Teaching with AI

OpenAI has put out a guide aimed squarely at teachers, and it reads less like marketing copy and more like a syllabus for adults who never signed up to be AI experts overnight. The document walks through how ChatGPT actually works under the hood, in plain terms, and spends real time on where it breaks down—hallucinated facts, confident wrong answers, the usual suspects.

What stands out is the section on AI detectors. OpenAI is blunt about their unreliability, which is notable coming from the company whose product those detectors are supposedly catching. Schools have leaned hard on these tools since ChatGPT launched in late 2022, often suspending or failing students based on shaky evidence. This guide basically tells teachers: don't trust that number on the screen the way you've been trusting it.

There's also a chunk devoted to bias, walking through how a model trained on internet-scale text can reproduce stereotypes or skew outputs in ways that aren't obvious unless you're looking for them. Pairing that with a set of suggested prompts turns the guide into something closer to a lesson plan than a policy memo—teachers get concrete starting points for using ChatGPT in lessons, not just warnings about what could go wrong.

The timing fits a pattern. Two and a half years after ChatGPT upended homework norms overnight, schools are still catching up, and most training has come from districts scrambling on their own or third-party edtech vendors with their own agendas. OpenAI stepping in with its own explainer changes who's setting the terms of that education, for better or worse.