Schools Bet on AI Tutors. Stanford Finds Kids Barely Use Them

Last updated: June 25, 2026

If you have ever bought an app that promised to fix your child’s reading and then watched it gather digital dust, the newest research will feel familiar. Two school districts handed hundreds of elementary students a well-known AI reading tutor and set aside class and after-school time for them to use it. Then researchers at Stanford counted the minutes. The children touched the tutor for two to five minutes a week, and the ones who needed the practice most were the least likely to log on at all. The lesson is not that technology failed your child. It is that access to a tool was never the same thing as learning.

TL;DR

The Stanford SCALE Initiative study, published June 17, 2026, followed about 350 elementary students across two randomized controlled trials in two districts.
Despite scheduled time, average weekly AI reading-tutor use was 2.18 minutes in one district and 5.23 minutes in the other, and only 53 to 60 percent of students ever logged on.
The platform recommends about 30 minutes of weekly use for measurable reading gains.
Pairing students with a human tutor raised the number of stories completed by 71 to 80 percent, but neither district saw a statistically significant improvement in reading scores.
Students who used the tutor on their own tended to be higher-achieving and less likely to receive special education services.

Schools are handing kids AI tutors faster than the research keeps up, and parents are left wondering whether the tools help. Here are the questions that come up most.

Common questions

Does this mean AI tutors do not work for kids?

No, and the researchers were careful not to claim that. The study did not test whether the AI itself teaches reading. It found that access alone did not lead to use, that most children barely logged on, and that the human who provided accountability is what moved engagement. Read it as a caution about implementation, not a verdict on the technology.

My child is a struggling reader. Will an AI tutor help them catch up?

Be cautious about assuming it will on its own. In this study, the students who needed extra reading practice most were the least likely to use the tutor when left alone with it. If you try one, sit alongside your child and build the routine together. A screener is a useful starting point for understanding where your child stands, 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 too, since that is the only route to those supports.

How much should my child use a learning app for it to matter?

The platform in the study recommends about 30 minutes a week for measurable gains, far more than the two to five minutes most children logged. Consistency matters more than novelty. A short daily routine with a person checking in beats a tool left to run on its own.

If the app is not the answer, what is?

The relationship. Adding a human who checked in, encouraged, and troubleshooted raised the number of stories children finished by 71 to 80 percent. The companionship and accountability a person provides is hard for a screen to replace, which is why your involvement is the highest-leverage thing in the room.

Two districts gave kids a well-known AI reading tutor and scheduled time to use it. Stanford counted the minutes: two to five a week. The fix was never the app. It was the human beside them. Tweet Now Follow Learning Success

What the Stanford study found

The study, from Stanford’s SCALE Initiative and published on June 17, 2026, followed about 350 elementary students across two randomized controlled trials in two districts. Each district carved out dedicated time and expected students to log on for at least two thirty-minute sessions a week. That was the plan on paper.

What happened was thinner. Of the students assigned to work on their own with the AI, only about 60 percent in one district and 53 percent in the other ever logged on at all. Average weekly use across all students came to 2.18 minutes in one district and 5.23 minutes in the other. Even the children who did use it averaged 13 to 26 minutes a week for only four or five weeks, well under the 30 minutes a week the platform itself says is needed for measurable reading gains.

Then the researchers tried the obvious fix and added a human. Pairing each child with a tutor who checked in, encouraged them, and helped with snags lifted the number of stories students finished each week by 71 percent in one district and 80 percent in the other. Yet neither district saw a statistically significant bump in end-of-year reading scores. “Having these tools available, even if they’re really good, doesn’t necessarily mean they’re going to get used if they’re not being embedded into kids’ learning experiences,” said lead author Carly Robinson, research director for the Stanford SCALE Initiative.

Author Quote "

Having these tools available, even if they’re really good, doesn’t necessarily mean they’re going to get used if they’re not being embedded into kids’ learning experiences.

What the coverage gets wrong

Most coverage of AI in schools runs on a single loop: boosters promising a tutor for every child against skeptics warning of screens. This study fits neither lane, and the nuance is the point. The researchers did not find that AI tutoring fails, and it would be wrong to say they did. What they found is narrower and more useful: access is not use, use was minimal even with scheduled time, and the human who provided accountability is what moved engagement. The popular line that handing every student an AI tutor will close achievement gaps has the causation backwards, because the children who most needed the practice used the tool the least.

The frame the science supports

The story sells itself one way: AI tutors deliver personalized instruction at a scale no human teacher matches, so the path to closing the gap is to put one in front of every child. This study quietly takes that frame apart. The expensive part was never building the tool. The hard part was getting a child to sit with it, stick with it when the reading got hard, and come back the next day.

That points straight at the thing the technology was supposed to replace. Personalized instruction has always run on a relationship, and the research bears it out here: adding a person who provided accountability and motivation moved engagement, while access on its own moved almost nothing. The active ingredient was the human, not the algorithm. “There is just this component of accountability that a human can provide, where it’s so easy to look away or check out of something when it gets hard when you’re dealing with a screen,” Robinson said. A struggling reader does not need a slicker interface. They need someone who notices when they stall and stays in the room, which is the part the science of how children actually build reading skills has said for years.

The sharpest finding cuts at the kindest argument for these tools. Among children who used the tutor on their own, the ones who logged on tended to be higher-achieving and less likely to receive special education services. The students with the most to gain from extra practice were the least likely to get it. “I think it should give us pause about treating AI tutoring as an equity solution,” Robinson said. An equity fix that the neediest learners use least is not closing a gap. It risks widening one.

Key Takeaways:

The news: A Stanford SCALE study of about 350 elementary students found average AI reading-tutor use of two to five minutes a week despite dedicated time to use it.

The angle: Adding a human tutor raised stories completed by 71 to 80 percent, while access on its own moved neither usage much nor reading scores. The relationship was the active ingredient.

For parents: Treat an AI tutor as a supplement to practice with a person, not a substitute, and ask whether any tool builds your child’s skill or does the work for them.

What this means for your child

Here is the honest tradeoff, because none of this makes an AI tutor worthless. Used with a person beside them, children finished far more stories, and for an older kid who has the foundation, a patient explainer or a reading companion has real value. The danger is the quiet assumption underneath the pilots: that buying access is the same as delivering help. It is not, and the most vulnerable readers are the ones that gap hurts most.

So treat any learning app as a supplement to a relationship, not a substitute for one. If you bring a tutor home, sit with your child for the first stretch, build the short daily routine together, and be the accountability the screen lacks. Consistency beats novelty, and a few focused minutes with you checking in does more than an hour of a tool left to run alone. When you weigh any tool, ask the one question that governs all of it: is this building my child’s skill, or doing the work in its place? A learning skills analysis is a calm place to see where your child actually stands before you spend a dollar on software.

Author Quote "

At this stage, there has been little evidence that GenAI chatbot tutors meaningfully impact learning outcomes for students, or that they are developmentally appropriate for students in elementary schools.

Your child’s growth was never waiting on a better algorithm. It was waiting on someone who notices when the work gets hard and stays in the room anyway, and that someone is you. The villain in this story is not artificial intelligence. It is the comfortable belief that more technology always means more learning, the belief that hands a child access to a tool and calls it help while the kids who need the most are served the least. You are the one who sees whether a tool is building your child up or doing their growing for them. If you want every reading, math, and confidence-building tool in one place, with a personalized Action Plan and a real human approach you keep either way, the All Access program gives you a free trial to start today.