Every few years, a technological shift forces the SEO profession to reconsider its fundamentals. The move from desktop to mobile required new strategies. The advent of voice search changed how practitioners thought about query phrasing. Now, the rise of large language models represents the most significant rethink since the Penguin update. Understanding exactly what changes — and what doesn't — is essential for building a strategy that works in 2026 and beyond.

The old model: optimising for crawlers and 10 blue links

Traditional SEO is built on a remarkably stable foundation. A search engine crawler — Googlebot, BingBot — visits web pages, follows links, and indexes content. An algorithm then ranks that content based on hundreds of signals: keyword relevance, page authority (derived from the quantity and quality of inbound links), technical performance, user experience signals, and more.

The user's interface with this system is the SERP (Search Engine Results Page): a ranked list of ten blue links, potentially enriched with featured snippets, knowledge panels, and People Also Ask boxes. The goal of traditional SEO is to appear as high as possible on that list for queries that matter to your business, and to earn the clicks that flow from high rankings.

Every metric in traditional SEO is click-centric: impressions, click-through rate, organic sessions, conversion rate from organic traffic. The entire system assumes that content is discovered through the SERP and consumed on the publisher's website.

The new model: optimising for AI citation and recommendation

LLM SEO — also called Generative Engine Optimization (GEO) — operates on entirely different assumptions. The user doesn't see a ranked list; they receive a synthesised answer. There is no SERP to rank on, no position one to achieve. The question isn't "do I rank?" but "am I cited?" and "is my brand mentioned in the answer?"

AI assistants synthesise responses from a combination of training data and, increasingly, real-time retrieval. Your brand's presence in that synthesis depends on how strongly and positively you are represented across the sources that fed the model's training, and across the sources the model retrieves at query time. There is no algorithm with published ranking factors — but there are patterns in what gets cited and what doesn't. To understand these patterns more deeply, read our introduction to GEO.

What stays the same: authority, relevance, and trust

Despite the profound differences in mechanism, the strategic north star of both traditional SEO and LLM SEO is identical: be the best answer to the question being asked. In traditional SEO, "best answer" is judged by Google's algorithm. In LLM SEO, "best answer" is judged by what an AI model has learned to associate with credibility, accuracy, and relevance.

The following signals retain importance in both worlds:

  • Content quality: Accurate, well-researched, clearly written content performs better in both traditional and AI search. Google's quality raters and AI training pipelines both reward depth and precision over shallow volume.
  • Authority: High-domain-authority backlinks in traditional SEO translate roughly to high-authority third-party citations in LLM SEO. The mechanism differs but the underlying concept — credibility conferred by trusted sources — is identical.
  • Trustworthiness: E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness), Google's quality framework, is also a useful lens for LLM SEO. For a full treatment, see our article on E-E-A-T and GEO.

"The core premise of SEO — be the best answer to a question — hasn't changed. What's changed is who's doing the answering."

What changes: from keywords to entities

The biggest tactical shift between traditional SEO and LLM SEO is the move from keyword optimisation to entity optimisation. Traditional SEO practitioners think in terms of keyword clusters: which terms do users search for, what is the search volume, how competitive is the landscape, where can I rank? The content strategy follows from this keyword map.

LLM SEO practitioners need to think in terms of entities — named concepts with well-defined attributes. Your brand is an entity. The category you compete in is an entity. The problems you solve are entities. AI models reason about entities and their relationships, not about keyword co-occurrence patterns. If your brand's entity is clearly defined — consistent name, clear category association, verifiable attributes — it will be represented accurately and cited reliably. If your entity is fuzzy or inconsistent, even high-volume content won't reliably surface it in AI responses.

In traditional SEO, the backlink is the foundational authority signal. A link from a high-DA site to your content is a vote of confidence that the algorithm counts and weights. The entire link-building industry exists to earn and manage these votes.

In LLM SEO, the analogous signal is the brand mention — a reference to your brand in authoritative third-party content, whether or not that content includes a hyperlink. When TechCrunch writes about your product, when an industry analyst includes you in a comparison report, when a Wikipedia editor cites you as an example — all of these create the kind of third-party signal that influences how AI models perceive your brand's authority.

This doesn't mean backlinks are irrelevant. Many AI assistants use retrieval-augmented generation, which means they fetch web content at query time. A backlink-rich page that ranks well in traditional search is also more likely to be retrieved by an AI. But the conceptual frame shifts from "link equity" to "brand authority."

Measuring success: rankings vs share of voice in AI

Traditional SEO success is measured with well-established metrics: keyword rankings, organic impressions, organic sessions, conversion rate from organic. These metrics are imperfect but consistently measurable.

LLM SEO requires a different measurement framework centred on AI share of voice: what percentage of AI responses to category-relevant queries mention your brand? How does this compare to key competitors? Is the sentiment positive? Is the description accurate? Our article on tracking AI share of voice covers the full methodology.

Building a strategy that covers both worlds

The practical implication is that brands need a strategy that serves both the traditional search world and the AI answer world — and fortunately, these are more complementary than they are in conflict. The tactics that improve traditional SEO authority (earning coverage in high-authority publications, building a clear topical content cluster, improving technical signals) also improve LLM SEO. The difference is in the additional layer of entity management, structured data, and direct AI visibility monitoring that GEO requires.

The brands that will win the next decade of search are those that treat GEO not as a replacement for SEO but as its natural extension — the next chapter in the same continuous effort to be the most visible, credible, and cited authority in their category. Start measuring your AI visibility with Sight →