Communications drafts a recycling campaign with a large language model then fact-checks diversion rates. What GenAI literacy practice is essential?

Generative AI literacy for municipal communications means using large language models to accelerate drafting while staff remain accountable for factual accuracy before release. A recycling campaign draft from an LLM can help with tone and structure, but diversion rates, tonnage claims, and program eligibility details must be verified against city data or program owners. The essential practice is generative assistance plus human fact-checking—especially for quantitative civic claims residents may rely on.

Using an LLM to draft campaign copy while staff verify diversion rates against training data before release uses the wrong source of truth: an LLM's training data is not a current, auditable source for city diversion rates, which must be verified against city data or program owners. Using an LLM to draft campaign copy while staff ask the same LLM to reconfirm facts before release keeps verification inside the same model: the LLM can confidently repeat its own hallucinated diversion rates, so staff must still verify against city data or program owners before release. Using an LLM to draft campaign copy while staff accept LLM-provided citations for facts before release trusts unverified references: LLM citations can be fabricated or irrelevant, so diversion rates must still be verified against city data or program owners before release.

Exam caveat: fluent campaign copy is not evidence that embedded statistics are correct; fluency and truth are separate. Operational check: maintain a source-of-truth sheet for diversion metrics, require communications and program staff sign-off, watermark drafts as unverified until checked, and refuse to publish any numeric claim the city cannot cite.