GEO.

GEO is the work of increasing the likelihood that a site is chosen as a source in the answers generated by AI systems.

The term came into circulation as search engines shifted from offering a list of links to producing synthesised answers. Classic optimisation aimed at ranking; GEO aims at being mentioned inside the generated answer. Both are fed by the same index, but their selection logic differs.

The most important framing is this: GEO is not an alternative to SEO but a layer on top of it. Google’s own guidance points the same way — there is no separate technique, special markup or registration method for appearing in generative features. A page that is not indexed cannot appear in any answer engine.

At the centre of the difference is a change of unit. In classic ranking what competes is a page and its whole is assessed; in an answer engine what competes is a PASSAGE. When the model produces an answer it takes not the whole page but the section that answers the question on its own.

1. Access — can the crawler get in? robots.txt, server response, JavaScript dependency.
2. Readability — heading structure and passage boundaries.
3. Intelligibility — structured data and entity links.
4. Trustworthiness — author, date, sources, verifiability.
5. External corroboration — being mentioned in other sources.

The fifth item lives entirely off-site and is the most neglected: models prefer information that several independent sources agree on. On-site perfection has a ceiling, and rising above it requires existing off-site as well.

The term itself may be temporary, and probably will be. Google’s wording points that way too: there is no separate discipline for generative features, SEO itself is widening. Within a few years GEO will most likely stop being a separate heading and dissolve into SEO — much as mobile optimisation, once counted a separate speciality, is an ordinary requirement today.

What lasts is not the name but the work: a site that is crawlable, readable, intelligible and trustworthy. These were true twenty years ago too; the only thing that changed is the cost and the return.

1. Access — can AI crawlers get into the site? robots.txt permission, server response, content delivered from the server.
2. Readability — does the heading hierarchy draw the passage boundaries? That is how a model knows where something ends.
3. Intelligibility — are structured data and entity links in place?
4. Trustworthiness — who is the author, what is the date, what do the claims rest on?
5. External corroboration — are you mentioned in other sources?

These five items have to be taken in order, because each rests on the one before. Writing schema for a page that is not indexed is like arranging the window of a shop that is closed.