When AI Starts Shaping What's “Normal”

Imagine asking:

It sounds like a simple question.

But where does the answer come from?

Search results. TikTok. YouTube. Podcasts. LinkedIn. Forums. Articles. Creators. Friends. Experts.

And increasingly, AI systems that summarise all of this into one apparently coherent response.

AI isn’t necessarily creating the underlying ideas.

But it is increasingly sitting between us and the information environment.

It selects.

Summarises.

Ranks.

Combines.

Rephrases.

And gives us something that can feel surprisingly authoritative.

That becomes particularly interesting when the question isn’t factual.

What’s a healthy relationship?

What does career success look like?

What’s normal behaviour?

When should I leave?

What makes a good parent?

What makes somebody a good employee?

These questions involve culture, experience, values and context.

There may not be one correct answer waiting in a database.

So when AI compresses hundreds or thousands of perspectives into a neat response, something important happens:

This is another reason Human-Centred AI needs to go beyond interface design.

We need to understand how people interpret AI responses.

How much authority do they give them?

Do they understand where the information came from?

Can they recognise uncertainty or disagreement?

What perspectives disappeared during summarisation?

And how might an apparently neutral answer influence a real human decision?

The challenge isn’t simply creating AI that gives good answers.

It is helping humans maintain judgement in an environment where increasingly persuasive systems are helping shape the questions, information and answers around them.