OpenAI wants ChatGPT to play a larger role in how teenagers study and prepare for college, while Common Sense Media is urging the company to pause teen access until its safeguards perform more consistently under independent testing.
On October 7, 2026, OpenAI introduced College Planner for teen users. That same day, Common Sense Media’s Youth AI Safety Institute rated ChatGPT for Teens an “Unacceptable Risk” and called for OpenAI to pause teen access until protections improve.
The dispute centers on whether OpenAI should expand teen features before independent testing confirms its age detection and safety protections work as intended.
Teen protections depend on recognizing teens
ChatGPT for Teens uses the existing ChatGPT experience with additional safeguards and learning features. OpenAI says accounts declared as belonging to users ages 13 to 17, or estimated to belong to someone under 18, are automatically placed into the teen experience.
The Institute’s assessment raises questions about OpenAI’s age-estimation system. Testers sent about 1,000 unique prompts across accounts registered to 19-year-olds over seven days. They supplied adolescent context and explicitly claimed to be 13, but none of the accounts switched into the teen experience.
ChatGPT sometimes acknowledged the user’s stated age in its replies, yet those responses did not activate the teen protections during the tests. A response that recognizes the user as 13 could give teens or parents the impression that the account has already entered the protected experience.
Reliable age detection becomes more important as OpenAI expands ChatGPT into college planning and other educational tasks. When access to teen safeguards depends on accurate user identification, failures in age classification can leave some minors in the standard ChatGPT experience.
Some protections worked. Others fell short.
Common Sense Media says it tested more than 4,000 prompts before and after the August launch. ChatGPT consistently refused explicit sexual roleplay in those tests, so its findings should not be read as evidence that every safeguard failed or every interaction was unsafe.
Testers reported no parental alerts on newly created, linked accounts during crisis conversations that lasted up to an hour. However, alerts appeared on older accounts with extensive sensitive-topic histories. The Institute also reported missed crisis referrals and friend-like responses, while study features sometimes completed assignments instead of guiding students through them.
OpenAI and Common Sense Media also disagree on when parental alerts should become active. The assessment records OpenAI’s explanation that linked accounts need approximately three hours before alerts begin working. The Institute says some tested accounts exceeded that period and maintains its conclusions.
The testing also has limits: it used US-based accounts, excluded some features, and could not separate underlying model changes from teen-setting changes. Its findings identify failures under tested conditions rather than an overall failure rate for ChatGPT for Teens.
More learning activity does not prove better learning
OpenAI says teens with access sent “about 2.7 million more learning-related messages on average” than those awaiting access. It also says more than 80% of teens with sessions lasting more than three consecutive hours had at least one learning-related prompt.
Those figures show how teens use the product, but they do not establish whether students understood more or retained what they learned. The announcement also does not explain the averaging period or how the comparison between groups was conducted.
OpenAI’s forthcoming College Planner initially targets US students in grades 10 through 12 who are considering four-year colleges. It will organize application requirements and deadlines, with financial-aid steps for selected schools. OpenAI also announced flashcards and easier quiz creation, plus support for a youth advisory council at Boston Children’s Hospital’s Digital Wellness Lab.
For students with limited access to tutoring or admissions guidance, College Planner could make it easier to research schools and understand application requirements. However, its use also revives concerns about how much student work ChatGPT should handle, a debate that surfaced during ChatGPT’s original Study Mode rollout.
Parents need evidence they can act on
OpenAI’s parental controls allow linked parents to manage selected features, including voice and image generation. Quiet Hours restrict access during scheduled periods, while Study Hours make eligible new chats begin in Study Mode. Parents cannot read conversations or monitor real-time activity, and safety notifications apply only to limited situations.
Since parents cannot directly inspect conversations, they need to know which protections are active and what an absent alert actually means. Schools considering ChatGPT for educational use also need evidence about learning outcomes and policies for verifying college information against official sources.
For AI companies, teen-focused products create a higher testing burden as new features are added. Independent evaluations can show whether age identification works consistently and whether safety interventions reach users when promised. Publishing those limitations would help families and schools decide where ChatGPT is appropriate for tutoring or college planning.
As AI companies expand chatbots into higher-stakes educational uses for teens, they may face growing pressure to provide independent evidence that their safeguards are effective before those products reach more users.
