What the AI Update email is
Your AI Update is the recurring email that reports how 9 AI systems covering 90%+ of AI answers represent your business, sent on your plan schedule (monthly on Standard, weekly on Pro, daily on Custom). It is built to be honest about what the numbers can and cannot say, so a quiet week reads as quiet and a real change reads as real.
The headline number: a 7-day window with a plus/minus band
The accuracy figure at the top is usually a 7-day rolling window — the share of answerable factual questions the AI engines got right about you over the last seven days — shown as "N% ±E". The ±E is a 95% confidence interval: the margin of error given how many probes ran, so a smaller sample gives a wider band. A reading of "82% ±4" means the true value is very likely between 78 and 86 percent.
Why it matters: a move that stays inside the band is measurement noise, not a real change. AITWIRE calls a change real only when it clears the combined band of the two windows being compared.
A brand-new or very sparse account may still see a 30-day figure until there is enough 7-day history. The label always names the window.
"Latest cycle incomplete — not scored"
If the most recent cycle collected far fewer probes than normal (for example, a pipeline hiccup), AITWIRE does not report that partial number. Instead it shows a note like "Latest cycle incomplete (n=12 vs typical ~40) — not scored" and headlines the last full reading instead. This is deliberate: a half-collected day can look like a crash when nothing actually changed.
Suppressed change figures
The change-versus-last-period figure compares two 7-day windows. When either window is incomplete, the change is suppressed — shown as an em-dash labelled "delta suppressed (partial data)" — instead of a possibly misleading number. A glitch day can never render as a real move.
Two different accuracy rulers
The email carries two accuracy numbers that measure different things. Never compare them directly:
- Answerable-factual accuracy — of the factual questions AI engines can reasonably answer about you, the share answered correctly. The headline is the 7-day window; the stat strip carries the 30-day window. Same ruler, two windows.
- Cycle accuracy — the blended per-cycle score, which also includes category and visibility probes. A different ruler, and not comparable to answerable-factual accuracy.
Each is labelled where it appears so the two are never confused.
What to do with it
Read the AI Update for the trend, not the daily wiggle. When accuracy moves outside the band — up after a fix you published, or down as a drift alert — that is the signal worth acting on. Judge published fixes over weeks, not hours (see How the AI Update Cycle Works).