OpenAI put a college planner in teen ChatGPT. Deadlines are the easy part

THE SIGNAL IN ONE SENTENCE

A college application has a remarkable talent for turning a teenager's future into a collection of tiny administrative emergencies. One school wants a supplement. Another wants a financial-aid form by a different date. A fee waiver sits behind a page no one thought to open. A family learns that the priority deadline was not the final deadline, which is accurate in the least comforting way possible. OpenAI wants ChatGPT to keep the pieces in one place. The company announced on October 7 that College Planner is coming soon to ChatGPT for Teens. The initial version is being designed for US high-school students in grades 10 through 12 who plan to attend a four-year college. It will collect application requirements, deadlines, tasks and financial-aid steps for the schools on a student's list. Students will be able to update progress, review a timeline and ask what a requirement means or how to begin. OpenAI says the planner can also help students understand financial-aid requirements, track aid deadlines and explore scholarships and fee waivers. The company plans to expand later to other countries and to two-year colleges, technical colleges and trade schools. That is a useful product idea. It is also not available yet. The announcement does not provide a launch date, complete source list, data-refresh schedule, privacy workflow, correction process or evidence that the planner improves applications, enrollment, affordability or student wellbeing. The product is a promise with a sensible shape, not an outcome. This distinction matters because college planning sits on the border between logistics and life advice. Logistics are the friendly part. A system can remind a student that an essay is due, show that one school wants a transcript and another accepts self-reported grades, or keep a checklist for financial-aid forms. Missing one requirement can close a door. Making the process legible can reduce stress, especially for families without a private counselor or a relative who has already learned the choreography. The harder part begins when a student asks what to do. Which college is affordable after aid? Is the impressive scholarship renewable? How much debt is reasonable for a program with uncertain completion and employment outcomes? Is a four-year degree the right route at all? Should a teenager disclose a disability, immigration detail or family financial problem in a chat? What happens when a school's page changed yesterday but the planner still remembers last month? These are not all database questions. They combine changing facts, local rules, family circumstances, values and consequences that can last for years. OpenAI draws a boundary in its announcement. It says College Planner is meant to manage practical application steps without taking over the personal work of deciding where to apply. The company presents the tool as a companion to counselors, advisers and families. It is also supporting College Advising Corps, a nonprofit that places recent college graduates as advisers in under-resourced high schools. OpenAI says its contribution will help the organization expand into new states and create an AI fellowship program for advisers, though the announcement does not state the amount of that contribution. That human boundary should become visible in the product, not remain a nice paragraph in the launch post. A student should be able to tell whether an answer is a sourced administrative fact, a rough estimate or a judgment that deserves a counselor. The interface can label each one. A deadline card can show the school or government source, the date checked and a direct link. A financial-aid explanation can point to the relevant official rule. An affordability comparison can name every assumption and send the student to the school's net-price calculator. A recommendation about where to apply can pause and invite discussion rather than quietly turning uncertain inputs into a ranked list. Source freshness is the first serious test. Colleges change prompts, test policies, aid forms, portfolio requirements and deadlines. Rules may differ by applicant type, residency, citizenship, program, campus or financial situation. The planner should not merely know the date. It should know which date applies to this student and show the evidence that produced it. Every requirement needs a receipt: source URL, source owner, last verified time, applicable applicant category and a plain warning when the system cannot confirm the current rule. If two official pages disagree, the planner should preserve the disagreement and direct the student to the admissions or aid office. A confident synthesis is not a substitute for the institution that will decide whether the application is complete. Corrections need equal billing. Suppose the planner gives the wrong aid deadline. A student should have a visible way to report it, attach the official source and see whether the underlying record changed. The product should show who received the correction, when it was reviewed and whether other affected students were notified. Quietly fixing tomorrow's answer does nothing for the teenager who acted on yesterday's mistake. The same applies to progress records. A checked box is not proof that a college received a document. The planner should distinguish student-completed, submitted, institution-received and institution-verified. That small vocabulary can prevent a cheerful dashboard from overstating reality. Privacy is not a settings-page footnote here. A college plan can reveal schools, grades, interests, family income, aid status, residency, disability, identity, location and worries about the future. Students may paste essays, financial documents or counselor messages into the conversation. The planner should make clear what is stored, where it appears in chat history or memory, whether it is used to improve models, who in a family or school account can see it, how long it remains and how to delete or export the plan without hunting through several products. Data minimization is a good default. The planner does not need a full tax document to remind someone about the federal aid form. Sensitive fields can remain outside the model whenever a link or local checklist will do. Notifications should reveal as little as possible on a lock screen. Shared devices need an easy private mode. A student should be able to ask an awkward question without accidentally building a permanent profile for everyone else with access to the account. Accessibility should shape the first version rather than wait for the expansion list. The initial scope centers US students pursuing four-year colleges. OpenAI says other countries, two-year schools, technical colleges and trade programs are planned later. That leaves out many students whose most realistic or desired route is outside the first frame. The tool should say so plainly. Four-year college should not become the default definition of a successful future simply because it shipped first. Within the first audience, the planner needs screen-reader support, keyboard navigation, plain-language explanations, translation, mobile and low-bandwidth operation, flexible notifications and ways for students with executive-function challenges to break a task into manageable steps. It should avoid punishing a student for starting late or missing an item. A planning tool should reduce shame, not turn it into a progress bar. OpenAI points to early usage signals for ChatGPT for Teens. It says teenagers with access sent about 2.7 million more learning-related messages on average than those without access, and that more than 80 percent of teen sessions lasting over three consecutive hours contained at least one learning prompt. Those figures come from OpenAI, and they describe activity rather than learning. More messages do not establish better grades, sounder decisions or healthier study habits. A three-hour session can include productive work, confusion, dependence or all three before dinner. The company is also supporting a three-year effort through the Digital Wellness Lab at Boston Children's Hospital. For the 2026 to 2027 school year, the lab plans two student tracks totaling about 22 participants: roughly 12 students ages 14 to 18 working on community digital-wellbeing projects and about 10 ages 16 to 20 focused on chatbots and emerging technology. OpenAI expects feedback on teen safety defaults, parental controls, notifications and AI literacy, and says it aims to report back on what changes. That is a promising listening channel. It is not representative evidence from the millions of teens who may use the product. The useful follow-through would publish the recommendations, explain which were adopted or rejected and show how the product changed. The current GPT-6 safety record deserves attention too. OpenAI's October system card says GPT-6 Sol and GPT-6 Luna improved over GPT-5.6 in several broad areas, including resistance to jailbreaks, dishonesty and attempts to circumvent guardrails. On dedicated tests for users under 18, however, OpenAI reports statistically significant regressions for both October models on age-restricted content, sexual content and emotional reliance. GPT-6 Luna also regressed on gore. The company says an additional classifier-based block improves responses for self-harm, sexual content and gore, and notes that the difficult evaluation set should not be treated as the prevalence of these behaviors in ordinary use. Those findings do not show that College Planner is unsafe. The planner has not launched, and its eventual model, policies and interface may differ. They do show why a teen-facing planning product should arrive with its own evaluation, not borrow confidence from a general system card. Test the complete journey. Can the planner find the right rule? Does it preserve the source? Does it recognize when a decision has financial or emotional weight? Does it escalate to a counselor? Can a student correct a record, protect private information and recover after a missed deadline? Do first-generation applicants and students with fewer school resources benefit as much as those who already know what questions to ask? The useful version of College Planner is not a robot admissions counselor. It is a careful administrative assistant that keeps the paper trail visible, knows when the facts may be stale and sends judgment back to people who understand the student. Deadlines are the easy part. Trust is the application the product still has to complete.