Research Confirms AI Builds Student Skills One Way, Undermines Them Another

Last updated: June 29, 2026

Here is what every school’s AI policy misses, and what a Middlebury College economics experiment proved in spring 2025. Students who used AI to push their own thinking further — brainstorming, asking for feedback, questioning what they did not understand — learned more by week two than students who never touched AI at all. Students who used AI to automate their work — having it generate essays, produce the output, do the thinking — scored worse in week two than those no-AI peers. Same tool. Same access. The only thing that differed was what role the student’s brain played. That is not a nuance. That is the whole story.

TL;DR

Middlebury College economists Germán Reyes and Zara Contractor surveyed 634 students (December 2024–February 2025) and found 80 percent use AI for coursework; 61 percent use it for augmentation and 42 percent for automation.
In a spring 2025 experiment, students who used AI to automate their work scored worse than students who used no AI at all by the second week.
Students who used AI for augmentation — pushing, explaining, and building on their own thinking — scored better than no-AI students by week two.
The research is available as a working paper through IZA (Discussion Paper No. 18055).
For children who already struggle with reading, math, or attention, AI automation carries extra risk: it produces the output while bypassing the active processing their brains need to build the underlying skill.

Schools are racing to set AI policies, and most of them are asking the wrong question. Here are the questions parents actually need answered.

Common questions

What is the difference between AI augmentation and AI automation?

Augmentation means using AI to push your own thinking further: asking it to explain something you did not understand, give feedback on what you wrote, or brainstorm options you then evaluate yourself. Automation means letting AI produce the output: generating the essay, writing the code, completing the task for you. The Middlebury experiment found they produce opposite outcomes for long-term learning, even though students used the same tool.

Does this mean kids should not use AI for schoolwork?

No. The augmentation users in the Middlebury study outperformed students who used no AI at all by the second week of the experiment. The finding is not that AI hurts learning but that automation does. A useful test: after using AI, can your child explain the material in their own words without looking at what AI produced? If yes, it is likely augmentation. If not, it is likely automation.

My child has dyslexia or struggles with reading. Should I be more concerned about AI automation?

Yes, and here is why: children who already find the cognitive work harder are more likely to reach for shortcuts. AI automation is an unusually efficient shortcut because it produces acceptable output with almost no effort from the learner. But that effort is often exactly what the brain needs to build the underlying reading or reasoning skill. A screener is a helpful starting point for identifying which processing systems need the most support and is not a diagnosis. For formal accommodations such as an IEP or 504 plan, or if you suspect a vision, hearing, or medical cause, pursue a professional evaluation, since that is the only route to those supports.

How do I tell whether my child is using AI in a way that builds or replaces their learning?

Three questions: Did they understand what AI produced, and can they explain it to you without it? Did they use AI on a draft they had already written, or to produce a first draft they never wrote? Could they answer a follow-up question on the topic without AI? If AI produced the work and they cannot explain it, that is automation. If they used AI to go further on thinking they had already started, that is augmentation, and the research says it builds skill.

A Middlebury experiment found AI automation made students learn less than no-AI peers. AI augmentation made them learn more. Same tool. The question was always who was doing the thinking. Tweet Now Follow Learning Success

What the Middlebury research found

The study comes in two parts, both led by Middlebury College assistant economics professor Germán Reyes and colleague Zara Contractor. The first was a survey conducted between December 2024 and February 2025. They reached 634 students — more than 20 percent of Middlebury College’s undergraduate body. What they found: 80 percent of students already use AI for coursework. Among those, 61 percent reported augmentation use — asking AI to explain a concept, generate feedback, brainstorm possibilities they then evaluated themselves. Forty-two percent reported automation use: having AI produce essays, generate code, or complete tasks on their behalf.

The second part was an experiment. In spring 2025, students were randomly split into two groups: one with standard online tools, one with those tools plus AI. Both groups researched CRISPR gene-editing technology and wrote essays — first in week one, then again in week two. The results split cleanly along usage type. Automation users got a short-term grade lift in week one. By week two, they performed worse on average than students who used no AI at all. Augmentation users showed only a modest improvement in week one — and a significantly larger one by week two. “The effects of AI largely depend on how students use the tools,” Reyes said. The research is available as a working paper through IZA (Discussion Paper No. 18055).

Author Quote "

The effects of AI largely depend on how students use the tools.

What the coverage gets wrong

Most reporting on this study leads with the reassuring finding: the majority of students use AI for augmentation rather than cheating. That is accurate but incomplete. The sharper finding — that automation produced short-term grade boosts and long-term learning loss while augmentation produced the opposite — is the result that changes what parents and teachers should actually look for. The popular framing of AI in education as a binary (ban it or embrace it) misses what researchers Germán Reyes and Zara Contractor actually measured: that the same tool, used two different ways, produces opposite outcomes for learning. Coverage that buries that finding in favor of the “most students are responsible” headline leaves parents without the one lens that would actually help them.

The frame the science actually supports

Coverage of this study tends to highlight the comfortable finding: most students use AI responsibly. That framing lets everyone off the hook and misses the harder lesson, which is the one parents of struggling learners actually need.

AI automation produced short-term grade boosts and long-term learning loss. That pattern has a precise explanation in decades of cognitive science. The brain builds durable skill through active engagement — through struggle that requires the learner to retrieve, connect, and construct. Brain-imaging studies led by Sally Shaywitz at Yale and by researchers at Stanford show that the neural pathways reading and reasoning rely on are physically built through practice. When a student has AI generate the essay, those pathways get no exercise. When they use AI to challenge their own draft, explain what confused them, or think through why an argument fails — they are doing exactly the kind of effortful, engaged processing that builds lasting understanding. The science on how the brain builds new pathways has said this for years. The Middlebury experiment now shows it in a classroom with an outcome measure.

The system getting this wrong is not the students — the majority of them are already reaching for augmentation. It is the binary framing that dominates the adult conversation: either ban AI (it is cheating) or celebrate any AI use (it is personalized learning). A school policy that bans all AI forecloses the augmentation that helps. An ed-tech platform that celebrates any AI use hides the automation that costs learners. Neither camp is asking what the student’s brain is doing. For children who already find the cognitive work harder — those struggling with reading, math processing, or attention — the automation trap runs deeper. These are the children already looking for ways to make the work easier. AI automation is the most efficient shortcut yet: it produces the output that satisfies the assignment while the brain misses the practice it needs to build the underlying skill. The IDA 2025 definition’s acknowledgment that reading and learning draw on multiple processing systems — language, attention, working memory, processing speed — means there are multiple systems that develop through active engagement and get bypassed through automation.

Key Takeaways:

The study: A spring 2025 experiment by Middlebury College economists Germán Reyes and Zara Contractor randomly assigned students to AI access or none, had them write essays twice, and measured outcomes by usage type.

The finding: Students using AI for automation scored worse than no-AI peers by week two; students using AI for augmentation scored better — same tool, opposite outcomes depending on who did the thinking.

For parents: The question is not whether a child uses AI but whether the AI is doing the thinking for them — because the long-term learning outcome depends entirely on the answer.

What this means for your child

None of this makes AI a problem to avoid. The augmentation users in the Middlebury study outperformed the no-AI group by week two. A student who uses AI to test their understanding, evaluate options they have generated, or improve a draft they already wrote is using a cognitive scaffold that works. The research says so.

The honest tradeoff: AI automation is efficient in the short run and corrosive in the long one. AI augmentation asks more of the student in the short run and builds more skill by week two. A few questions worth asking when your child uses AI for schoolwork: Did they read what the AI produced and explain it back to you in their own words? Did they use AI to generate a first draft, or to improve a draft they had already written? Could they answer a follow-up question on the topic without AI? If the answer to all three trends toward no, the AI did the thinking. That is the usage pattern the Middlebury experiment showed costs learners. If the answers trend toward yes, that is augmentation — the pattern that beat the no-AI group. The question to ask of any tool is the same one that governs all of it: is this building my child’s skill, or doing the work in its place?

Your child’s brain builds skill the same way it always has: through active effort, through doing the hard thinking themselves, through engaging a problem rather than routing around it. The villain in this story is not artificial intelligence — it is the framing that treats all AI use as equivalent, the comfortable belief that any use of the tool is progress, while the question of who is doing the thinking goes unasked. AI automation gives a child the answer without the growth. AI augmentation gives them a scaffold for work they still do themselves. If you want to understand exactly which processing systems your child needs to strengthen — and build a plan around that, not around a label — the Learning Success All Access program gives you that picture today.