By the third presentation, you can predict slide four.
The problem statement will be clear. The analysis will arrive in three neat parts. The proposed solution will sound sensible.
Nothing is obviously wrong.
Then the next student begins, and the same feeling returns. The structure is familiar. The language is familiar. Even the “innovative” idea seems to have passed through the same filter.
Each project is competent on its own.
Placed side by side, it becomes harder to see where one student’s judgment ends and another student’s begins.
Generative AI is helping more people reach a respectable level of completion. That is useful. A student who struggles to begin can get a starting structure. Someone with limited design experience can build a clear presentation. A beginner can ask basic questions without feeling embarrassed.
When competent output becomes abundant, the student’s contribution has to become easier to see.
One study makes the pattern easier to see
In an online experiment published in Science Advances, participants wrote short stories. Some received story ideas from generative AI.
Access to those ideas led to higher average ratings for creativity, writing quality, and enjoyment, with particularly strong gains among participants who had scored lower on creativity before the task.
The AI-assisted stories were also more similar to one another.
This was one experiment involving a specific short-story task. It does not show that AI makes every student think alike, and it should not be stretched into a claim about every kind of schoolwork.
It does show a real trade-off. The same tool can help individuals produce stronger work while nudging a group toward familiar patterns.
UNESCO has raised a related concern. Generative systems learn from existing material, which can cause dominant viewpoints and common forms of expression to reappear in their output.
AI has not made creativity disappear.
It has made polish a weaker signal that original thinking occurred.
Before the answer, what did the student notice?
Suppose a student is asked to propose an improvement to a public space.
An AI tool can immediately suggest more lighting, additional seating, better signage, safer traffic flow, and more green space. All of those ideas may be reasonable.
But why did this student care about the space?
Perhaps they noticed that students step into the road when the sidewalk floods.
Perhaps a first-time visitor could not find the community centre entrance.
Perhaps an older resident mentioned that every bench sits in direct afternoon sun.
A worthwhile problem often begins before anyone asks for solutions. It begins when someone looks closely enough to notice that the ordinary explanation is incomplete.
Originality is not always dramatic. Often, one student simply pays attention to a detail that everyone else accepted as background.
Then they keep asking:
Who experiences this problem most?
Does the proposed solution work for the people expected to use it?
Are we answering the wrong question?
When the student does not know, what happens next?
AI can make information easier to reach. That is different from knowing how to learn.
A student investigating accessible public spaces may need some design knowledge, policy documents, observations of real routes, interviews with users, and a way to organize conflicting feedback.
No single class will provide the complete package.
The student has to discover the next gap:
I have plenty of opinions, but I do not know which source to trust.
I have an idea, but no evidence that the problem is common.
I know how to create a survey, but I do not know how to ask a neutral question.
The OECD Learning Compass connects student agency with setting goals, reflecting, making responsible choices, and learning to navigate unfamiliar situations rather than simply following fixed directions.
Tools will keep changing.
The ability to move from “I have no idea” to “I know what I need to learn first” is more durable.
Then the work has to meet the real world
A solution can look perfect on a screen.
The real world is less cooperative.
A customer may politely praise a product and never use it again. A community may appear to agree with a proposal until someone asks the people who were missing from the first conversation.
Empathy and storytelling belong inside the work itself.
Empathy helps a student notice whose experience never made it into the brief.
Storytelling helps the student choose the details that matter, explain how the evidence led to a conclusion, and make the reasoning understandable to someone who was not in the room.
They shape the problem, the evidence, and whether the proposed solution still makes sense outside the presentation.
Leave a trail of judgment
The next time a student uses AI for an assignment or project, look beyond the question, “Did AI write this?”
Instead, examine four points in the process.
Before opening the tool
What had the student already noticed? What were they unsure about? What did they currently believe?
After receiving an answer
Which suggestion did they use? Which one did they reject? Why?
After further research
What evidence changed their mind? What remains unresolved?
Before calling the work finished
Who has firsthand experience with the problem? Has that person responded to the idea?
This process will not make the assignment faster.
It will make the student’s thinking visible.
Students do not need to reject every common answer to prove that they are original. But they should be able to say:
“This is what I noticed before I asked.”
“This suggestion came from AI, and this is why I did not use it.”
“This evidence changed my conclusion.”
As good answers become easier to generate, the valuable part of a student’s work may not be a style nobody has seen before.
It may be the visible evidence that someone paid attention, made real choices, and remained responsible for the final judgment.