(Credit: Thai Liang Lim/Getty Images)
Most American teenagers have grown comfortable with treating AI chatbots as substitute teachers, but far fewer say they've ever used one as a therapist, according to a new report from the Pew Research Center.
The top use for these conversational apps is something less exciting than either of those person-replacement scenarios: searching for information, which 57% of teens have done at least once. Those results might make Google's efforts to weave AI into its web-search experience seem more understandable.
Pew's finding that only 12% have ever used an AI chatbot for "emotional support" may also provide modest reassurance to parents of online teens who have seen so much coverage of AI chatbots hurting kids' mental well-being, sometimes to the point of self-harm or suicide.
As for parents worried about whether AI is the real reason behind their teens' suddenly improved grades? That's complicated. The survey found that 54% of teenagers have ever used AI chatbots for help with schoolwork. But 10% report using them to do "all or most" of it, 21% to do "some" of it, and 23% to do "a little" of it.
The top application of AI for schoolwork is "researching a topic," which 48% of respondents say they have ever done; 43% reported ever using a chatbot to help solve a math problem, and 35% said they have used one to edit their own writing.
Applying AI to learning can mean many different things. As one example, a Feb. 22 Washington Post article reported how English teachers have invited their students to use chatbots as writing coaches—while teaching them to question that advice and recognize ways it might lead them astray.
As another, a startup named Companion.ai is now pitching an Einstein chatbot that has drawn scathing criticism on social media for its touted ability to let a student skip remote learning on the widely used Canvas platform: "[Einstein] logs into Canvas every day, watches lectures, reads essays, writes papers, participates in discussions, and submits your homework—automatically."
"Students actually using the product have been really positive," Companion.ai CEO Advait Paliwal said in an email. "On the educator side, we've received everything from constructive concerns to threats telling us to take it down or we won't ‘sleep well' and that we're causing the downfall of society."
(The site for Einstein doesn't list pricing—a real dark pattern—but Paliwal says the service offers $40, $100, and $200 monthly plans. A PR rep for Instructure, the developer of Canvas, did not answer an email requesting comment.)
The Pew data suggest many students aren't ready to put that much trust in an AI to do their work. While 26% said chatbots have been "extremely or very helpful" for completing schoolwork, almost as many—25%—described them as only somewhat helpful, 3% said they were not helpful, and the remaining 45% weren't using chatbots for schoolwork.
A majority of respondents also don't trust their own classmates to use AI honestly in school: 34% reported that students at their school use AI to cheat very or extremely often, and 25% did so "somewhat often." Only 14% said fellow students rarely or never use AI to cheat, and 15% weren't sure.
The report includes additional details on the demographics of teens who lean on AI for most or all of their schoolwork, finding the highest proportion — 20%—in households earning less than $30,000 a year. In households making more than $75,000 a year, that figure was just 7%.
The other categories of even-once AI chatbot use reported in the Pew survey: "fun or entertainment," reported by 47% of respondents; summarizing an article, book, or video, 42%; creating or editing images and videos, 38%; getting news, 19%; and "casual conversation," 16%.
The Pew survey also asked some bigger-picture questions about how teens see AI affecting them and the world around them, and they don't seem quite sold on the whole thing. While 36% say they expect AI to have a positive effect on them over the next 20 years, 32% think it would be equally positive and negative, 15% expect negative consequences, and 17% aren't sure.
Respondents are less bullish when asked what they thought AI would do for society at large: 31% predict positive results, 34% are equally positive and negative, 28% are negative, and 8% are unsure.
The report includes some quotes from anonymized respondents that suggest they have thought more deeply about these possibilities than certain tech CEOs.
One teenage girl voiced optimism about AI's ability to free up time for human creativity: "It will do tasks that can be automated and allow people more time to do what they like." But a teen boy complains that "it's hard to tell what's real or AI online anymore" while another teen girl notes the obvious potential for abuse: "There are evil people in this world, and the wrong person could make AI turn against humans."
This survey did not ask respondents which chatbots they use. A separate Pew survey released in December found that 59% of teens had used OpenAI's ChatGPT at least once, far above the figures for Google Gemini (23%) and Meta AI (20%).
Pew's researchers also spoke with parents and found them supportive of most AI-chatbot uses, with 79% saying they're OK with using them to look up information, but opposed to turning to them for casual conversations (45% of parents didn't support that) or emotional support (58% not OK). And the most striking number in this study? Pew found 42% of parents had not talked to their teenagers about AI chatbots.
Pew used the research firm Ipsos to conduct this online survey via its KnowledgePanel, which drew responses from 1,458 US teens and their parents online from Sept. 25 to Oct. 9, with an overall margin of error of plus or minus 3.3 percentage points.