What is the actual difference between SEO and AEO?

Deterministic against probabilistic

A ranking is a position in an ordered list. Run the query again and you get substantially the same list. That stability is what makes rank tracking a coherent product.

An answer is generated. Run the prompt again and the wording changes, the sources cited may change, and the ordering of what gets mentioned may change. Sampling variance is a property of the system, not a defect in the measurement.

This is why single-run answer-engine tests are unreliable, and why a vendor showing you a screenshot of one good result is showing you one draw from a distribution.

What that changes in practice

• Measurement. A position number is meaningless for a generated answer. What can be measured is frequency across repeated sampling, and the honest report is a distribution rather than a rank.
• Attribution. A click carries a referrer. A citation inside an answer often produces no visit at all, and the influence on a later direct visit cannot be traced. Attribution here is not a hard problem, it is currently an unsolved one.
• Unit of optimisation. SEO optimises pages. Retrieval operates on passages, so the unit becomes the extractable chunk, which changes how documents are structured.
• Corroboration. Ranking rewards authority on your own domain. Generation rewards agreement across sources, which puts weight on surfaces you do not own.

What does not change

Everything under the AEO layer. The page still has to be fetchable, render its content without JavaScript, resolve to a clear entity, and load fast enough to be worth fetching. AEO does not replace that work and cannot compensate for its absence.

This is the reason audits here run bottom up rather than starting at the layer the client asked about. The full sequence is set out in the method.