OpenAI shipped a teen ChatGPT. The over-65 cohort doubled to 23% and got nothing.

On August 18, OpenAI announced ChatGPT for Teens, a version of the chatbot with stronger safety protections, parental controls, and more age-appropriate responses ^[1]. The durable part of that launch is not the feature list but the precedent: age cohort is now a thing you ship a build for, and the segment that doubled its usage in a year did not get one ^[2][1].

Pew Research Center found earlier this year that 23% of more than 5,000 surveyed adults aged 65 and up use ChatGPT, against 10% the year prior ^[2]. That is a 2.3x increase over twelve months ^[1]. AARP's internal survey, reported by Business Insider, put AI use among people over 50 at 30% in 2025, up from 18% in 2024 ^[3], a 12-point move ^[2]. AARP is not a bystander here: the organisation recommends ChatGPT for companionship and mental stimulation ^[4].

What the cohort does with the product is the part worth reading twice. One survey cited by Fast Company, not yet peer reviewed, found that 63% of older adults had used ChatGPT for medical advice ^[5]. Fast Company also reports that older users are skipping intermediate platforms and going straight to systems like Claude and ChatGPT, partly because they can text or talk directly to them ^[6], and that for some of these users Chat is starting to look like the new search bar ^[7]. The same piece notes they bring sensitive material with them, including intimate health details and retirement planning ^[8], and that older adults are losing billions of dollars to online fraud and AI-enabled scams ^[9].

That is a different failure surface from the teen one. A teen build is mostly subtractive: restrict, refuse, escalate to a parent ^[1]. The over-65 case runs the other direction. Someone using a chatbot as a search engine for a niche surgeon and a confidant for retirement numbers is exposed less by content that should have been blocked than by answers that are confidently wrong and links that should not be followed. And there is no obvious analogue to the parental control account in the senior case, which is probably part of why the build does not exist.

The gap is being noticed outside product orgs. A handful of Democrats wrote earlier this year to tech giants including OpenAI about implementing safeguards for older users ^[10].

Two caveats, both from the source making the argument. It is not clear how effective the teen controls will turn out to be ^[11]. And Fast Company's own conclusion is not that OpenAI should ship a senior edition, but that teenagers are far from the only vulnerable group using the product, because excessive credulity and misplaced authority are age-neutral risks ^[12] and dependent relationships with chatbots form across every age group ^[13].

Watch whether ChatGPT for Teens behaves like a template, meaning age tiers plus verification that get reused, or like a one-off built for a specific news cycle. Watch Pew's next wave for whether 23% moves again at the same rate ^[2]. And watch whether the congressional letters harden into anything with a date attached ^[10], because until then the cohort with the most money to lose and the most medical questions to ask keeps getting the default build.