Be Readable by the Assistants People Now Ask First

Answer engine optimization is structuring a page so an assistant can lift a direct answer out of it. In practice that means a short, self-contained answer near the top of the page, matching FAQ schema, and prose that states the subject rather than opening with "it" or "they". Every page in the BayouEdge catalog carries one.

Generative engine optimization is the wider job of being the source a model reaches for: publishing something that exists nowhere else, dating it, saying how it was measured, and making the entity behind it verifiable. Explanation alone is reproducible by anyone, so it rarely earns a citation on its own.

WebMCP lets a page register tools a browser-resident agent can call — checking a service area, pulling pricing, starting an enquiry — instead of the agent guessing from the layout. It is an unratified proposal with experimental browser support, so the tools are feature-detected and the page works normally without them.

Probably not many. BayouEdge measures this on its own site and publishes the figures: of 5,119 logged AI agent requests, only 22 were an assistant fetching a page to answer someone in the moment. The rest were search indexing and training harvest. Treat this as readiness work whose schema and answer structure help ordinary search today.

No, and nobody can. A server log shows that an assistant fetched a page; it cannot show whether the answer used it or named the source. Anyone selling a guaranteed AI citation is selling something they are unable to measure.

Requests from named AI agents are logged and split by why the agent came: harvesting for training, indexing for later, or fetching a page to answer a live question. The third is the one that matters, and it is reported as a fetch count rather than relabelled as citations.