AI features are getting added to student-facing websites faster than most institutions can write policy for them — chat widgets, recommendation panels, auto-summarised content. It's worth pausing on a few basics before adding another one.
Disclose it. If a student is talking to an AI, they should know within the first message, not discover it later. This isn't just an ethics preference — it changes how people phrase questions and how much they trust the answer.
Always leave a human door open. Every AI-facing feature needs a visible, low-friction way to reach a person. "Contact us" buried three pages away doesn't count — if someone's frustrated enough to want a human, make that the easiest click on the page.
Say what it doesn't know. An AI feature that never expresses uncertainty will eventually state something wrong with full confidence. Interfaces that show "I'm not certain about this — please confirm with [department]" build more trust over time than ones that always sound sure.
Keep the data trail short. Student-facing AI tools should collect only what's needed to answer the question in front of them, and institutions should know exactly where that data goes and who can see it.
Make it accessible too. A chat widget that isn't keyboard-navigable or screen-reader friendly just adds a new barrier while claiming to remove one.
None of this requires exotic technology — it requires the same discipline as any other public-facing service: be honest about what it is, make it easy to escalate, and don't let it speak with more confidence than it's earned.