How it works
Four things decide whether a CBD account grows or gets restricted. This page shows how each one is decided here, and lets you test the strictest of them on your own sentence.
Topics from demand, not from a model
Product types from your catalogue are turned into keyword sets per country and language. Volumes, trends and the questions people ask become the themes of the month. A niche is cached for 30 days, because every one of those lookups is paid.
The account stage is decided by rules
Growth, trust or conversion is picked from numbers: followers, engagement, posting regularity, whether lab reports exist. A model only explains the verdict in plain language. If the stage were chosen by a model, the same account would get a different plan on a different day.
A library that gets better, not longer
26 structures, 58 hooks and 21 events, each tied to a goal, a funnel stage and a niche. Every hook was run through the compliance engine before it entered the library, and rejected if it did not pass.
The whole catalogue gets sold, not one hero product
Positions are spread across product types by priority, so the slow movers get their turn instead of the same bestseller every week.
The compliance engine
Seven layers run in order: banned claims, medical language, required disclaimers, jurisdiction rules, sales pressure, platform policy and brand stop topics. A finding is either a fix, which rewrites the line, or a fail, which stops the item before a single cent is spent on rendering it.
Test a sentence
QA looks at the render, not the prompt
The finished image goes to a second model, which scores readability, brand fit and whether the text on the image matches the caption. Below the threshold the item is fixed and re-rendered, at most twice, and then handed to a human rather than published.