AI Visibility Forecasting
What each confidence label means
A forecast without a confidence level cannot be wrong in any useful way. Four labels, what each one requires, and what it permits you to do.
Why the label is the product
A directional statement with no confidence attached is unfalsifiable. If conditions move as described it was right, and if they do not it was directional. That is comfortable for the forecaster and worthless to a reader.
Attaching a level does three things: it tells a business how much to weight the forecast in a decision, it creates a record that can be scored later, and it forces the forecaster to separate what was observed from what was inferred.
The four levels
High. A trend observed across multiple engines, over multiple periods, with a mechanism that explains it. Rare. Permits acting before the change completes.
Moderate. A consistent trend on one engine or a shorter window, or a clear trend with no established mechanism. Permits preparing, not committing.
Low. A pattern that may be real, insufficient duration or sample to distinguish it from noise. Permits watching. Published because a low-confidence reading published honestly is a success.
Speculative. Reasoning from a mechanism rather than from an observation. Labelled clearly and never used to justify spending.
What raises and lowers confidence
It rises with duration, with agreement across engines, with a mechanism that predicts the direction independently, and with the size of the effect relative to normal variation.
It falls when a trend appears on one engine only, when it coincides with a known product change that could explain it trivially, and when the sample is small enough that reversal would not be surprising.
The rule that matters most
A low-confidence reading is never upgraded because it would be more useful at a higher level. That single pressure is where forecasting practices go wrong, and it goes wrong quietly, because nobody scores the forecast afterwards.