NEW YORK — Schools are entering a complicated new phase with artificial intelligence. Students are gaining more access to AI tools – while teachers are still working out how to police their use.

The tension has sharpened during the 2026 school year.

Google is expanding Gemini inside Classroom and school-issued accounts. Meanwhile, New York City Public Schools have taken a very different path for younger students.

The nation’s largest school system has banned student-facing generative AI across grades 2K through 8 for the 2026–27 school year. High school students can still use approved programs under tighter rules.

At the same time, Google is bringing Gemini tools to students of all ages through Classroom. The company says students can use the system for quizzes and study guides built around class materials. These two developments show how divided education has become.

Schools want students prepared for an AI-heavy future. They also worry that the same technology can do too much schoolwork for them. This leaves teachers with another difficult question: how can they tell when AI crossed the line?

AI Use Is Already Common Among Students

New research suggests schools are trying to catch up with behavior already happening outside official policies. A nationally representative Common Sense Media survey published in August questioned 1,017 American teenagers. It found that 70% of teens already use AI for schoolwork.

Among those students, 63% said they had used AI to get answers for school assignments. Some copied answers directly – while others changed or improved the generated response afterward. The numbers become more interesting when schools try blocking access.

Among teens using AI for schoolwork – 44% said their school had blocked an AI service. Of those students – 59% simply switched to a personal device instead. This makes a blanket ban difficult to police.

It also explains why schools have turned toward detection software.

Detection Tools Are Facing Their Own Limits

An AI detector can examine submitted writing and estimate how much text may come from a generative system. For teachers handling dozens of essays, that sounds useful. The problem starts when an estimate gets treated like proof.

Turnitin acknowledges that false positives can happen within its own system. Its current reports no longer provide exact scores when detection falls between 1% and 19%. The company says results in that range are less reliable.

Turnitin also gives schools a direct warning about disciplinary use. Its guidance says detection “should not be used as the sole basis for adverse actions” against students. Instead, the company recommends further review and human judgment.

This distinction has become more important as students use AI in different ways.

One student may ask Gemini to explain algebra. Another may ask it to write an entire history essay. A detector cannot always explain which process happened.

Schools Are Taking Different Approaches

New York City is trying to draw a clearer line by age. Its new policy blocks student-facing generative AI through eighth grade. High school students will get limited access alongside required AI literacy lessons.

Schools will require high school students to complete two 45 minute AI literacy modules. These lessons cover data privacy and appropriate classroom use.

Chancellor Melissa Aviles-Ramos framed the policy around keeping human learning central. “AI will never replace our teachers,” the school system says in its guidance.

The challenge becomes even more complicated in career focused classes – where students may be expected to use digital tools as part of practical learning. In business or entrepreneurship courses – for example – students might work with everything from AI writing assistants to an invoice generator when completing realworld assignments. That makes it difficult for schools to create one rule that treats every use of technology in the same way.

Google is heading in a more open direction. The company announced in June that Gemini in Classroom would reach students of all ages. Study notebooks can build personalized lessons around a student’s learning goal.

Google also announced expanded AI training for educators during September. A $4 million commitment through Google.org will support free training delivered with Digital Promise. So the argument is no longer simply about banning AI. It is increasingly about deciding which uses support learning.

Students Are Asking Similar Questions

The Common Sense Media findings show students understand some of the trade-offs themselves. Among teens using AI – 38% said the technology causes them to generate fewer original ideas. Another 39% said they miss some learning when AI completes school assignments. Still 66% said AI also helps them understand schoolwork.

These numbers show why detection alone cannot settle this debate. A student asking for an explanation may be learning. A student submitting a generated answer without understanding it presents another problem entirely.

Yvette Renteria of Common Sense Media argues that schools need broader conversations around AI policies. “Policies, rules, and bans” cannot work without community discussion, she said. Her organization found that only 27% of teens had discussed how AI works with a teacher.

Detection May Become Only One Part of the Process

Schools are unlikely to stop using detection software soon. Teachers still need some way to flag assignments that deserve closer attention. Detection tools can provide one piece of that process. However, current guidance points toward a wider approach.

Draft history can show how an assignment developed. Classroom discussions can test whether students understand what they submitted.

Teachers can also ask students to explain sources or defend an argument in person. These methods take more time than checking one percentage. They also provide information an AI detector cannot provide by itself. As classroom AI expands, schools may have to accept a less convenient answer.

The question is no longer simply whether students used AI. The harder question is how they used it, and whether they still did the learning themselves.