Common Prompting Mistakes (And What Fixes Them)

A quick tour of the mistakes that come up most often once you're past the basics — and the specific, small fix for each one.

Mistake 1: Bundling multiple unrelated asks into one prompt

Asking for a summary, a tone rewrite, and a list of follow-up questions all in one message tends to get a shallow pass at all three instead of a good job on any one of them.

Fix: Split unrelated asks into separate messages, or clearly numbered, distinct sections if they truly need to happen together.

Mistake 2: Assuming the AI remembers a previous, separate conversation

Each new conversation typically starts with no memory of a different one you had yesterday — unless the specific tool you're using has an explicit memory feature you've enabled.

Fix: If earlier context matters, restate it briefly, or continue within the same conversation thread instead of starting a new one.

Mistake 3: Not specifying format, then being annoyed by the format you got

If you don't say you want a table, a short paragraph is a completely reasonable default response — the model wasn't wrong, it just wasn't told.

Fix: State the format explicitly, every time it matters: "as a table," "in 3 bullet points," "under 50 words."

Mistake 4: Accepting a confident-sounding factual claim without checking it

Fluent, confident phrasing isn't evidence of accuracy — language models can state incorrect information with the same tone as correct information.

Fix: For anything factually important — a statistic, a citation, a specific claim — verify it independently before relying on it.

Mistake 5: Giving up after one disappointing response

A single mediocre response often just means one piece of the four-part structure (role, task, context, format) was missing, not that the tool can't help with the task at all.

Fix: Look at what's actually missing from your prompt, add it, and try again — see Why Vague Prompts Get Vague Answers for how to spot the gap.

Mistake 6: Over-trusting output on high-stakes topics

Treating AI output as a final answer on something medical, legal, or financial — where being wrong has real consequences — skips a verification step that matters.

Fix: Use AI output as a starting point or a second opinion on high-stakes topics, not a substitute for a qualified professional.

Key takeaway: most of these aren't really "mistakes with AI" — they're the same communication gaps that would confuse a person, just easier to overlook when you're typing quickly.