From ranked links to generated answers
Large language models moved search from "here are ten pages that might have your answer" to "here is your answer, assembled from several sources." Google's AI Overviews — tested publicly as the Search Generative Experience (SGE) before its wider rollout — now surface AI-written summaries directly above traditional organic results for a large share of queries. Perplexity and ChatGPT Search were built around conversational, cited answers from the outset rather than adding them on top of an existing results page, and Microsoft integrated its Copilot assistant directly into Bing.
The common thread across all of these systems is retrieval-augmented generation: rather than relying purely on what a language model memorized during training, the system retrieves current web content in response to a query and uses it to ground a generated answer, typically with citations back to source pages.
New disciplines: AEO and GEO
This shift has produced new terminology and, increasingly, new specialties within digital marketing. AEO (Answer Engine Optimization) and GEO (Generative Engine Optimization) both describe the practice of structuring content so that it is accurately retrieved, cited, and summarized by AI systems — a related but distinct goal from traditional SEO's focus on ranking position. Clear structure, direct and quotable answers to specific questions, credible sourcing, and well-marked expertise all appear to matter more in this context than they did for traditional ranking alone, though the field is new enough that best practices are still being established through observation rather than settled research.
The publisher traffic question
AI-generated answers have raised a real and unresolved tension for the sites whose content trains and grounds them: a zero-click search, where a query is fully answered on the results page itself, generates no visit to the source site even when that site is cited. Publishers, platforms, and search engines are still actively negotiating what fair attribution, licensing, and traffic-sharing should look like in this model, and the economics are far from settled.
What's next
The consistent pattern across more than three decades of search history is that each generation of technology has been disrupted by a better way of assessing relevance and intent: from filenames, to keywords, to links, to language understanding, to generative synthesis. Whether AI-generated answers represent a durable end state, or simply the next stage in that progression, is at this point an open question rather than a settled one — and it is very likely that whatever comes after generative search will follow the same pattern of closing a gap that today's systems still leave open.