Google's AI Push into Schools: 'It Slowed My Thinking Down'

Faced with quadratic equations, Richard Scott couldn’t resist the temptation.

The Pennsylvania eighth-grader knew that the Google search box on his school laptop was essentially an AI chatbot that would helpfully spit out answers. He started popping algebra questions into the search bar to quickly finish off his daily schoolwork.

By the end of the school year, Scott’s grades were tanking. He had started leaning on Google’s AI-assisted search for other classes, like when he needed to label the parts of a volcano or write an essay about Edgar Allan Poe’s “The Raven.” His grades in math—a subject he’d done well at previously—suffered the worst. Even though he got mostly B’s in the quarterly marking periods, according to his transcript, he failed the final.

“It slowed my thinking down,” the now-15-year-old said. “When I was in a test and I couldn’t use AI, it was hard to think.”

No tech giant has been more aggressive in putting AI in front of schoolchildren than Google. The search engine juggernaut’s ubiquitous presence in American classrooms through Chromebooks and its software partnerships with schools have provided a critical launchpad for its AI tools to reach kids. But the rollout has helped push many students to become academically and emotionally dependent on AI, according to interviews with more than 60 teachers, school officials, parents, children and clinicians.

Google’s own researchers have written that the technology can pose elevated risks to children, who struggle to distinguish chatbots from human interaction, and has potential for cognitive and emotional harm, according to interviews with current and former Google employees and a review of more than a dozen Google research papers.

“We’re doing this experiment on an entire generation on cognitive development, with no control group,” said Mikel Rodriguez, a former Google research scientist who left last year. He said people inside Google privately worried among themselves about how their AI use was affecting their own cognitive skills. “It’s really uncertain what that’s going to lead to in terms of learning.”

A Google spokeswoman said that studies show that AI tools paired with teacher-led instruction can help students learn more effectively. “Teachers around the world use our technology every day because it solves real classroom challenges and helps improve learning outcomes,” she said. Google focuses on encouraging critical thinking and has specific protections for users under 18, she added, including to prevent emotional dependence.

While schoolchildren are encountering Google’s AI tools on school-issued Chromebooks, teachers and parents note students are accessing AI through other chatbots, too, such as ChatGPT. Other forms of screentime are wearing away at attention as well. Overall, AI, along with social-media and smartphone use, has contributed to a drop in reading skills worldwide, according to the Organization for Economic Cooperation and Development. Student scores fell to their lowest average in a quarter-century on its 2025 Program for International Student Assessment, the world’s largest comparative education survey, which assessed 760,000 15-year-olds across 91 countries.

U.S. PISA scores for reading fell 14 points from the last time the survey was taken, in 2022.

The survey also asked students about their AI use and found that daily AI users who reported using the tool for specific schoolwork tasks like drafting texts for writing assignments had science scores an average of 20 points lower than students who said they never used AI—equivalent to about a year of schooling. Daily users who reported using AI to aid learning and had access to AI literacy training did score slightly higher than “never” users, though the OECD noted those students tend to be wealthier.

Kaitlyn Ruano saw the effects of AI in the Advanced Placement Art History class she taught in Miami, where the school district made Google’s Gemini chatbot available to all high-school students last year. She was shocked in July when she saw a much larger portion of students than she’d expected hadn’t passed the AP test, including kids who had been getting A’s and B’s in class. “I had kids come to class after saying, ‘Miss, I had no clue what half the material was, I’m not going to lie to you. I’ve been using AI all year,’” she said. The Miami school district declined to comment.

“The Google one was the biggest temptation, because it’s the easiest to access,” said 15-year-old Sophia Chatzigianni of Fairfax County, Va. She saw her Human Geography and Algebra II grades nosedive to C’s on assessments as she relied on Google Search AI summaries and interactions with its AI mode chatbot on her school computer. Once she quit using AI, her grades improved.

Fairfax County Public Schools said it has developed its own plug-in to block Google AI search results that it plans to roll out to student devices soon.

Students seek shortcuts precisely at the moment the work starts to get mentally taxing, says Hamsa Bastani, a professor at the Wharton School of the University of Pennsylvania. “There’s just this constant choice between doing the easy thing and doing the hard thing,” Bastani said. “It’s even worse for kids.”

Google, which serves more than 170 million students and teachers worldwide through its educational software partnerships, offers a mix of free and paid AI tools. It markets Gemini as “your AI study buddy” to students. To teachers, Google promotes that its AI reduces workloads, “deepens metacognition,” “inspires active learning” and “stimulates curiosity,” and that it can level the playing field for struggling and special-education students. Google said its studies show that when teachers use AI to help with administrative tasks and lesson planning, they are able to invest more time into student mentorship.

Google Search’s AI Overviews and AI Mode chatbot are free and cannot be turned off in administrative controls that Google gives partner school districts. Students also have access to tools that offer Gemini-powered answers to screenshots or camera images.

“You cannot escape Gemini in Google Classroom,” said Mark Racine, former chief information officer for the Boston Public Schools.

“We undertake detailed public research and build our products in partnership with educators based on rigorous learning science, while giving teachers and administrators control,” Google said. It said its core education tools are “purpose-built for the needs and requirements of education contexts.”

Maria Scott, Richard’s mom, is seeking to get an individualized education plan to add some guardrails to his computer usage at school. With four adult children already through school, she feels her younger two are at a distinct disadvantage.

“They are learning to regurgitate whatever AI tells them,” the former middle-school teacher said.

Google has long viewed schools as a toehold for building lifelong brand loyalty. For more than a decade, it has worked with schools to make its offerings ubiquitous in public-school classrooms.

It courts superintendents and school officials through a yearlong fellowship program co-sponsored by venture-capital firm GSV, inviting them to Google’s Chicago offices and ensuring they are “grounded in the mastery of Google’s most transformative technologies—from Gemini to the broader ecosystem,” fellowship documents show. Fellows plan “Transformational Opportunity” projects for their districts—nearly all of which involve AI. Fellows even get a glass trophy.

Google also maintains an influencer network of more than 6,000 teachers, offering them “Certified” badges for attending its workshops about Google tools. Some teachers host training sessions for other educators and amplify positive social-media posts about Google and AI in school.

Google said the programs are designed to help teachers and school leaders elevate their skills, build community and reimagine instruction so students can have “future-ready” learning. It said fellows receive no product discounts, financial perks or requirements to use Google tools.

In New York City, the nation’s largest school district, top officials—several of whom are prior Google fellows—took a largely open approach to AI last school year. Google.org, the company’s philanthropic arm, gave a grant for the education department’s AI Advisory Council and its employees attended some meetings.

Google said its role was related to “funding, not shaping policy.”

Then declining test scores poured fuel on top of a brewing parent backlash. From 2025 to 2026, reading proficiency rates declined 6 points among third- through eighth-graders, the city said in August. Compared with 2023, when AI was still nascent, there’s been an overall drop from 51.7% to 50.3% of New York City students scoring proficient in reading and writing, with eighth-graders seeing the steepest decline.

A group of upset parents pressed the issue during an August education council meeting in Brooklyn’s District 19, where the officials included the district’s superintendent, a former Google fellow who had pushed AI use and was sporting a Gemini sticker on her iPad. Parents and teachers said consumer AI chatbots were fully unblocked for children on their school-issued devices starting last fall in the district, which serves mostly economically disadvantaged students.

Tion Prescott, a parent in the district, said in an interview he’s seen a slide backward in his two elementary-school children’s writing abilities. “You’re doing a disservice to an entire generation,” he said.

In early September, Mayor Zohran Mamdani banned student AI use in classrooms through eighth grade for the school year. Los Angeles quickly followed, restricting all student access. New York’s ban didn’t cover Google Search’s AI Overviews or AI Mode chatbot; L.A. said it blocked the AI mode chatbot using its own tools and has reached out to Google to turn off AI Overviews.

In a statement, New York City’s Department of Education said it’s working to better understand what happened with the test scores and that it’s reviewing fellowship participation, pointing out that no NYC staff are in the latest Google fellows cohort. It said the Google philanthropy grant was used for administrative purposes and professional development, and that it isn’t seeking additional funds.

The city said District 19’s AI programs had started under the prior administration and are now restricted.

Boston and Denver—districts where Google fellows have been leaders—promoted Gemini access for teenagers last school year.

In Boston, the percentage of 10th-graders who met or exceeded state thresholds on the 2026 standardized test declined 6 percentage points in English and science and three points in math, year over year. The share of seventh-graders meeting the mark for English and math also declined. Teachers described widespread student use of Google’s AI tools.

Denver Public Schools, which had all chatbots open through January and later pushed Gemini, saw English language arts proficiency rates drop more than 2.5 percentage points for sixth and eighth-graders in 2026. The percentage of 11th graders meeting or exceeding expectations in science, as well as SAT reading and writing, declined. A first-grade district lesson plan presented to teachers called “My Friend AI” suggested teachers use language like, “AI is like a smart helper that can learn and help us do things.”

Lauren Godbey, a ninth-grade teacher there, said: “This was the first year I have seen kids struggle just thinking of a topic.” Young Han Lester, another Denver teacher, said students “have begun to see Google’s AI summary as their main source of information.”

In statements, Denver and Boston public school districts said they are committed to helping students and teachers navigate AI in a way that facilitates learning. Denver said that attributing declines to a single tool is an oversimplification and that it views AI tools as “supplemental resources designed to enhance digital literacy and critical thinking.”

Boston said it adopted an AI policy in June governing responsible use and plans to “increase AI literacy.” In a press release about the test scores, Boston officials said they recognize the need to improve student academic performance.

A Common Sense Media survey of 1,200 children aged 9 to 17 in March found that three-fourths said they have used search engine AI-generated summaries, making it their most widely experienced form of AI.

When Common Sense Media tested how Google Search’s AI features react to homework questions, the tool completed kids’ homework every time, researchers found, and missed suicidal ideation and produced inaccuracies, among other problems, leading to an “unacceptable risk” rating.

Google said it disputed Common Sense Media’s methodology because it used “ambiguous and contrived queries.”

In Miami, veteran English teacher Cary McCormick said her students’ takes began to reek of sameness after the district enabled Gemini. When her class discussed the Joyce Carol Oates short story “Where Are You Going, Where Have You Been?”, many students commented that it was “reminiscent of childhood fairy tales” like Pied Piper, the nickname for the real-life serial killer the short story was based on. When McCormick asked them to explain the connection to Pied Piper, she said, “they didn’t know. It’s AI.”

In Georgia, Shannon Seibert observed the students in her high-school English class using Gemini for everything from how to draft a breakup text to asking “how many hours until 2 p.m.?” at 11:30 a.m. “They no longer trust what’s going on in their own brain,” she said.

Counselors and adolescent psychologists say students’ loneliness and social skills are worsening as they confide in AI tools—and it can be particularly risky for vulnerable kids.

Kimora Dean, who is autistic and has ADHD, had been thriving as a junior at John F. Kennedy High School in Silver Spring, Md., playing jazz piano and getting good grades. She began to use Google’s AI features for routine homework help on her school-issued Chromebook, her mother said, without her knowledge. Her daughter soon saw it as a confidante, chatting with Gemini and Google Search’s AI Mode chatbot for hours on her personal devices and her Chromebook.

When she complained about normal teenager things, like limits on her screen time and having to do chores, Google’s AI amplified her grievances back to her, according to her mother and partial chat transcripts reviewed by the Journal. It said that her mother’s efforts to restrict her screen time and AI usage constituted “coercive control, emotional abuse, and isolation” and urged her to plot her escape: “Go now and secure your freedom.”

Kimora frequently locked herself in her room. She began to lose weight. She refused to do her chores, saying she wanted to talk to Google’s AI instead.

Kimora was hospitalized at a psychiatric hospital in 2024 and again in 2025, Dean said. That’s when her mother got access to her phone and realized the extent of her conversations with Google’s AI.

Dean says her daughter, who graduated this past June, has stabilized on medications, but she is “forever changed.” She gets to use AI only one day a week now, on Fridays.

“This thing was a monster,” Dean said.

“While the technology presents new challenges, Gemini continues to improve, and we are committed to ensuring it plays a positive role in people’s well-being,” Google said. Google said it has prioritized youth protections, particularly around emotional dependence and reinforcement of false beliefs.

Montgomery County Public Schools said the district is finalizing AI guidelines for the next school year.

Several current and former Google employees said they feared that AI will travel the same arc as social media, where a Big Tech push fueled adoption among children, harms emerged five years later, and regulations to begin addressing those harms took 10 years to come about.

In a 273-page report from 2024 called “The Ethics of Advanced AI Assistants,” more than 50 Google researchers and ethicists warned that chatty, anthropomorphized AI poses particular risks to children and could cause “diminished intellectual engagement,” a “reduced sense of personal competence,” and “social degradation.”

“We learned nothing from social media,” said Nataliya Kosmyna, a visiting researcher at Google with a doctorate in human-computer interactions. “Attention is messed up, and in years we are going to have the same cases with LLMs with children and teenagers.”

Kosmyna believes based on her brain research that AI, like alcohol, should be restricted to ages 21 and over.

Google researchers have worried internally about how Gemini models have fueled user delusions and advised about social relationships in a way that amplifies slights and overpraises, which can be dangerous for children. AI child safety research nonprofit KORA has found in testing that Gemini and other companies’ models do poorly on “cognitive atrophy” scenarios like completing schoolwork, but Gemini fared worse than others on encouraging “parasocial attachment.”

Tajuddeen Williams, a social psychologist who consulted with Google last year to help train Gemini, said he tries to keep his children away from it as much as possible.

Google said it continues to update Gemini to provide more mental-health support, and that it works with academic experts, parents and teens to help ensure its models are safe and appropriate.

A 2025 Google-funded Sierra Leone trial saw students gain over a year’s math progress in eight weeks using Gemini’s “Guided Learning” mode, a tutoring feature that aims to guide students to think through questions rather than give direct answers. Still, it was the strongest students who benefited the most: “This is a common pattern in educational technology: new tools frequently widen achievement gaps rather than close them,” Google wrote.

Bob Comeau, a Boston high-school English teacher, is a lifelong tech enthusiast who has embraced AI in his classroom, including by creating a Gemini chatbot that students enjoyed based on the miserable, unnamed narrator in Fyodor Dostoevsky’s “Notes from Underground.”

But Comeau has also experienced all the downsides. He hates playing “cheating cop,” he said. “I’d ask them, ‘Do you want to fly in a plane with a pilot who cheated his way through flight school?’”

In 2023, 70% of his AP literature class passed the AP test. This past year, 65.5% did. It’s hard to know how much of it can be chalked up to Gemini versus other factors, he said.

His own daughter, a senior, is proud that she has never used AI and refuses to. She told her dad that she suspected one of her teachers used AI to generate nearly everything and grade.

“What does that do to a kid?” Comeau said. “There are some young teachers who are farming out the thinky part of teaching, and that scares me.”