AI: the essential starting point
Before using AI to research, write, transcribe or verify material, understand the kind of tool in front of you. Generative systems can produce fluent language and convincing media. Their ability to sound right must remain separate from the journalist’s task of establishing what is true.
The essentials
AI already appears in familiar tools: recommendation feeds, automatic transcription, translation and image processing. Generative AI produces text, images or audio from a prompt. Distinguishing the task and system matters more than treating everything carrying the AI label as a single editorial capability.
A language model generates text by predicting what is likely to follow in a context, using patterns learned during training. Fluent output is therefore not, on its own, the result of a verified investigation. A system can state a correct name or date and an invented one with similar confidence. Fabricated output of this kind is commonly called a hallucination.
Keep the two properties together: speed and fluency can be useful, while the same ease can disguise unsupported detail. Treat the tool’s answer as something to examine, not a source that settles the question. The responsibility for recognising uncertainty, checking facts and deciding what can be published remains with the journalist.
Take it to the field
Ask the tool about something you know well, then check names and dates. The exercise reveals misleading confidence: an elegant formulation is never, by itself, evidence that an answer is true.
Your turn
Ask an AI system five factual questions about your city or beat whose answers you know. Classify each answer as accurate, plausible but false, or an admission of uncertainty. Use the table as a first audit.