What the ± margin means

Ask an AI engine the same question twice and you can get two different answers. So each prompt is asked three times per engine, per run by default, and every score is an average over those answers.

• The ± is one standard error: roughly how far the score would move if the same prompts were asked again.
• 4 ± 4 means the score is likely somewhere around 0 to 8. Too few mentions to tell apart from none.
• 62 ± 3 is a tight reading: the engine answered consistently.
• More prompts mean more answers, and more answers mean a narrower margin.

The shaded band around each line on the trend chart is this same margin.

• A change between two runs counts as real only when it's more than twice their combined margin. Smaller moves are within normal variation.
• Change alerts use this same rule, so they don't fire on noise.
• An engine with a single answer has no margin to measure, so a change there is never flagged as real.