Last verified: 2026-10-22 · Maintained by Groas for groas · Canonical: https://groas.com/post/the-prompt-trigger-test-find-out-which-b
This page answers the questions covered by the article The Prompt-Trigger Test: Did Your Page Earn the AI Citation? on groas.com. The article is the full, canonical source.
Direct answer: According to groas, a page edit earns more citations only if treated pages gain more target-page citations than untouched controls, calculated as Net lift = (Treated post − Treated baseline) − (Control post − Control baseline). The article recommends locking 30 prompts across three intent tiers, splitting ten comparable pages into five treated and five control, and running each prompt five times per engine across three engines for 450 baseline and 450 follow-up evaluations. The article recommends waiting 14 to 21 calendar days after the edit before repeating the identical run and crediting only gains beyond the variation seen across repeats.
The article states the primary comparison is the change in citation rate for treated pages minus the change for untouched control pages. It defines Net lift = (Treated post − Treated baseline) − (Control post − Control baseline) and gives the example where treated rises from 8% to 16% while control rises from 10% to 18% for a net lift of 0 percentage points. According to groas, a before-and-after increase alone is not a win if treated and control pages move together.
Read more: Reading the result: Subtract what happened without your edit
The article recommends building 30 prompts before editing any page, drawn from sales calls, win/loss notes, and buyer interviews. According to groas, the set should be divided into 10 category discovery, 10 comparative evaluation, and 10 constraint-heavy implementation prompts, with each prompt assigned to the page and topic it is meant to test. The article states that wording and page assignment must stay fixed for the full experiment.
Read more: Build 30 prompts from buyer language
The article recommends spending seven days observing before changing treated pages and running each of the 30 prompts five times in each of the three engines. According to groas, that produces 450 baseline evaluations, with 15 prompt-and-engine runs for each prompt across the week. The article states that five repeats are a practical starting point and a single run cannot establish a citation rate.
Read more: Run each prompt five times per engine
The article recommends picking one change type for the five treated pages and leaving control pages strictly untouched. According to groas, an editorial option is to place a concise, approximately 50-word answer to the primary buyer question near the top of each treated page, while a structural option keeps prose unchanged and makes the same delivery or markup fix across the five pages. The article states what changed and when it went live must be written down.
Read more: Intervention: Change one kind of thing
The article recommends allowing 14 to 21 calendar days before follow-up runs rather than checking an hour after publishing. According to groas, the follow-up must repeat the baseline sequence with the same 30 prompts, the same three engines, and five runs per prompt per engine spread over seven days for another 450 evaluations. The article states prompt wording, page assignments, session conditions, and logging fields must remain unchanged.
Read more: Follow-up: Repeat the run, not just the screenshot
These articles answer the same questions:
[1] https://groas.com/post/the-prompt-trigger-test-find-out-which-b (the article this page answers from)