ChatGPT for Teens: What It Is and How to Use It for Studying
ChatGPT for teens explained: age rules, parental controls and Study Mode, plus what research says about using the chatbot to study effectively.
OpenAI rolled out parental controls for ChatGPT on September 29, 2025, its most direct response yet to the millions of teenagers using the chatbot for schoolwork. The launch followed Study Mode, a tutoring-style feature released in July 2025. Together, the two updates signal that ChatGPT for teens has become a product category in its own right, not an afterthought of the adult version.
The stakes go beyond one company’s roadmap. Teen adoption of chatbots has doubled in two years, schools are rewriting homework policies, and researchers warn that the wrong usage patterns can quietly erode learning. How a 15-year-old opens ChatGPT tonight — and under which settings — now shapes both grades and study habits.
- According to Pew Research Center (2025), 26% of U.S. teens used ChatGPT for schoolwork in 2024, double the 13% recorded in 2023.
- OpenAI requires users to be at least 13 years old, and minors under 18 need parental consent under its terms of use.
- Study Mode, launched in July 2025, makes ChatGPT guide students through problems step by step instead of handing over answers.
- Parental controls, released in September 2025, let a parent link accounts, set quiet hours and disable voice, memory and image generation.
What ChatGPT for Teens Actually Includes
ChatGPT for teens is not a separate app but a layered configuration of the standard chatbot: a minimum age of 13, parental consent under 18, automatic content restrictions for linked teen accounts, and optional features such as Study Mode, all introduced by OpenAI between July and September 2025.
The teen experience rests on three pieces. First, the account rules: OpenAI’s terms have long required users to be 13 or older, with parental permission up to 18. Second, Study Mode, which any user can toggle on. It uses Socratic questioning, checks understanding before advancing, and withholds direct answers to homework-style prompts.
Third, parental controls. Since late September 2025, a parent can send an invitation from their own account to a teen’s account. Once linked, the teen’s version of ChatGPT applies stricter content policies by default. Parents can also reduce exposure further, though they cannot read their child’s conversations — a deliberate privacy line OpenAI has defended publicly.
OpenAI has also said it is building an age-prediction system to estimate whether a user is under 18 and route them to the restricted experience automatically. Where the system is unsure, the company stated it will default to the teen version.
«We prioritize safety ahead of privacy and freedom for teens; this is a new and powerful technology, and we believe minors need significant protection.»
The Numbers Behind Teen Adoption
According to Pew Research Center (2025), 26% of U.S. teens aged 13 to 17 said they had used ChatGPT for schoolwork, up from 13% in 2023. Awareness is near-universal: 79% of teens in the same survey had heard of the chatbot.
The growth is uneven by subject. Pew found teens judge the tool most acceptable for researching new topics, and least acceptable for writing essays. That distinction matters, because essay-writing is precisely where detection is hardest and where academic-integrity disputes concentrate.
Usage also extends well beyond homework. Common Sense Media reported in 2025 that 72% of U.S. teens had tried AI companions — chatbots used for conversation rather than tasks. The finding pushed regulators and OpenAI itself toward the safety features now bundled into teen accounts, including distress-detection routing announced after a high-profile lawsuit in August 2025.
For schools, the practical consequence is that blanket bans have quietly lost ground. If a quarter of students already use the tool for coursework, policy has shifted from prohibition toward defining acceptable use.
How to Use ChatGPT for Studying Without Hurting Learning
A 2024 study from University of Pennsylvania researchers found that students using a standard GPT-4 tutor scored 48% better on practice problems, but 17% worse on the final exam once access was removed. A version constrained to give hints, not answers, erased that damage.
The Wharton-led experiment, run with nearly a thousand high-school math students, is the most cited evidence in the teen-AI debate. Its lesson is not that chatbots harm learning, but that answer-extraction does. Students who outsourced the reasoning learned less; students who received scaffolded guidance did not.
That is the design logic behind Study Mode, and it maps onto a broader distinction between delegating work and delegating thinking — a line explored in depth in AI Can Do Your Homework: How to Use It and Still Actually Learn. The same prompt can be a shortcut or a tutor depending on what the student asks for.
| Study pattern | What the student asks | Effect on learning (per available research) |
|---|---|---|
| Answer extraction | «Solve this problem» / «Write this essay» | Better short-term output, worse exam performance when access is removed (UPenn, 2024) |
| Scaffolded tutoring | «Give me a hint, don’t tell me the answer» | No measured harm; mirrors Study Mode’s default behaviour |
| Retrieval practice | «Quiz me on this chapter and correct my mistakes» | Aligns with testing-effect research; the student still does the recall |
| Explanation on demand | «Explain this concept in a different way» | Low risk; comparable to consulting a second textbook |
EdTech startups have converged on the same conclusion. Study apps ranging from Khanmigo to Spanish startups such as Modo Cheto increasingly restrict direct answers by design, betting that guided practice is what parents and schools will pay for.
The Limits of Parental Controls
OpenAI’s parental controls, live since September 2025, require the teen to accept the account link and can be unlinked by the teen at any time, with a notification sent to the parent. They restrict content and features, but they do not grant parents access to conversation transcripts.
The design reflects a trade-off OpenAI has made explicit: teen safety over surveillance. Parents can set quiet hours, disable image generation and opt the teen out of model training. They receive alerts only in narrow cases, such as when the system flags signs of acute distress and a human reviewer confirms the risk.
The gaps are equally clear. A teen can create an unlinked account with a different email, use a logged-out session, or switch to a competing chatbot with weaker controls. Age verification remains self-declared until the age-prediction system matures. Child-safety groups, including Common Sense Media, welcomed the controls in 2025 while noting they depend heavily on voluntary family adoption.
For families deciding how to configure access, the more durable lever may be usage norms rather than settings — agreeing on when the chatbot explains and when it stays closed, as covered in StudyVerso’s guide to learning with AI assistance.
What It Means for Students, Parents and Schools
The consolidation of ChatGPT for teens as a distinct, controllable product shifts responsibility outward: OpenAI now provides the guardrails, but their effect depends on whether families link accounts, whether schools define acceptable use, and whether students choose tutoring over answer extraction.
For students, the practical takeaway from the research is narrow but firm. Used as a quiz generator, an explainer or a hint engine, the chatbot behaves like an always-available tutor. Used as an answer machine, it produces grades that evaporate under exam conditions.
For parents, the controls are a floor, not a solution. They filter content and limit features, yet the highest-impact decisions — which mode, which prompts, which habits — happen inside conversations no parent will read.
For schools, the burden is curricular. Assessment formats built on take-home essays predate a world where 26% of students carry a fluent writing engine. The institutions adapting fastest are moving evaluation toward in-class work and oral defence rather than detection software, whose reliability remains contested.
The open question is no longer whether teenagers will study with ChatGPT — a quarter of them already do — but whether the version they meet is the tutor or the answer machine. That outcome will be decided less by OpenAI’s settings screen than by the first prompt a student types on a Sunday night.
