The number of AI tools marketed at students grows every month, and most institutions have no structured way to help students tell a genuinely useful one from a gimmick. A directory can fix that — but only if it's designed around evaluation, not just a list of logos.
Start with categories that match how students actually search: by task ("summarise a reading," "practise a language," "check my writing"), not by vendor name. A directory organised by brand assumes the student already knows what they're looking for.
Each entry needs the same short profile: what it actually does, what it costs, what data it collects, and — critically — where it's known to get things wrong. A directory that only lists strengths reads as marketing, not guidance, and students learn to distrust it fast.
Build in a feedback loop from day one. A static list goes stale within a semester; a small "was this useful / accurate for you" prompt keeps the directory honest and gives the team maintaining it a reason to revisit entries regularly.
Finally, decide who's accountable for keeping it current. A directory with no owner is the fastest way to end up recommending a tool that changed its pricing, its data policy, or shut down entirely six months ago.
This is the concept behind a small AI family-task-assistant project I'm currently testing — the same logic (classify by task, be honest about limits, keep it current) applies whether the "student" is a university student or a parent managing a household.