AI Visibility Forecasting
The method, and its limits
What we forecast, what we refuse to, how confidence is assigned, and how a forecast gets scored afterwards. Published so the record can be checked rather than trusted.
How a forecast is made
What each confidence label means
Four levels, what each requires, and the pressure that corrupts them.
A signal is not an event
Most coverage reports events. A forecast is a claim about what an event implies.
Scoring a forecast after the fact
The only mechanism by which forecasting improves, and the part nobody publishes.
The conditions we track
Citation concentration, and why it matters more than volume
Whether the set of sources is widening or narrowing changes what work is worth doing.
Source class weight: what is gaining and what is losing
Classes are forecastable where individual sites are not.
Volatility as a condition worth tracking
How much answers move sets the noise floor for every other reading.
Leading and lagging indicators
Almost everything businesses monitor arrives last.
The limits
Why we do not forecast your ranking
The thing everyone wants is the thing least possible to forecast.
How to use a forecast in an actual decision
Four legitimate uses and three reliable misuses.