Does AI Have Feelings?

It can certainly seem like it does.

Tell an AI chatbot that you've had a terrible day and it might respond sympathetically. Give it some good news and it may sound pleased for you. Be rude to it and it might tell you that it understands you're frustrated.

But does the AI actually feel any of those things?

As far as we know, no.

Today's AI systems can produce language that sounds remarkably human, but there is no good evidence that they experience happiness, sadness, fear, embarrassment, pain or any of the other feelings that people do.

Why does it seem to have feelings?

AI has been trained using enormous amounts of human-created information, including examples of how people use language. Understanding how this training works is key to grasping why AI behaves the way it does—you can learn more in our article about how AI works.

That means it has learnt patterns in the way we communicate.

If you say:

"My dog died yesterday and I can't stop thinking about him."

an AI might respond with sympathy.

It isn't necessarily feeling sad about your dog. It is generating the sort of response that fits what you've told it.

The result can be surprisingly convincing.

So is it just pretending?

"Pretending" isn't quite the right word either.

Pretending suggests that the AI knows it doesn't have feelings but deliberately acts as though it does.

That's giving it too much human intention.

A better way to think about it is that AI is producing a response appropriate to the conversation.

If you're upset, sympathetic language is appropriate.

If you're celebrating, enthusiastic language is appropriate.

If you tell it a joke, it may respond as though it found it funny.

That doesn't mean there's a feeling behind the words.

But it sometimes says "I feel…"

This is one reason things can become confusing.

AI assistants use everyday conversational language, including words such as "I", "think", "understand" and sometimes expressions that sound emotional.

Don't take these phrases too literally.

When an AI says "I understand how frustrating that must be", it isn't necessarily describing an internal emotional experience in the way another person would.

It's communicating in a natural, familiar way.

Does AI understand me?

This is a more complicated question.

Modern AI can recognise remarkably subtle relationships in language. It can often work out what you mean even when you haven't explained yourself particularly well.

That can certainly feel like understanding.

But this shouldn't automatically be confused with human understanding or consciousness. The same pattern-recognition that makes AI seem understanding can also lead it astray—which is why AI sometimes gets things wrong while sounding confident.

An AI doesn't need to experience sadness itself to recognise language associated with sadness and produce an appropriate response.

Could AI have feelings in the future?

Nobody can give a certain answer to that.

Scientists, philosophers and AI researchers still debate what consciousness actually is, how it arises and whether a machine could ever possess it.

Future AI systems may also be very different from the ones we use today.

What matters when using today's AI is not to mistake convincing human-like conversation for evidence of human-like feelings.

Do I need to be polite to AI?

You don't need to worry about hurting its feelings.

Saying "please" and "thank you" isn't necessary to protect the AI from becoming upset.

But there's nothing wrong with being polite either. Many people simply find it more natural to communicate that way.

If you type:

"Thanks, that's exactly what I needed."

the AI isn't going home feeling pleased with itself.

Equally, if you type:

"No, that's completely wrong. Try again."

you haven't hurt its feelings.

In fact, telling an AI clearly when its answer isn't useful can help you get a better response.

Why does this matter?

Because AI can be extremely convincing.

After a long conversation, particularly about something personal, it can begin to feel as though there is somebody on the other side of the screen who knows you.

It's worth remembering what you're interacting with.

AI can be useful for talking through ideas, explaining things and even providing a sympathetic-sounding response when you're having a difficult day.

But it isn't a human relationship.

AI can sound caring, excited, amused or concerned. That doesn't mean it actually feels those emotions.

The words may feel human.

The feelings behind them aren't.