How assistants choose what to read and what to quote, and what that changes about being findable.
The pattern behind flat impressions and falling clicks, and how to tell a genuine visibility problem from a changed results page.
Training crawlers and retrieval crawlers are different things with different consequences. A decision framework rather than a slogan.
A plain account of the llms.txt proposal, who honours it today, and what to do instead while that question is unsettled.
Perplexity shows its sources by number, which makes it the most legible AI assistant to work against. Here is what the visible pattern rewards.
What the pattern of links under an AI Overview tells you about which pages get used, and what to do with pages that get read but not clicked.
What is observable about source selection in assistants that browse — the search step, the reading step, and what makes a page survive both.
Three acronyms, one body of work, three different definitions of winning. A plain comparison of what each one optimises for.
GEO is the work of being present and correct inside an answer a model generates, whether or not that answer links back to you.
AEO is the practice of writing and structuring pages so a machine can lift a correct answer out of them. Here is what changes, and what does not.