AI doesn't teach you: it just gives you the answer
17% worse when the tool is removed: the academic shortcut sends the bill later, according to a Pennsylvania study.
You ask for an essay and it comes out in seconds. In an analysis on Xataka México, PISA 2025 is cross-referenced with MIT research and an uncomfortable conclusion emerges: AI doesn't teach you, it just gives you the answer. It may sound exaggerated until you see the numbers.
Twenty points you pay for later
According to the OECD's PISA 2025 assessment, students who rely on AI to summarize texts or write essays get significantly worse results in science. The difference is around 20 points: almost a lost school year.
A University of Pennsylvania study followed nearly a thousand math students. With GPT-4, grades went up at first, but when the platform was removed, the group that had studied with AI performed 17% worse than the control group.
«An answer that cost you nothing didn't form anything in you.»
That's what worries me. The tool doesn't make anyone stupid; it lets you turn in the result without doing the work that shapes judgment. An answer that cost you nothing didn't form anything in you.
Proverbs says it without decoration: "In all toil there is profit, but mere talk tends only to poverty" (Proverbs 14:23, ESV). Fruit is born from the process, not from the file you submit.
The ban for those under 21 doesn't fix the root
The Wall Street Journal published a report where banning these tools for those under 21 is proposed. The idea sounds blunt and hard to apply: no school filter stops a young person with a cell phone.
PISA 2025 brings a more useful nuance than any prohibition. Those who use AI frequently and also receive formal instruction in AI literacy end up slightly above average.
There's the clue I'm following. The problem isn't the chatbot in the backpack, but the effort evaded with it. Banning can reassure adults; it doesn't educate any student. Teaching how to use it with judgment moves the needle.
Questions, not answers
A clinical trial from MIT found that daily use of these tools is associated with more isolation, more dependence, and more feelings of loneliness. The American Psychological Association warned that many young people find it hard to distinguish simulated empathy from a program from a real human conversation.
On that I want to be clear: a chatbot doesn't know you, doesn't correct you with affection, and doesn't rejoice for you. It can sound kind, but it doesn't love you.
Researchers call this "metacognitive laziness": solving in seconds and avoiding any real mental effort. Their proposal is an AI tutor that asks guiding questions instead of delivering the solution instantly. That change of direction seems to me the most sensible of the whole debate.
I wonder if we are teaching how to use a tool or surrendering our head to it. A concrete step: before asking for the answer, write your own attempt; then compare, ask, and correct. Ask for questions, not solved tasks.
"Do not be conformed to this world, but be transformed by the renewal of your mind, that by testing you may discern what is the will of God, what is good and acceptable and perfect" (Romans 12:2, ESV).
Studying with honesty is a form of worship, and the judgment you gain serves you for a lifetime. Who do you hand your head to when you study?
