From blue links to cited answers: how search optimization evolved into AEO and GEO

It is tempting to treat AI search as a clean break, a before-and-after line drawn at the launch of ChatGPT. It is not. Search optimization has been through four or five ruptures already, and each one created a new discipline with a new acronym, a new success metric, and a chorus of people declaring the old playbook dead. AEO and GEO are the latest, and understanding the lineage is the fastest way to see what genuinely changed and what is the same game with new surfaces.

The eras, at a glance

The classical era: SEO

SEO was born when search became the front door to the web. Google's insight was that the link graph encoded authority: a page linked to by many authoritative pages was probably authoritative itself. For a decade, optimization meant earning links, matching keywords, and keeping a site crawlable. The worldview was page-centric. Pages had authority, links passed it, domains accumulated it.

This era produced both the durable fundamentals (crawlable URLs, fast pages, semantic HTML, clear titles and headings) and the manipulable games (keyword stuffing, link farms, exact-match domains). Google spent the next fifteen years algorithmically killing the games while keeping the fundamentals. That pattern repeats in every era since: the structural fundamentals carry over, the manipulation tactics get devalued.

The paid layer: SEM

Paid search (SEM in its narrow modern sense) arrived alongside organic and never left. The auction for commercial queries turned the top of the SERP into real estate sold by the click. SEM and SEO have coexisted ever since as the two halves of search marketing: rent the top with ads, or earn it with content.

The reason SEM matters to this story is what is happening to it now. As AI Overviews answer commercial queries directly, the paid inventory beneath them gets fewer eyes, and the AI engines themselves are starting to introduce their own ad formats. The auction is migrating, not disappearing.

The social detour: SMO

Between roughly 2008 and 2015, a lot of people believed feeds would replace search. Discovery moved to Facebook, Twitter, and later Instagram and TikTok; social media optimization became its own discipline. It turned out to be a parallel channel rather than a replacement. People still went to search engines with intent (to evaluate, to buy, to learn how), and went to feeds to be entertained and to stumble onto things.

The lasting lesson from the social detour is about audience ownership. The brands that survived every subsequent platform shift were the ones that used social to build direct relationships (email lists, communities, owned audiences) rather than renting reach. That lesson is load-bearing again in the zero-click era, where search-engine intermediation is eroding traffic the same way.

The extraction turn: early AEO

Around 2014 Google began lifting passages out of pages and showing them as the answer. The featured snippet was the first time the click became optional: a user could read the answer without leaving the SERP. People Also Ask, knowledge panels, and voice-assistant readouts followed. This was the birth of Answer Engine Optimization, even though the term came later.

The optimization unit changed for the first time since SEO began. It was no longer the page and its rank; it was the passage and its extractability. A clean definitional paragraph under a question-shaped heading could win position zero even if the page ranked third. The fundamentals that mattered, structured data, semantic HTML, direct answers, were the same ones that would matter for AI search a decade later.

Zero-click and the answer SERP

By the early 2020s the cumulative effect of snippets, PAA, panels, and ads was a SERP that answered most informational queries within the page. Zero-click search crossed from a minority to a majority of queries. Publishers felt the squeeze first: the same rank that drove ten thousand clicks in 2018 drove a fraction of that by 2023, because the user got the answer without leaving Google.

This is the moment the business model of search-dependent publishing started to break, and it set up the central strategic question of the GEO era: if the click is no longer guaranteed, what is the durable visibility? The answer turned out to be the citation.

The generative rupture: GEO

Perplexity shipped the consumer-facing form of the answer engine in 2022: ask a question, get an assembled answer grounded in retrieved sources, with citations. ChatGPT Search, Claude with web search, Gemini, Google's SGE (rebranded to AI Overviews), and Bing Copilot followed. The mechanism underneath them all is retrieval-augmented generation: retrieve sources, ground the answer in them, cite them.

This is the deepest rupture since SEO itself, because it changes the optimization target twice over. The output is an answer, not a list, so ranking matters less than being one of the cited sources. And the answer blends multiple sources, so the unit becomes a page that contributes citable sentences to someone else's synthesis. Generative Engine Optimization is the discipline for that target, and citation share is its headline metric, the way rank was for SEO.

What carried over, and what broke

The through-line across thirty years is remarkably stable. Three things have survived every rupture:

• Entity authority. From the link graph to the knowledge graph to AI engine source selection, the engines have always tried to answer "is this a recognized, trustworthy source on this topic." Entity SEO and E-E-A-T are the current names for the oldest signal in search.
• Structural cleanliness. Crawlable, fast, semantically-marked-up pages have helped in every era. Schema.org is the modern expression; it mattered for rich results, then snippets, now grounding.
• Direct, specific, dated content. The page that answers the question clearly and early has won since 2014 and wins harder now.

What broke is the manipulation layer, again. Keyword density, link-volume games, thin doorway pages, and stuffed schema each had their moment and each got devalued. The pattern is reliable enough to use as a filter: if a tactic exploits a proxy rather than improving the actual content, assume the engines will eventually price it out.

Where this is heading: AIO and convergence

The current vocabulary debate (AIO, LLMO, and the rest) is mostly a sign that the disciplines are converging. AI Overviews already behave extractively for simple queries and generatively for complex ones. The practical bet for 2026-2028 is that AEO and GEO collapse into a single "AI search optimization" practice running against a shared content foundation, the same way SEO and SEM settled into two halves of one programme.

The teams that navigate this well are the ones who internalize the lineage. The fundamentals have not changed in thirty years; the surfaces have. Invest in the through-line (entity authority, structural cleanliness, specific dated content), track the surfaces that actually drive your business, and treat each new acronym as a new place to apply old discipline rather than a reason to start over. The AEO vs GEO bridge guide and the full disambiguation guide go deeper on where the current terms genuinely diverge.