One loop, six stages, on a schedule.

The loop starts with instruments, not opinions. Every week it reads your whole site the way a search engine does: titles, headings, page structure, internal links. Every day it pulls your real search and traffic numbers, including a dedicated count of visitors who arrive from AI answers. It also keeps an eye on speed, uptime, and whether AI assistants can actually reach your pages.

Raw signals become findings: a specific problem or opportunity, with a severity, an estimated value, and the evidence behind it. And before anything gets blamed on your site, the system checks whether every site it watches dipped at the same time. If they did, that's Google changing the rules, not something to rewrite your pages over.

This is the only stage where AI writes anything, and what it writes is a proposal: change this title, add this answer, retire that page, each marked with a risk level. Facts come from your own pages or your own data, never from thin air.

Every proposed change lands in your review queue in plain terms: what changes, why, and what the risk is. You approve, edit, or reject. Routine low-risk fixes can go on autopilot once their track record on your own site has earned it. Bigger changes always wait for you.

Approved changes are published through the same system you already use to publish, never behind its back. WordPress, Webflow, Shopify, Wix, Framer, Ghost, Notion, or a developer-managed site: WhoTheyFind connects to what you have and uses its normal publishing flow, history included.

Nothing is declared fixed on the day it changed. The loop waits days to weeks, compares before and after, and records one of three outcomes: worked, no effect, or made things worse. Anything in the last group gets undone automatically. Being able to undo is a requirement for every platform WhoTheyFind publishes to.