Learn AI SEO

The umbrella term for optimising toward machine-generated answers. How it relates to generative and answer engine optimization, and how results are measured.

A recognised subclass of search engine optimization concerned with being quoted by language models rather than ranked by crawlers. Definition, scope and measurable outcomes.

Shared foundations, divergent objectives. Ranking competes for a click; citation competes for a sentence. Where the disciplines overlap and where they part.

Passage extraction, corroboration across domains, and the index prerequisite. What decides which pages appear beneath a generated summary.

Retrieval runs against an external index before the model writes. Absence from that index removes a site from consideration entirely, whatever its content.

A dedicated crawler plus third-party search, weighted toward pages carrying clean, attributable factual claims.

A conversational surface that decomposes one question into many hidden queries. What that fan-out means for a page built to answer only the original.

Answered questions produce no visit. Which enquiries still convert, and how to capture value when the click never arrives.

Publishers protecting inventory and businesses seeking recommendation need opposite directives. Which agents to permit, which to refuse, and the exact syntax.

Which types carry weight, the properties most often omitted, and why an identifier linking your record to a public database outperforms additional markup.

Machines resolve names to records. A business without a resolvable record is a string, and strings are not recommended.

Depth across neighbouring questions raises selection probability. Why answering one well matters less than answering its adjacent set.

A proposed plain-text file summarising a site for language models. Current adoption, which systems honour it, and an honest assessment of its present value.

The technique set in execution order, with the measurable effect of each and the three most commonly skipped steps.

A fixed prompt set, run on a schedule, logging which domains each assistant names. Method, cadence and the metrics that mislead.

What each platform actually measures, which assistants it covers, and which questions still require manual checking.

Observed market ranges by engagement type, the variables that move a quote, and the conditions under which the spend is not yet justified.

Twelve questions that separate a measurable engagement from a repackaged retainer, including the three answers that should end a sales conversation.