Use AI as a Tutor, Not a Crutch: A Practical Workflow

By Andrew Reed· October 25, 2026

MIT professor Eric So argues that leaning too hard on AI can make our brains less nimble, according to a Fox News report. His concern is that AI can weaken memory, problem-solving, and foundational mastery by removing the mental "friction" that learning needs. But So also says AI is valuable when used as a tutor or cognitive trainer, and that distinction is the useful part.

Below is a practical way to put it to work. You'll learn how to spot when you're using AI as an answer machine, and how to switch any chatbot into tutor mode for studying, writing, and brainstorming. These steps work in any chatbot app on a phone or computer; no specific version is required.

Why instant answers cost something

Per the report, So argues that AI exploits the brain's natural tendency to conserve mental effort and seek instant gratification. That creates a feedback loop: the easier it is to ask a chatbot, the more people ask it, even for the simplest tasks.

The key idea is friction. Struggling to recall a fact or work through a problem is uncomfortable, but according to So, that effort is part of how learning happens. When a chatbot removes it, you get the result without building the skill behind it.

That's So's argument, not a settled law, and the report doesn't give a number for how much damage occurs. Treat it as a good reason to be deliberate, not a reason to panic.

Step 1: Try first, then ask

The simplest habit is also the most effective. Before you open a chatbot, spend a few minutes on your own attempt.

  1. Write down what you already know about the question, even if it's rough.
  2. Make a first attempt at the answer, outline, or solution.
  3. Only then open the AI and bring your attempt with you.

The five minutes of struggle is the point, not wasted time. (The five-minute figure is my suggestion, not So's.) You're keeping the effortful part of the work for yourself.

Step 2: Tell the chatbot to act like a tutor

Most chatbots default to handing over a finished answer. You can change that with your first message. Paste in something like this:

"Act as a tutor. Don't give me the answer. Ask me one question at a time to guide me, and tell me when I'm on the right track."

This turns the tool from an answer machine into the tutor or cognitive trainer So describes. If it slips back into giving answers, remind it: "Hint only, please."

Step 3: Use AI to check your work, not do it

After you've produced something, ask the AI to critique it.

  • For studying: "Here's my explanation of this topic. What did I get wrong or leave out?"
  • For writing: "Here's my draft. Point out the two weakest paragraphs and explain why, but don't rewrite them."
  • For problem-solving: "Here's how I solved it. Does my reasoning hold up?"

The order matters. You do the thinking, the AI reviews it. Reverse that, and you're back to outsourcing.

Step 4: Quiz yourself with it

Memory is one of the skills So says overreliance can weaken, so practice recall directly. Give the chatbot your notes or a topic and ask:

"Quiz me with five questions, one at a time. Wait for my answer before telling me if I'm right."

Pulling information out of your own head is the exercise. Keep your notes closed while you answer.

Step 5: Brainstorm after you've brainstormed

For creative work, write your own list of ideas first, even a short one. Then ask the AI: "What angles am I missing?" You get the benefit of a second perspective without letting the chatbot set the direction from the start.

A quick self-check: tutor or answer machine?

Ask yourself after any AI session:

  • Could I explain this to someone else without the chat open?
  • Did I make an attempt before asking?
  • Did the AI ask me anything, or only tell me things?

If the answers are no, no, and only tell, you were probably using it as an answer machine. That's fine for low-stakes tasks. It's a problem for skills you actually want to own, like the material for an exam or the core of your job.

Where answer mode is fine

Nothing in So's argument, as reported, says you should never ask a chatbot for a direct answer. The concern is using it that way by default, including for the simplest tasks. Reserve answer mode for things you don't need to master: reformatting a list, drafting a routine email, or looking up something you'll never need to recall.

A decent rule of thumb, and this is my take rather than So's: if you'd be embarrassed not to know it in six months, do the work yourself first.

The one habit to start today

The next time you reach for a chatbot, type your own attempt first, then ask the AI to critique it instead of replace it.