Why a Smart AI Will Say 'I Don't Know'

Good AI knows when to admit it is not sure.

7 min · Reviewed 2026

The big idea

A well-trained AI will say 'I'm not sure' instead of making up a guess. That's a good sign, not a bad one.

Some examples

AI saying 'I don't know' is safer than a fake answer.
Older AI used to make stuff up more often.
Newer AI is taught to flag when it is unsure.
You should trust an honest 'I don't know' more than a confident lie.

Try it!

Ask AI a very strange question, like 'What did I have for breakfast?' A good AI will say it cannot know that.

Here's why "Why a Smart AI Will Say 'I Don't Know'" matters: Learning about AI is one of the most important skills you can build for the future! Good AI knows when to admit it is not sure — and knowing how to apply this gives you a concrete advantage.

Learn what "uncertainty" means and why it's important
Learn what "honesty" means and why it's important
Learn what "limits" means and why it's important

Find out more about Why a Smart AI Will Say 'I Don't Know' by asking an AI a question about it
Talk to a grown-up about what you learned
Write down one new thing you learned today

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-explorers-foundations-AI-and-why-it-says-i-dont-know-r9a5

What should a well-trained AI do when it doesn't have the right information to answer your question?
1. Stay completely silent until you ask something else
2. Change the topic to something it does know
3. Make up a guess so you get an answer
4. Say it doesn't know or isn't sure
Why is it safer for AI to say 'I don't know' instead of making up an answer?
1. Made-up answers can be wrong or even harmful
2. Guessing helps AI learn faster
3. AI gets tired if it thinks too much
4. Questions without answers are bad questions
What is one way newer AI is different from older AI?
1. Newer AI can read your mind
2. Newer AI is taught to flag when it is unsure
3. Newer AI never answers any questions
4. Newer AI only works at night
When an AI says 'I don't know,' is that a sign of weakness or a sign of being smart?
1. It's a sign of smartness because the AI knows its limits
2. It means the AI is ignoring you
3. It means the AI is broken
4. It's a sign of weakness because the AI failed
You ask an AI 'What did I have for breakfast this morning?' What should a good AI say?
1. It should say it cannot know that information
2. It should pretend to remember from your mind
3. It should make up a funny answer for entertainment
4. It should guess something common like cereal or eggs
Why do AI developers teach AI to say when they're unsure about something?
1. To help the AI get in trouble less often
2. To make the AI slower at answering
3. To give users more accurate and helpful information
4. To make the AI seem less smart
What does the term 'uncertainty' mean in the context of AI?
1. When AI refuses to work
2. When AI doesn't have enough information to give a sure answer
3. When AI is worried about failing
4. When AI is confused about what language to use
What kind of question might make a good AI say 'I don't know'?
1. Questions about your personal thoughts or feelings
2. Questions about things that happened long ago in history
3. Simple math problems like 2+2
4. Questions about well-known facts like the capital of France
Which is more trustworthy: an AI that says 'I'm 100% sure' but is wrong, or an AI that says 'I'm not sure'?
1. They are equally trustworthy
2. The one that sounds 100% sure
3. The one that says 'I'm not sure'
4. Neither can be trusted
Why did older AI systems make stuff up more often than newer ones?
1. Older AI was not trained to recognize when it didn't know
2. Older AI was lazier and didn't want to work
3. Older AI was angry at users
4. Older AI had better imagination
If you ask a good AI about something it genuinely knows, like 'What color is the sky?' what should it do?
1. Ask you what color you think it is
2. Change the subject
3. Give you the correct answer
4. Still say it doesn't know
What should you look for to tell if an AI is being honest about what it knows?
1. What color its text is
2. Whether it admits when it doesn't have enough information
3. How fast it answers
4. How long its answers are
If an AI makes up an answer that sounds very real but is completely false, what is this called?
1. Being smart
2. Confabulation or making things up
3. Making a joke
4. Being helpful
Why is it important for AI to know the difference between what it knows and what it doesn't know?
1. So it can win trivia games
2. So it can decide when to sleep
3. So it can talk to other AI
4. So it doesn't spread wrong information to people
What is a sign that an AI has been well-trained to handle questions it cannot answer?
1. It only answers in one-word responses
2. It always gives very short answers
3. It says it doesn't know or isn't sure when appropriate
4. It never answers any questions

← Back to interactive lesson

Tendril · Explorers · AI Foundations

Why a Smart AI Will Say 'I Don't Know'

Good AI knows when to admit it is not sure.

7 min · Reviewed 2026

The big idea

A well-trained AI will say 'I'm not sure' instead of making up a guess. That's a good sign, not a bad one.

Some examples

AI saying 'I don't know' is safer than a fake answer.
Older AI used to make stuff up more often.
Newer AI is taught to flag when it is unsure.
You should trust an honest 'I don't know' more than a confident lie.

Try it!

Ask AI a very strange question, like 'What did I have for breakfast?' A good AI will say it cannot know that.

Learn what "uncertainty" means and why it's important
Learn what "honesty" means and why it's important
Learn what "limits" means and why it's important

Find out more about Why a Smart AI Will Say 'I Don't Know' by asking an AI a question about it
Talk to a grown-up about what you learned
Write down one new thing you learned today

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-explorers-foundations-AI-and-why-it-says-i-dont-know-r9a5

What should a well-trained AI do when it doesn't have the right information to answer your question?
1. Stay completely silent until you ask something else
2. Change the topic to something it does know
3. Make up a guess so you get an answer
4. Say it doesn't know or isn't sure
Why is it safer for AI to say 'I don't know' instead of making up an answer?
1. Made-up answers can be wrong or even harmful
2. Guessing helps AI learn faster
3. AI gets tired if it thinks too much
4. Questions without answers are bad questions
What is one way newer AI is different from older AI?
1. Newer AI can read your mind
2. Newer AI is taught to flag when it is unsure
3. Newer AI never answers any questions
4. Newer AI only works at night
When an AI says 'I don't know,' is that a sign of weakness or a sign of being smart?
1. It's a sign of smartness because the AI knows its limits
2. It means the AI is ignoring you
3. It means the AI is broken
4. It's a sign of weakness because the AI failed
You ask an AI 'What did I have for breakfast this morning?' What should a good AI say?
1. It should say it cannot know that information
2. It should pretend to remember from your mind
3. It should make up a funny answer for entertainment
4. It should guess something common like cereal or eggs
Why do AI developers teach AI to say when they're unsure about something?
1. To help the AI get in trouble less often
2. To make the AI slower at answering
3. To give users more accurate and helpful information
4. To make the AI seem less smart
What does the term 'uncertainty' mean in the context of AI?
1. When AI refuses to work
2. When AI doesn't have enough information to give a sure answer
3. When AI is worried about failing
4. When AI is confused about what language to use
What kind of question might make a good AI say 'I don't know'?
1. Questions about your personal thoughts or feelings
2. Questions about things that happened long ago in history
3. Simple math problems like 2+2
4. Questions about well-known facts like the capital of France
Which is more trustworthy: an AI that says 'I'm 100% sure' but is wrong, or an AI that says 'I'm not sure'?
1. They are equally trustworthy
2. The one that sounds 100% sure
3. The one that says 'I'm not sure'
4. Neither can be trusted
Why did older AI systems make stuff up more often than newer ones?
1. Older AI was not trained to recognize when it didn't know
2. Older AI was lazier and didn't want to work
3. Older AI was angry at users
4. Older AI had better imagination
If you ask a good AI about something it genuinely knows, like 'What color is the sky?' what should it do?
1. Ask you what color you think it is
2. Change the subject
3. Give you the correct answer
4. Still say it doesn't know
What should you look for to tell if an AI is being honest about what it knows?
1. What color its text is
2. Whether it admits when it doesn't have enough information
3. How fast it answers
4. How long its answers are
If an AI makes up an answer that sounds very real but is completely false, what is this called?
1. Being smart
2. Confabulation or making things up
3. Making a joke
4. Being helpful
Why is it important for AI to know the difference between what it knows and what it doesn't know?
1. So it can win trivia games
2. So it can decide when to sleep
3. So it can talk to other AI
4. So it doesn't spread wrong information to people
What is a sign that an AI has been well-trained to handle questions it cannot answer?
1. It only answers in one-word responses
2. It always gives very short answers
3. It says it doesn't know or isn't sure when appropriate
4. It never answers any questions

← Back to interactive lesson