Education news

Stanford Gave Kids an AI Tutor. Many Never Used It.

Stanford handed 355 elementary students their own AI reading tutor. Many never opened it, and the ones who did used it far too little to help. The fix was not a better app. It was a person in the room, and even that has limits worth understanding before you hand your child a screen.

Laura Lurns Last updated August 27, 2026 · 10 min read

Stanford Gave Kids an AI Tutor. Many Never Used It.

If you have ever handed your child a shiny learning app and assumed the learning would follow, a new Stanford study is worth your attention. Researchers gave 355 elementary students their own AI reading tutor and tracked what actually happened next. Across two school districts, between 40 and 47 percent of the children never logged on even once, and the ones who did used it far less than the tool’s own makers said was needed to help their reading. Adding a real person to encourage them lifted how much they used it, and still did not move reading scores in a way the study was able to measure. Access, it turns out, is not the same thing as learning.

TL;DR

A Stanford working paper (EdWorkingPaper 26-1451, June 2026; Robinson, Gormley, Ribeiro & Loeb) ran two randomized trials with 355 students in grades one through five across two high-poverty school districts, each given access to an online AI reading tutor.
Left to use it on their own, about 40 percent of children in one district and 47 percent in the other never logged on once; among those who did, average weekly use was 13.2 and 25.8 minutes, both below the 30 minutes a week the tool’s makers said was needed for a reading gain.
Adding a human focused on engagement rather than instruction raised the number of stories children completed by 71 to 80 percent.
Neither district showed a statistically significant reading gain from the human support, but usage stayed too far below the recommended dose for the study to test whether the tool itself raises scores, so this is not evidence AI tutoring “does not work.”
The paper is not yet peer-reviewed, and the comparison group still had the AI tutor, so there was no group without any AI tool to serve as a true baseline.

Context the coverage leaves out

What the coverage gets wrong

The reporting on this study was mostly careful, and the dominant frame, that AI tutoring hit a wall because students would not use it, is accurate as far as it goes. The risk is in what a hurried reader does with it. The first misread is to treat the flat reading scores as proof that AI tutoring does not work. The study does not support that, because usage never came close to the dose the tool was designed around, so the trial simply never tested whether it raises scores. The second, quieter misread is to file this under “engagement problem” and assume a slicker app or a better nudge would fix it. That skips the harder lesson the study actually teaches: access is not use, use is not learning, and a rising engagement number is not evidence a child learned anything. The story is not that one AI tutor flopped. It is that the whole industry keeps selling the handoff as if it were the outcome.

Stanford researchers gave hundreds of kids their own AI reading tutor and watched most of them barely touch it. Here are the questions parents are asking about what that means for the apps in their own homes.

Common questions

Does giving my child an AI tutor help them read better?

On its own, giving access is not enough, which is the whole point of this study. When 355 children were handed their own AI reading tutor, up to 47 percent never logged on, and the ones who did used it far below the amount the makers said was needed. The tool only has a chance to help a child who actually uses it, and in this study the thing that got kids to use it was a person encouraging them, not a better app. Access is the start of the question, not the answer.

Why wouldn’t kids use a free tutor that was supposed to help them?

Because a screen and a good intention are not much of a reason for a six-year-old to show up. Young children rarely self-start hard practice alone, and an AI tutor asks them to do exactly that. In the study, adding an adult whose only job was to encourage participation and build a routine raised how much the tool got used by 71 to 80 percent. The lesson is not that children are lazy; it is that engagement is something a caring adult creates, not something the software supplies by existing.

Should I avoid AI learning tools altogether?

No. The study does not show AI tutoring fails, only that access without use teaches nothing, and use in this trial never reached the level needed to test whether the tool raises scores. A tool built around effort, one that makes your child recall and work through mistakes rather than tap through for points, has real value. Use it as a supplement to your involvement, sit with your child for it, and judge it by whether they are working, not by how many minutes a dashboard logged.

How do I figure out where my child actually needs help?

Start by getting specific, since reading, math, writing and attention pull on different underlying skills and a general study app will not sort that out for you. A free parent screener is a low-stakes way to see where to focus first. A screener is a starting point, not a diagnosis. If your child might need formal accommodations such as an IEP or 504 plan, or you suspect a vision, hearing, or medical cause, pursue a professional evaluation too, since that is the only route to those supports. If the difficulty spans more than one area, a multi-area analysis keeps you from framing a whole-child pattern as a single-subject problem.

Share the idea

Stanford gave 355 kids their own AI reading tutor. Up to 47% never logged on. A caring adult got them using it, but reading scores didn't follow. Access to a tutor isn't the same as learning.

Post on X Follow

What the Stanford researchers found

The study, a not-yet-peer-reviewed working paper released in June 2026 by Carly Robinson and colleagues at Stanford’s SCALE Initiative, ran two randomized controlled trials, one in each of two school districts serving high-poverty communities. All 355 children, in grades one through five, were given access to an online AI literacy tutor. Half were left to use it on their own. The other half got the identical tutor plus a human whose only job was engagement, not teaching: encouraging kids to show up, sorting out tech problems, building a little accountability and routine. Children were meant to do two thirty-minute sessions a week, and the platform’s makers said a child needed at least thirty minutes a week on it to see any reading gain.

Almost nobody got there. In one district, a little over 60 percent of the students ever logged on at all, meaning about 40 percent never touched it; in the other, only 53 percent ever logged on. Among the children who did use it, average weekly use came to 13.2 minutes in one district and 25.8 in the other, both under the thirty-minute floor the makers had named. Robinson put the headline plainly: “having access to this AI tutor isn’t the same as using it,” and called it “a key finding that we weren’t even meaning to test.”

Adding the human helper did move the needle on use. The number of stories children completed rose by 71 percent in one district and 80 percent in the other. What it did not do was raise reading achievement by an amount the study could detect in either district. That is not the same as proving the tutor does not work, and the researchers are careful about the difference: use stayed so far below the recommended dose that the trial never got to test whether the tool lifts scores. As Robinson said of the achievement result, “We never really got close enough to the dosage needed to find out.” Reporters covering the study set it in a wider context, noting that even Sal Khan, whose Khanmigo tutor is among the best known, has acknowledged that getting students to actually use an AI tutor is a real and unsolved problem.

Having access to this AI tutor isn’t the same as using it.
Laura Lurns · Learning Success expert

Engagement is not the same as learning

The popular story about artificial intelligence in education treats the handoff as the whole event: give a child the tutor, and progress is presumed to be underway. This study takes that story apart in the cleanest way possible. Access did not produce use. Boosting use did not produce a measurable gain. Between the app a family signs up for and the reading a child does, there are two gaps, not zero, and most of the attention in the AI-tutoring conversation is spent pretending the first one is the finish line. As The 74 summarized the paper, access to AI tutoring does not ensure meaningful use, and while human support increases engagement, it might not be enough on its own to generate measurable learning gains.

Here is the trap the whole field keeps stepping into, and it is our reading of what this study exposes rather than a settled verdict. Time-in-app and lessons-completed are easy to count and easy to mistake for learning, so a product that raises them looks like it is working even when the test scores stay flat. That is exactly what happened here: the human helper drove engagement up sharply, and the thing engagement is supposed to stand in for, actual reading skill, did not follow in a way anyone could measure. A number going up on a dashboard is not the same as a child who reads better, and confusing the two is the most common move in ed-tech marketing.

Notice what did the work, such as it was. The tool got used more when a caring adult was standing next to it. That is the part worth holding onto. The strongest role for a piece of software is to free up an engaged adult to do the relational, motivating work no app has ever replaced, not to serve as a stand-in for that adult while everyone hopes for the best. This is why supporting your child without taking the work over matters more than the particular tool on the screen. A tutor amplifies a relationship. It does not manufacture one.

Key takeaways

Access is not use: handed their own AI tutor, 40 to 47 percent of the children never logged on even once.
Engagement is not learning: a human helper raised how much kids used the tool, but reading scores did not follow.
A tool supplements a caring adult: the tutor got used most when a real person was there in the room.

What this means when you hand your child a screen

None of this makes AI tutors bad, and it is not a reason to unplug the family. A tool built around effort, one that makes a child recall, work through their mistakes and try again, has genuine value. The honest tradeoff is narrower and more useful than a verdict: no tool teaches a child who does not open it, and in this study the single biggest lever on whether the tool got opened at all was whether a person was there to make it happen. The technology was never the intervention. The child using it, with support, was.

So the practical move is to stop treating access as progress. Signing your child up for a well-reviewed app is the beginning of the question, not the answer to it. Sit with them for the first sessions. Be the reason they show up. Watch closely whether the tool is making them do the hard part, retrieving and reasoning, or whether it is entertaining them while a progress bar fills. If your child is one of the many who quietly never logs on, that is not a failure of willpower on their part; it is the predictable result of leaving a young child alone with a screen and a good intention.

The forward-looking version is simpler than the marketing. A learning tool works best as a supplement to your involvement, not a substitute for it. Decide what you want the app to do, put yourself in the room while your child does it, and judge the tool by whether your child is working, not by how many minutes the dashboard logged. That is the difference between a screen that helps and a screen that only looks like help.

Your child is capable of real learning, and the most powerful ingredient in it is not an app, it is you. You do not need to fear the technology or become an expert to make a good call here; you need to know that access is not the same as learning, and that no tool works for a child who is left alone with it. The villain is not artificial intelligence and it is not any company that builds a tutor. It is a comfortable and expensive idea: that handing a child a clever screen counts as teaching them. Cutting through that confusion is what Learning Success is built to do. Our All-Access membership opens an assessment that asks about the processing systems your child’s learning runs on, and a roadmap that names what to build first and in what order, worked alongside the reading, spelling or math instruction your child is already getting.

See what All Access gives your child

Is your child struggling in school?

Get your free personalized learning roadmap

You describe what you see at home. We turn it into a plan you start this week.

Answer 5 short questionnaires about what you already notice, 30–45 minutes at your own kitchen table
Your child sits no test and gets no score: nothing to schedule, nothing for them to dread
You do the answering, the AI does the writing, and a person reviews it before it reaches you
Access all 40+ courses instantly: reading, math, focus, processing and more, with new ones added regularly

Why we use AI, plainly: it writes from a knowledge base our team maintains and audits. We work through it line by line and pull anything the evidence stops supporting. The roadmap you get on Tuesday reflects what we corrected on Monday, and a human still reads it before you do.

Your school district must evaluate your child free of charge if you ask in writing, whatever your income and whatever the outcome (US, 34 CFR 300.111 and 300.301(b)). That route takes time and answers a different question than you do. This one starts today, from what you already know.

Get my free assessment & 7-day trial Cancel anytime during your trial.

Your answers stay yours. We do not sell your personal information, and we do not hand identifiable assessment data to outside AI companies to train their models.

A screener is a starting point, not a diagnosis. If your child might need formal accommodations (an IEP or 504 plan), or you suspect a vision, hearing or medical cause, pursue a professional evaluation too. That is the only route to those supports.

References

Stanford NSSA / SCALE — Access is Not Enough: Human Support Improves Engagement with AI Tutoring
EdWorkingPapers (Annenberg Institute, Brown) — Paper 26-1451 abstract: Access is Not Enough
Chalkbeat — AI tutoring research ran into a problem: students wouldn’t use it
The 74 — Even With Human Help, Kids Need Motivation to Use AI Tutors

Laura Lurns · Learning Success expert Writes about the learning brain for parents who want plain answers. Every article is grounded in current neuroscience and classroom practice.