Three acronyms are circulating for what looks like one job, and they get used as synonyms in pitch decks. They are not synonyms. They measure different things, at different layers, and only one of them is under your direct control.
SEO measures a position
Search engine optimization has a number at the end of it. For a given keyword, in a given country, on a given day, a page sits at position 4. The number moves, you can chart it, and everyone agrees what it means. The whole reporting practice of the last twenty years is built on the fact that a ranked list has positions in it.
GEO measures a presence
Generative engine optimization has no position, because an AI answer is not a list. Ask ChatGPT, Gemini, Perplexity or Claude for the best accountant in Vienna and you get a paragraph naming three or four firms. You are in it, or you are not. Ask the same question in slightly different words and the answer changes.
That makes GEO a fundamentally different measurement problem. There is no single ranking to check — there is a distribution of answers across engines and phrasings, and the useful metric is how consistently you appear across it. One answer proves nothing. A hundred answers over ten weeks is a trend.
Two things worth measuring beyond bare presence: where in the answer you appear (first name mentioned carries more weight than a footnote), and how you are described. A model that names you and then steers the reader elsewhere has mentioned you, and that mention is not a win.
AEO measures a page
Answer engine optimization is the one that is actually yours. It asks whether a page can be extracted from: does it have a clear heading structure, does it answer a question in a self-contained paragraph near the top, does it carry structured data, does it state facts in sentences rather than burying them in a carousel that renders after JavaScript.
This is the part you can verify directly, because it is a property of your own HTML rather than of someone else’s model. It is also the part that people skip, because it looks like the SEO work they already did.
Why the distinction matters commercially
If you sell all three as one thing, you end up promising an outcome you cannot control. Nobody can guarantee a language model will recommend a client — the model is not an ad account and there is no bidding surface. What you can promise is honest measurement of GEO, real improvement on AEO, and the SEO work you were already doing.
The dishonest version of this market sells “get recommended by ChatGPT” as a deliverable. The defensible version sells the measurement, the trend, and the specific page-level work that plausibly moves it — while being explicit that the last step belongs to a model nobody owns.
What to put in a client report
- SEO: position per tracked keyword, and what moved.
- GEO: mention rate per engine across a fixed prompt set, checked on a fixed cadence, with the answers archived so a claim can be re-read later.
- AEO: the page-level findings — missing headings, thin or absent meta descriptions, invalid structured data, pages that return an error but still look fine to a scanner.
Three sections, three different kinds of claim. Mixing them into one “AI visibility score” without saying what went into it is how you end up unable to answer the only question that matters when the number drops: which part of it moved?