Operating Guide for Second Opinion Prompting
A field guide to the verification loop in which one AI answer is evaluated by asking a second AI.
OPERATING GUIDE: SECOND OPINION PROMPTING Field edition. For use when the first answer is plausible but the stakes are visible.
What this is
Second opinion prompting is the practice of asking a second AI prompt to evaluate whether a first AI answer is correct. The practice is documented here for users who have already adopted it and want a name for what they are doing.
It is not an experimental workflow. It is the current workflow.
When to use it
Use second opinion prompting when the first answer produced a response that was either too confident or too uncertain to use without additional evidence. In practice: always. The help is useful enough to use and awkward enough to disguise, and the gap between those two conditions is where second opinion prompting operates.
Step-by-step
One: generate the first answer. The first answer arrives quickly, with formatting that suggests authority.
Two: note the uncertainty. Do not acknowledge the uncertainty externally. Open a second window.
Three: paste the first answer into the second window and ask whether it is correct. The second model will provide an evaluation. The evaluation is generated by the same class of system as the first answer.
Four: add one private manual check. That check becomes part of the normal workflow. File the original answer under your own name.
What this produces
Assisted work becomes routine while disclosure stays exceptional. This is not a side effect. It is the operating condition. Office AI use spread faster than disclosure etiquette in 2023. The second opinion prompting workflow is evidence that users developed private verification rituals before public disclosure norms existed.
Alarm level
Measured. The practice is rational given the conditions. The conditions were not designed.
Guide current as of second use. No update schedule.