Getting cited inside ChatGPT and Perplexity answers is real and measurable. It is also smaller than the people selling it suggest.
Generative Engine Optimisation is the work of being named inside an answer produced by ChatGPT, Perplexity, Claude or Google's AI results, rather than ranking in a list of links below it. It is a genuine discipline, it is measurable, and we do it. It is also being sold at roughly ten times its current weight, which is worth saying plainly before anyone reallocates a budget.
The mechanism is different enough from search to need its own attention. A ranking is a position in an ordered list. A citation is a decision by a model that your page is the most quotable source for one specific claim. Quotability turns out to be a property you can design for: complete standalone sentences rather than ones that depend on the previous paragraph, facts in tables, explicit definitions near the top, and claims that are checkable against something external. Content written to be skimmed by a person and content written to be lifted by a model are not the same shape.
The unglamorous half is access. A great many sites still serve almost nothing to a crawler that does not run JavaScript, and most AI crawlers do not. GPTBot, PerplexityBot, ClaudeBot and Google-Extended all fetch raw HTML. If the page is an empty shell that fills in after hydration, the model sees an empty shell. We hit this on our own site, which is why every page here is prerendered to real HTML at build time.
What GEO is not, yet, is a traffic channel that replaces search. The volumes are much smaller, and a citation frequently produces no click at all, which makes it look like nothing in analytics while still doing work. The honest framing is that it is brand presence in a place where a growing share of research now happens, and that measuring it means tracking mentions rather than sessions.
The other honest note: it moves. Answer engines change how they select and attribute sources on no notice, and anyone claiming a durable method is describing this quarter. What has held so far is unremarkable, which is a reasonable sign it will keep holding: be genuinely the best source on a specific thing, say so in extractable language, and make sure a crawler can read the page without executing anything.
We run this on our own site as well as for clients. It is a reasonable test of whether it works, and it is why we can say what it does and does not do without borrowing somebody else's case study.