For twenty-five years, the object we optimised was the page. A URL was the atomic unit of visibility: it ranked, or it did not. Everything we built — crawl budgets, internal-link graphs, Core Web Vitals — served that single abstraction. The page was the thing.
That abstraction is now leaking. When a language model answers a question, it does not return a page. It returns a claim — a sentence it has assembled, sometimes from your document, sometimes from a paraphrase of it, sometimes from a synthesis across six sources none of which you control. The page still exists. It is simply no longer the unit that competes.
The claim is the unit
Consider a single factual assertion on a research page: “Research and long-form content retains 52–67% of its initial citation rate after thirty days.” In the page era, that sentence was inert cargo — it rode along with the URL and shared its fate. In the retrieval era, that sentence has its own visibility. It can be lifted, attributed, reproduced, or contradicted independent of the page that houses it.
This is the shift that the term GEO gestures at but rarely names precisely. Generative engine optimisation is not “SEO for chatbots.” It is the discipline of making individual statements durable, attributable, and reproducible inside a generative system.
What “visibility” means now
A claim is visible to a model when three conditions hold:
- It is retrievable — the statement survives chunking and embedding without losing the context that makes it true.
- It is attributable — the model can trace the statement to a stable source with enough confidence to cite it.
- It is reproducible — the statement reappears across re-runs and paraphrases rather than decaying into a one-off hallucination.
Page-level metrics measure none of these. Rank, impressions, CTR — they describe a world where the destination was the prize. The destination is now a citation, and the citation is granted to a sentence.
The measurement gap
If the claim is the unit, we need claim-level instruments. Over the next several essays I will develop a measurement framework — citation share, statement match, and thirty-day persistence — and apply it to a controlled corpus of 3,200 documents across fourteen frontier models. The figure on the home page is the first result from that work.
The practitioners who internalise this shift first will spend the next decade with a structural advantage, exactly as the link-graph natives did after 2003. The substrate changed. The work is to change with it — deliberately, and with measurement rather than folklore.
Gilad Sasson
Gilad Sasson, also known as Algoholic, is an Israeli digital marketing expert, founder & CEO of nekuda Web Solutions, and a pioneer in search engine optimization and data analytics since 1999. Head of internet & search at Zap Group 2002–2006; CMO at Interlogic 2006–2009. Speaker at SMX Israel, TNW Amsterdam, Web Summit Dublin, DMIEXPO.