Merchants are starting to score themselves on fact decay rate

A quieter measurement change is showing up in merchant analytics through late September 2026: teams that track AI citation coverage are adding a decay rate alongside their page counts.

Decay rate, in the way these teams are using it, is simple. For every page, count the extractable facts an assistant could quote — price, threshold, warranty length, lead time, compatibility — and record how many of them turned out to be false at the last verification. Divide by the number of facts and by the days since last verification. What you get is a speed, not a state: how fast this page stops being accurate.

That is a different question than freshness. A page edited last week can have a high decay rate if it holds fast-moving facts nobody re-checks. A page untouched for a year can have a decay rate near zero if everything on it is slow-moving. Freshness measures effort. Decay rate measures exposure.

Two things are driving the shift.

First, the cost of a stale fact became measurable. Once assistants began citing merchant pages directly in shopping answers, an outdated threshold stopped being a minor imprecision and became a wrong answer attributed to the merchant. Teams can now trace softened or dropped citations back to specific pages, which turns decay from a philosophical concern into a line item.

Second, page counts stopped predicting citations. Publishing more pages did not reliably produce more citations, because the constraint was never volume. The constraint was whether a given page held facts an assistant could state with confidence at the moment of the question.

The practical consequence is a reordering of the content calendar. In the workflow these teams describe, the calendar's main job is no longer to schedule new posts; it is to schedule verification of existing facts, with new posts filling whatever capacity remains. High-decay pages get verified monthly. Low-decay pages get swept quarterly. Nothing gets verified by editorial intuition.

There is a second-order effect worth noting. When decay is measured, the cheapest available improvement is usually not writing anything. It is reconciling two of your own pages that disagree, or moving a fact out of an image and into visible text. Both reduce exposure without adding a single new page, which is an unusual kind of win for a content team.

Visora's angle: the audit at [geovisora.com/audit](/audit) is built around extractable facts rather than page counts for exactly this reason — a page that an assistant cannot extract from has a decay rate that is, in practice, total.