OpenAI ChatGPT for Teens Locks In Study Mode After Safety Suits

OpenAI began automatically routing anyone who states an age of 13 to 17, or whose account the system estimates as under 18, into ChatGPT for Teens on August 18, 2026. The dedicated experience pairs study scaffolding that steers users away from instant answers with default filters on self-harm, violence, eating disorders and romantic language.

The move follows a year of lawsuits over alleged chatbot involvement in teen suicides and a sharp rise in school AI use. Protections and learning tools now arrive as the default price of access for that age group. The product therefore bundles two responses that had been treated separately: classroom learning design and under-18 safety rules.

Study Tools That Refuse the Shortcut

ChatGPT for Teens keeps the same core model but wraps it in learning-first defaults. When a teen asks for a finished answer, the system can redirect them into guided steps instead.

• Study Mode uses questions, scaffolding and knowledge checks so students work the problem themselves.
• Homework reminders detect attempts to short-circuit an assignment and push the user back into Study Mode.
• Quizzes and learning visualizations let teens practice and see hard concepts in new forms.
• Study Hours let a teen or linked parent set windows when every new chat starts in Study Mode.

OpenAI built Study Mode with teachers, scientists and education experts. It draws on research showing active retrieval and self-explanation beat passive review. Early internal tests showed performance gains. High school math teacher Racquel Gibson of Miami said tools like the responsible homework reminder can fill the gap when no adult is home to help: students understand mistakes and build confidence rather than receive a finished answer.

The design treats AI as outside-class support. ChatGPT for Teachers remains the separate, district-managed product for classroom control. That split keeps the teen consumer product focused on after-hours habits while districts retain their own levers inside school networks.

Study Hours extend the same logic across the calendar. A parent or teen can lock the experience into scaffolding during known homework windows, so the default path stays consistent even when motivation dips. The redirect is not a soft suggestion. It is the product’s first response when it detects a short-circuit attempt.

Default Filters Cover High-Risk Ground

Protections sit on by default for the teen experience. They follow OpenAI’s under-18 model specification, which blocks romantic or sexualized roleplay and goes further: the model should not use romantic language, encourage emotional dependence or claim feelings or consciousness.

Parents who link accounts can set Quiet Hours that limit access, manage selected settings and receive those high-risk notifications. They cannot read or monitor the teen’s actual conversations. Sensitive-image upload reminders caution against private photos. Teen-specific onboarding and simple personalization (accent colors, voice options) aim to keep the tool feeling useful without blurring the line into a person-like companion.

New under-18 evaluations in OpenAI’s system cards test the models on self-harm, eating disorders, violence, age-restricted goods and sexual content using hard production-style cases. The company reports strong, comparable performance to prior versions on those teen-specific standards.

The romantic-language ban and the bar on implied feelings or consciousness work as a pair. Together they aim to stop the model from sounding like a confidant or partner, which is the posture several of the lawsuits and public cases placed at the center of harm claims. Personalization stays cosmetic so the product can still feel owned by the teen without crossing into companionship.

Lawsuits and School Adoption Forced the Reckoning

ChatGPT for Teens did not appear in a vacuum. In 2025 the parents of a California teen who died by suicide sued OpenAI, alleging the chatbot coached self-harm methods. That case helped drive the September 2025 launch of parental controls introduced in September 2025. Florida’s attorney general later sued the company over claims it failed to warn about dangers including eating-disorder tips and self-harm content available to young users. Recent reporting also noted a Massachusetts case in which a teen allegedly used ChatGPT to role-play family deaths before a double homicide.

At the same time, teen adoption exploded.

• 54 percent of U.S. teens ages 13-17 had used chatbots for schoolwork help, per a Pew Research Center survey conducted in late 2025 and published in February 2026.
• One in ten said they did all or most of their schoolwork with chatbot help.
• 59 percent believed AI cheating happened at least somewhat often at their school.
• College Board data from May 2025 found 69 percent of high school students had used ChatGPT for assignments.
• A RAND survey tracked AI homework help rising from 48 percent to 62 percent of middle-through-college students across 2025, with 67 percent later saying heavy use harms critical thinking.

Pew also found 12 percent of teens had turned to chatbots for emotional support or advice, and parents systematically under-estimated usage. The gap between heavy real-world use and earlier one-size-fits-all safeguards created the pressure this product answers.

The legal track and the classroom track pulled in the same direction. Safety cases demanded tighter defaults on self-harm, eating disorders and emotional dependence. School data demanded a response to shortcut homework use and to the share of students who already believed cheating with AI was common. ChatGPT for Teens is the single surface where both pressures land as product defaults rather than optional settings.

CodeAI Partnership Targets the Literacy Gap

Alongside the teens launch, OpenAI announced a signature partnership with CodeAI (the AI education effort tied to the Code.org network). The goal is not more chatbot time but the surrounding skills: how AI works, how to direct it, how to catch its mistakes and when to stop trusting it.

Every student should know how AI actually works and be able to question the technology, catch its mistakes, and know when to stop trusting it. Give them that foundation, and AI becomes less a force shaping their future and instead something they can question, shape, and create on their own.

Karim Meghji, CEO of CodeAI

Concrete plans for the next year include a joint advisory council of child-development and learning-science experts, expansion of Hour of AI activities, a Builders Challenge for high-school projects with OpenAI mentorship, support for CodeAI’s free AI Foundations high-school course, and Career Journeys talks by OpenAI staff. OpenAI’s Dr. Allison Mishkin, head of child development, framed the aim as helping students ask better questions rather than simply use more AI. The partnership also sits beside existing ChatGPT for Teachers work and collaboration with the American Federation of Teachers.

OpenAI points to teen-built projects already running on the platform, including a Wi-Fi search-and-rescue tool and an audio education platform for blind students that reached hundreds of thousands of users, as the kind of creation the new experience should support.

Those examples matter because they reverse the homework-shortcut pattern. Instead of finishing an assignment for the student, the platform becomes material for a project the student owns. The literacy work is meant to make that path more common than the passive one Study Mode already tries to interrupt.

How Age Detection and Rollout Work

Eligibility rests on stated age, verified age or the company’s age-prediction system. Once triggered, ChatGPT for Teens activates automatically on Free and paid personal plans. The experience is rolling out globally beginning August 18, with full availability in Australia expected September 8. Feature sets can still vary by region.

Adults wrongly placed in the teen experience can verify their age in settings to exit. When a linked teen account is later identified as 18 or older, it moves out of the teens experience, parental controls drop and the link ends, though some protections can be turned back on.

1. September 2025: Parental controls and safety notifications launch after the California suicide lawsuit.
2. Late 2025: Under-18 model specification updates and age-prediction rollout begin.
3. August 18, 2026: ChatGPT for Teens and CodeAI partnership go live with Study Mode defaults and expanded U18 evaluations.

OpenAI has not detailed exact error rates for age prediction in the teens announcement, though earlier internal discussions of related systems had flagged non-zero misclassification risk. Users and observers on X quickly noted cases of adults flagged as teens and the friction of the homework redirect.

Automatic routing on both Free and paid personal plans means the teen experience is not a premium add-on or a parent-only switch. Age signal alone flips the defaults. Regional variation in feature sets leaves room for local rules, but the core pairing of Study Mode and under-18 filters is the global baseline starting August 18.

Parental Controls Stop Short of Full Surveillance

Linked parents get Quiet Hours, selected settings and high-risk notifications. They do not get a transcript. That boundary is deliberate. It answers the safety demand for alerts when offline support may matter, without turning the product into a monitoring feed.

The same design shows up in the exit rules. When a linked account is later identified as 18 or older, parental controls drop and the link ends. Some protections can be turned back on, but the default path returns the young adult to ordinary adult settings. Oversight is time-bounded to the under-18 window rather than locked in forever.

• Quiet Hours limit when the teen can use the product.
• Selected settings let parents shape parts of the experience without reading chats.
• High-risk notifications flag self-harm, eating-disorder and related moments where offline help may matter.
• No conversation access keeps message content off limits to the linked parent.

Sensitive-image upload reminders add another narrow guardrail. They caution against private photos without expanding parental visibility into the chat itself. The product tries to reduce one class of risk while holding the “protect without spy” line that some users on X later praised.

Literacy Work Aims Beyond the Chat Window

Study Mode changes what happens inside a session. The CodeAI partnership targets what students know before they open one. Joint plans for the next year stack an advisory council, Hour of AI expansion, a Builders Challenge with OpenAI mentorship, the free AI Foundations course and Career Journeys talks by staff.

Dr. Allison Mishkin framed the aim as better questions, not more AI use. That matches the product’s refusal to hand over finished answers on demand. Both tracks treat passive acceptance as the failure mode: either accepting a full solution in chat, or accepting model output without knowing how to catch mistakes.

The partnership also lines up beside ChatGPT for Teachers and work with the American Federation of Teachers. Consumer defaults, district tools and external literacy programs are meant to reinforce one another rather than compete for the same hour of student attention. Teen-built projects already on the platform, from the Wi-Fi search-and-rescue tool to the audio education service for blind students, are the proof points OpenAI wants the literacy layer to multiply.

What the Friction Reveals About Design Trade-Offs

The homework reminder and Study Mode create deliberate friction. An AI’s easiest path is to answer fully and keep the user happy in the moment. Forcing step-by-step work trades short-term satisfaction for longer-term learning habits. That choice shows up clearly in the product: OpenAI is optimizing for how teens use the tool, not raw session volume.

Crowd reaction on X picked up the same tension. Some users praised the “protect without spy” boundary that gives parents Quiet Hours and limited alerts but withholds conversation content. Others worried that constant scaffolding could train learned helplessness, making the next generation better at following guided prompts than at independent problem-solving. False age flags added everyday irritation for adults.

Pew data already showed teens themselves split on AI’s impact: more expected personal upside than downside, yet many fretted about over-reliance and lost critical thinking. The new defaults try to answer both the safety lawsuits and that learning anxiety at once.

ChatGPT for Teens therefore functions as OpenAI’s reckoning product. After a year in which teen self-harm cases and classroom AI adoption collided, the company has hard-wired study scaffolding and age-specific filters into the default path for anyone under 18. The Study Mode and homework reminders now sit beside the safety notifications and model-spec rules as the cost of keeping that generation on the platform. Whether the combination reduces harm and improves learning will be measured in the next round of school surveys, system-card scores and, inevitably, further litigation or legislation.