Reading a forecast

One platform moving is not the market moving

Most published AI visibility commentary describes one assistant and is read as describing all of them. They do not behave alike, they do not change together, and averaging them produces a number that is true of nowhere.

Why unlike things are never combined here

An assistant that answers from a live index and one that answers largely from training data are not two samples of the same population. They cite differently, they update on different clocks, and a change in one carries almost no information about the other.

Averaged into a single score, a sharp fall on one surface and a rise on another cancel to nothing, and the report shows a calm week during a week in which something significant happened to somebody. AI Visibility Forecasting keeps them separate for that reason, and a single headline index is the one thing it will not produce.

Three questions that decide whether a platform change is yours

Does it send you anyone? A material change on a surface that has never referred a visitor is a fact about the market, not an instruction to you. Interesting; not urgent.

Does it answer your kind of question? Assistants specialise in practice. One that is mostly used for code and one mostly used for local services will not both matter to a mobility retailer, and treating a change in the first as a threat wastes attention that the second deserves.

Is the change to retrieval or to phrasing? A surface that starts retrieving different sources is a visibility change. One that words the same answer differently usually is not, though it reads dramatically in a screenshot.

The correlated case, which is the one worth fearing

Occasionally several surfaces move the same way in the same week, and that is a different event entirely: usually a shared upstream source, a common index, or a change in what the underlying web is publishing. Those are rare, they are the ones that justify reorganising a quarter, and they are only visible if the platforms were kept apart in the first place.

A combined index cannot tell a correlated move from a coincidence, because it has already added them together before anyone looks.

What we hold, and what we do not

Each covered surface is reported on its own, with its own conditions and its own record, and cross-platform movements are called out explicitly when they occur rather than inferred from a total.

What we do not do is publish a single AI visibility score, weight platforms by a market share nobody can verify, or imply that a change on one is a change everywhere. Where a movement is confined to one surface, the report says which one, and that is usually the most actionable sentence in it.

When to ignore a forecast covers what to do with a change that is not yours, and both sit under AI Visibility Forecasting.