Chapter 1 of 6
What These Tools Actually Do
Prediction, not knowledge — and why the difference matters in class.
7 min read
A prediction engine
A large language model predicts likely next words given everything before them. It has no database of facts to consult and no belief about truth. Fluency is its output; accuracy is a side effect of good training data.
Why it invents
When the likely continuation is a plausible-sounding citation, the model produces a plausible-sounding citation. This is not lying. It is the system working exactly as designed on a task it was never designed for.
The classroom consequence
Every factual claim from a model is a hypothesis requiring a source. Teaching that habit is the single highest-value AI lesson you can deliver.
Practice exercises
0 of 3 complete1.A language model produces a fake but plausible citation because it…
2.Write the one habit every student should apply to any factual AI claim.
3.In two or three sentences, explain to a student what the tool is actually doing when it answers a question.