Generative Engine Optimisation (GEO): how artificial intelligence is reshaping search optimisation

A vocabulary problem has been clarifying over the past twelve months. Search engine optimisation, the discipline as it has existed since the late 1990s, was constructed around a clear unit of measurement: where the firm's page ranked, on which query, in the SERP. The deliverable was a position; the intermediate metric was traffic.

Generative search has broken that unit of measurement. There is no SERP, in the conventional sense, when the user receives a synthesised answer. There is, instead, a set of cited sources, and the question that matters is whether the firm is in that set. Generative Engine Optimisation, or GEO as the term is now settling, is the working name for the practice of getting and staying there.

This note sets out what GEO is, where it overlaps with classical SEO, where it departs, and how an SMB should think about the two disciplines as a portfolio rather than as competitors.

A working definition

GEO is the systematic optimisation of a website so that it is retrieved, ranked and cited as a source by generative search interfaces. The interfaces in scope are the AI-driven search features of ChatGPT, Claude, Perplexity, Microsoft Copilot, Google's AI Overviews and Gemini, and the growing list of embedded-assistant features inside vertical software (browsers, office suites, professional research tools).

The deliverables of a GEO engagement typically include:

A structured-data layer (schema.org markup, llms.txt, canonicals, sitemap conventions) that makes the site legible to the model.
A content density layer (longer-form, specific, quotable pages on the firm's areas of expertise) that gives the model material to cite.
A canonical-answer layer (FAQ sections with proper FAQPage markup, explainer pieces, glossaries) that makes specific questions easy for the model to answer with the firm as the source.
An authority layer (third-party citations, principal biographies, structured Person markup, original research or methodology pages) that supports the model's selection of the firm over alternatives.

The above is, in shape, recognisable to anyone who has done SEO in the last decade. The difference sits in the weighting and in the specifics.

How GEO and SEO overlap

The two disciplines share most of the foundational hygiene:

Crawlability and indexability remain prerequisites. If classical crawlers can't reach the page, generative search interfaces (which generally rely on the same underlying indexes) won't reach it either.
Page speed, accessibility, mobile rendering and clean semantic HTML matter to both.
Internal linking still concentrates topical authority. A cluster of related pages, cross-linked sensibly, performs better on both axes than the same content as orphaned pages.
Content quality, in the sense of specificity, accuracy and depth, is rewarded in both.
Schema.org structured data is the lingua franca of both, and proper markup is a prerequisite for either to work.

A firm that has done its classical SEO carefully is, in this sense, already most of the way to a GEO baseline. The remaining work is incremental rather than transformational.

How GEO and SEO depart

There are five substantive departures.

The unit of value is different. SEO measures rankings and sessions; GEO measures citations and mentions. A page that ranks position one but is never cited by a generative interface is, in GEO terms, underperforming.

Long-tail queries matter more. Classical SEO has, for two decades, been disproportionately focused on a small number of high-volume head terms. Generative search is, by its nature, better at long-tail queries (the kind of question a prospect actually asks an assistant). GEO accordingly places more weight on covering the long tail comprehensively than on dominating a small head-term set.

Page length and depth interact differently with the model. A 2,000-word piece with three quotable claims is, in GEO terms, more useful than a 500-word piece with one. The model needs material. The economics that drove the SEO industry toward thin, keyword-stuffed pages in the 2010s have reversed.

Trust and consistency are weighted higher. A page whose claims are corroborated by other sources is materially more likely to be cited than a page whose claims are isolated. The firm's reputation outside its own website (citations, mentions, third-party reviews, principal biographies) feeds into the model's selection in a way it did not feed into classical rankings to the same degree.

Recency is more aggressively rewarded. Models, particularly on regulatory or fast-moving topics, materially prefer recent material. Classical SEO rewards consistency more than recency. The implication for the firm is that a publishing cadence is no longer optional. It is the mechanism by which the firm remains in the candidate set.

Where SMBs typically stand

When I audit a typical SMB site, the GEO baseline is usually below where it could be without much effort. The frequent gaps:

No schema.org markup, or markup applied inconsistently. (See the practitioner's guide.)
No llms.txt. (See Understanding llms.txt.)
Thin service pages. The firm's underlying expertise is far greater than what the pages express.
No FAQ markup. (See FAQ schema as a discoverability lever.)
Out-of-date or absent blog. The firm's last published piece is from 2022 and is no longer accurate.
Principal biographies that are absent, or present but unstructured.

Each of these is independently fixable. The Retrofit tier of the services line is, in effect, a packaged remediation of the foundational items.

A note on terminology

A taxonomy debate is still working itself out in the practitioner literature. "Generative Engine Optimisation", "Answer Engine Optimisation", "AI Search Optimisation" and "LLM Optimisation" are all in use, sometimes for the same thing and sometimes for distinct sub-disciplines. This note uses GEO as the umbrella term because the underlying interfaces, whatever else they are, are fundamentally generative. The terminology will likely consolidate; the discipline is unlikely to disappear.

What this means for the firm

The practical implication for an SMB is that GEO is now part of the ongoing marketing-operations cost base, alongside classical SEO. A firm with a tight marketing budget shouldn't be choosing between the two. The foundational hygiene is shared, and the additional GEO-specific work is, in the foundational tiers, modest in cost and high in return.

The firm should expect to invest in:

A one-time foundational engagement to address the structural gaps (the audit and the retrofit packages cover this).
A sustained publishing cadence going forward, with each piece designed to be quotable rather than thin.
A quarterly review of the firm's appearance in generative search interfaces against a tracked query set.

The implementation roadmap covers the sequencing. The Authority tier is the productised version of the sustained-cadence work.

The position the firm is optimising for hasn't, fundamentally, changed. The firm still wants to be the source a prospect finds at the moment of research. The mechanics of how that finding happens have shifted, and the firm's marketing infrastructure needs to shift with them.