GEO vs SEO: what actually changes
The two disciplines share a foundation and diverge at the finish line. SEO competes for a position in a list of links, and the click is the outcome. GEO competes for inclusion in a generated answer, where there is no second place and often no click at all.
That difference reshapes what is worth optimising:
- The unit of success. SEO wins a ranked position. GEO wins a named mention, and a mention that omits you cannot be recovered by being slightly better on the next line.
- The shape of content. Ranking rewards depth and coverage. Citation rewards extractable claims: complete sentences that state a fact without depending on the paragraph around them.
- The role of consistency. SEO tolerates a brand described differently across the web. Retrieval systems do not, because they must resolve those descriptions to one entity before quoting anything.
- The measurement. SEO is measured in positions and clicks from a search console. GEO is measured by running questions through assistants and recording who gets named.
How we measure AI visibility
GEO fails as a service when it cannot be checked, so measurement is fixed before any content work starts. We agree a panel of questions your buyers genuinely ask, phrased the way they would phrase them, and that panel does not change between cycles.
The panel is run across ChatGPT, Perplexity and Gemini on a schedule. For each question we record three things: whether your brand is named, which specific page is cited if any, and which competitors appear in your place. Recording the competitor set matters more than it sounds, because it shows whether you are absent from a category or merely ranked behind a stronger source.
Comparison over time only works if the questions stay identical, which is why the panel is agreed up front rather than assembled at reporting time. Brand-name questions usually move first, since they are the easiest for a model to resolve. Category questions move last and are the real target.
What makes a page citable
Generative engines quote passages, not pages. A page earns citations by containing sentences a model can lift without risk, which is a narrower requirement than being well written.
Three formats do most of the work, and a strong page carries all three:
- A definition. One sentence in the pattern "X is a Y that does Z", placed near the top, before any qualification or marketing framing.
- A comparison. A structured contrast between two things the reader is choosing between, because comparison questions are among the most common prompts and the hardest to answer from a single vague page.
- A process. Numbered or sequential steps, which map cleanly onto how-to prompts and are quotable in full without losing meaning.
The prerequisite most GEO pitches skip
Retrieval systems behind AI assistants lean heavily on conventional search indexes. A page that cannot be crawled and indexed by a search engine is usually invisible to AI answers as well, which means GEO is not an alternative route for a site with a technical or discovery problem.
In practice this means the first GEO engagement often looks like technical work: crawler access, schema, entity consistency and the unglamorous question of whether anything on the open web links to you at all. A brand with no external references gives retrieval systems nothing to corroborate, and corroboration is what separates a claim a model will repeat from one it will not.
We would rather say this before an engagement than during one. If the foundation is missing, fixing it is the cheaper first move, and it makes everything downstream measurable.