible conditions, and you carry the wrong one into the room with you. The job of the prompt is to build your question list, not to guess at what's wrong.

What are the hidden costs people forget when they move?

First and last month, the deposit, utility and internet activation, building or HOA fees, the truck, and agent or broker fees. Those one-time costs routinely wipe out a year of a lower monthly rent.

Why does the moving prompt need web search turned on?

Local fees are the whole point. Without web search the model answers from training data, so a fee it learned 2 years ago comes back sounding current. Tell it to search and to label anything it estimated.

How do I get AI to argue against advice without inventing problems?

Name where the advice came from, ask what that source gains if you say yes, and have the objection argued by someone who would actually know. Then add a line giving it permission to say the advice is sound, so it doesn't manufacture 4 objections that don't exist.

Can I use these prompts for something other than doctors and moving?

Yes. The switch-or-stay prompt works on a job offer, a phone plan, or switching insurance. Anything where the price you see leaves things out.

Weekly Teardowns ✦ Model Agnostic ✦ Steal These Prompts ✦ Weekly Teardowns ✦ Model Agnostic ✦ Steal These Prompts ✦ Weekly Teardowns ✦ Model Agnostic ✦ Steal These Prompts ✦
Prompt Teardown

Weekly prompt teardowns for people who take AI seriously. Model-agnostic.

Site

  • Blog
  • Archive
  • Sponsor
© 2026 Prompt Teardown
Prompt Teardown

Prompt Teardown

Add your email to get weekly prompts in your inbox.

Click subscribe to get started. Tap subscribe to get started.

We hate spam, too.
Unsubscribe anytime.