Seven questions to ask any AI answer
To check an AI answer, ask seven questions: where it came from, what supports it, how current it is, who benefits if you believe it, what is missing, whether a new case would change it, and who decides what happens next. An AI answer can sound sure and still be wrong, so check before you use it.
Last reviewed: 20 October 2026 6 min read
Why should you check an AI answer at all?
Because an AI answer can be confident and wrong at the same time. A system that writes text puts together words that are likely to come next. Likely is not the same as true.
A US standards body, the National Institute of Standards and Technology, names this problem "confabulation": "confidently stated but erroneous or false content". People often call it a hallucination. A smooth, sure-sounding answer is not a sign that it is right.
So the habit is simple. Read the answer, then ask questions before you use it. The seven below fit on one card.
What are the seven questions?
Each question checks one thing. You do not need all seven every time: a quick fact may need two, and a big decision needs all seven.
| # | Ask | What it checks | A good sign |
|---|---|---|---|
| 1 | Where did this come from? | The source | The answer names a source you can open, such as a museum, a government agency or a textbook. |
| 2 | What supports it? | The evidence | You find the same fact on a trusted page you opened yourself, not only inside the answer. |
| 3 | How current is it? | The date | The source shows a date, and the fact is the kind that stays true, or it was checked recently. |
| 4 | Who benefits if I believe it? | The motive | No one gains money or attention from you believing it, or the answer says so plainly. |
| 5 | What is missing? | The gaps | The answer says what it does not know, or a second source fills the gap. |
| 6 | Would a new case change it? | The test | The claim holds when you try a different example, place or person. |
| 7 | Who decides what happens next? | The decision | A person, not the answer, makes any choice that matters, and knows they are responsible for it. |
How do the seven questions work on a real answer?
Here is one invented AI answer, checked step by step.
The question: "How many hearts does an octopus have?"
The answer: "An octopus has three hearts. Scientists first discovered this in 1998."
| # | Ask | What we found |
|---|---|---|
| 1 | Where did this come from? | The answer names no source. |
| 2 | What supports it? | A US government ocean agency says octopuses have three hearts: two pump blood to the gills and one sends it through the rest of the body. The first sentence checks out. |
| 3 | How current is it? | Body parts of an animal do not change from year to year, so an older source is fine here. |
| 4 | Who benefits if I believe it? | No one is selling anything. Low risk. |
| 5 | What is missing? | The answer gives no reason for the date "1998", and no source we checked supports it. |
| 6 | Would a new case change it? | Ask about a squid next. Does the answer still hold, or does it repeat the octopus fact? |
| 7 | Who decides what happens next? | You do. Keep the part you checked, and leave out the date. |
The result: half the answer held up, and half did not. That is common. An answer is not right or wrong as a whole; each claim inside it gets checked on its own.
How do the questions change from 8 to 18?
The questions stay the same. What a child does with them grows.
Ages 8 to 10: questions 1, 2 and 7, with an adult.
- Ask "Where did this come from?" out loud, together.
- Check one fact in a book or on a page the adult picks.
- Agree that a grown-up helps with any choice the answer suggests.
Ages 11 to 13: all seven, on homework and hobbies.
- Check each claim in an answer, not the answer as a whole.
- Write down where each fact was confirmed.
- Notice questions 4 and 5: who gains, and what was left out.
Ages 14 to 18: all seven, plus testing.
- Use question 6 on purpose: change one detail in the question and compare the answers.
- Look for answers that sound balanced but leave out one side.
- Follow the school's rules on AI use, and ask the teacher when they are unclear.
How is checking an answer different from checking the system?
The seven questions check one answer. To understand the machine that wrote it, you need a different set of questions about the system itself.
These six questions work on any AI system, from a toast sorter to a large text generator:
- What information does the system receive?
- What does it do with that representation?
- How did its behavior arise or change?
- How do we test it on fresh cases?
- Where can it fail or be manipulated?
- Who chose the goal, who is affected, and who has authority to act?
Every age revisits the same six questions with more depth. The seven answer questions are a daily habit. The six system questions explain why the habit is needed.
Where do questions like these come from?
Librarians have taught source checking for a long time. One widely taught method, called SIFT, has four moves: stop, investigate the source, find better coverage, and trace claims to their original context. Questions 1, 2 and 3 on this page follow the same idea. Questions 4 to 7 add what matters for AI answers: who gains, what is missing, whether a new case changes things, and who is responsible.
Printable: seven questions to ask any AI answer
Print this card and keep it next to the screen.
Printable: seven questions to ask any AI answer
- Where did this come from?
- What supports it?
- How current is it?
- Who benefits if I believe it?
- What is missing?
- Would a new case change it?
- Who decides what happens next?
The claim I checked: ______________________________ Where I confirmed it: ______________________________
What this guide does and does not do
It does:
- Give a checking habit that works on any AI answer, on any device.
- Show one worked example, claim by claim.
It does not:
- Rate or compare any AI product.
- Prove an answer true. Checking lowers the chance of a mistake; it does not remove it.
- Replace a teacher's or school's rules on AI use.
Where does Mechna Academy fit?
Mechna Academy is a city of machines for curious children aged 8 to 18, opening first to ages 8 to 10. In each mission, a child guesses how a machine decides, then opens its Peek Hatch to check, changes one thing and tests it on a case it has never seen. Children learn how AI works this way, and they never chat with an AI.
Questions
Check each claim against a source you open yourself, such as a museum, a government agency or a textbook. Ask where the answer came from, what supports it and how current it is. If you cannot confirm a claim, leave it out.
Yes. A system that writes text puts together likely words, and likely is not the same as true. A confident tone tells you nothing about whether an answer is right.
Treat it as a guess, not a fact. Ask a parent, teacher or librarian, or look in a book. For anything about health, safety or money, a trusted adult decides.
The habit can start as soon as a child meets AI answers, with an adult. Younger children can ask "Where did this come from?" together with a parent. Teens can use all seven questions and test answers by changing one detail.
No. Checking an answer tells you whether one reply holds up. Understanding how AI works means asking what the system receives, how it learned and how it is tested, which explains why answers need checking at all.
Mechna Academy is a city of machines for curious children aged 8 to 18, opening first to ages 8 to 10.
Mechna Academy is a city of machines for curious children aged 8 to 18, opening first to ages 8 to 10.
For grown-ups: join the waitlistPlaces are limited. The waitlist is for parents and guardians 18 and older.