Evidence at a glance
A Chat Assistant Takes On the Administrative Work
OpenAI has announced College Planner, a forthcoming addition to ChatGPT for Teens. It is designed to bring application requirements, deadlines, tasks, and financial-aid steps for a student’s chosen schools into a plan they can update over time. The point is not to choose colleges for students, but to help them see what needs doing, what remains, and when to move forward. The initial version is planned for US students in grades 10–12 applying to four-year colleges. It has not yet been broadly released.
The change matters because the product’s unit of work is shifting from a single answer to an ongoing process. A student might previously have asked ChatGPT what a requirement means. Planner aims to connect that explanation to a timeline and task list the student can revisit and update. OpenAI says hundreds of thousands of US teens already use ChatGPT each week to plan college applications, so the product is being built around a task users have already brought to the chat interface.
Organizing Requirements Is Not the Same as Verifying Them
Planner breaks application management into concrete steps: collect school requirements and deadlines, organize tasks, track financial-aid steps, and let students update their progress. Students can also ask what a requirement means or how to get started. For someone applying to several schools, keeping these items in one plan could reduce the effort of switching among webpages, notes, and reminders. OpenAI presents the tool as an addition to support from families, counselors, and advisers, not a replacement for them.
But centralizing information is different from ensuring that it is correct. The announcement does not explain how the system will source, verify, or update college requirements, or how it will alert students when those requirements change. Product teams should therefore evaluate more than the usability of the task list. They should test whether each requirement has a clear source and freshness signal, and what happens when information is wrong. A polished timeline can make an incorrect deadline feel more trustworthy, not less.
Adoption Is Not Evidence of Learning
OpenAI also reports substantial use of ChatGPT for Teens’ learning features. Nearly 1.2 million teens used Learning Visualizations in one week, and more than 180,000 used Study Mode. The product is also expanding the path from notes to revision materials: students can photograph multiple pages and combine them into a PDF, then generate flashcards or interactive quizzes. Flashcards can be marked as mastered or missed and practiced again. The multi-page note feature is available on iOS, while Android support is still in development.
Those figures indicate adoption, not that students have learned more. OpenAI also says that, as access expanded, teens with access sent about 2.7 million more learning-related messages on average than teens who did not yet have access. The announcement does not define the statistical unit behind “on average” or give the comparison period, and it provides no independent measure of learning outcomes. For an education product, message counts, opens, and generated flashcards show that activity occurred. They cannot stand in for measures of understanding, retention, or the quality of completed work.
Short Sessions Do Not Replace Youth Safety Evaluation
OpenAI reports that teens spend less than 15 minutes a day on ChatGPT on average, and that fewer than 2% use it for more than three consecutive hours. In conversations that receive a break reminder, nearly half of users take a break or end the conversation within five minutes. The announcement compares this with a publicly reported average of five hours a day on social media, and says that more than 80% of sessions lasting over three hours included at least one learning-related prompt.
These details help describe usage patterns, but they do not establish safety or learning outcomes. Social media and a chat assistant are different experiences, and an average cannot explain what is happening among the small share of heavy users. Nor does a learning-related prompt in a long conversation show that the student learned the material. Teams building for teens should evaluate break reminders, session duration, and learning tasks separately, including how each affects different groups of users.
Student Input Matters Only If It Reaches Decisions
OpenAI also says it will support the Student Advisory Council run by Boston Children’s Hospital’s Digital Wellness Lab. Teen participants will test tools and offer suggestions on topics that may include safety defaults, parental controls, notifications, and AI literacy. The council is independently operated by the lab, not a new internal OpenAI governance body. For the 2026–27 school year, the program plans to bring together about 22 students, including roughly 10 focused on AI chatbots and emerging technologies.
It is important to distinguish student participation from corporate accountability. An external council can provide feedback from young users, but its existence does not show that recommendations will change a product. The announcement does not say which suggestions will be adopted, how disagreements will be handled, or how decisions will be reported back to participants. For product leaders, the useful test is not whether an advisory seat exists, but whether feedback has a traceable route into requirements, safety reviews, and release decisions.
College Planner should not be judged by usage alone. More meaningful tests would ask whether students miss fewer application or financial-aid steps, whether information stays accurate and current, and whether students with less support gain access to guidance they can understand. A product can help manage a process, but it cannot replace the judgment of students, families, schools, and advisers. Until those boundaries and outcomes are made clearer, Planner is best treated as a workflow tool whose value remains to be demo