Retrieval and generation are different jobs
For twenty years search worked by retrieval. You asked, the engine fetched a ranked list of documents, and you chose. The entire discipline of SEO grew around one question: how do we get our document higher in that list so it gets clicked?
Generation works differently. The engine reads widely, then writes an answer. Your business either appears inside that written answer or it does not, and whether anyone clicks through is a separate question. Optimising for that is a different job, because the unit of success stopped being a position and became a mention.
This is the part worth being precise about. It is not that ranking stopped mattering. It is that ranking stopped being sufficient.
Side by side
| SEO | GEO | |
|---|---|---|
| Mechanism | Retrieval: your page is fetched into a ranked list | Generation: your business is written into an answer |
| Unit of success | Position, then a click | A mention, accurate and in context |
| What you optimise | Relevance and authority for a query | Quotability, entity clarity and corroboration |
| Typical measurement | Rankings, sessions, click-through rate | Named-in-answer rate across a defined prompt set |
| Where the traffic goes | To your page, measurably | Sometimes to your page, sometimes nowhere at all |
| What breaks it | Losing position to a better-optimised competitor | Being unreachable, inconsistent, or uncorroborated |
What did not change
Almost all of the groundwork, which is why the SEO-is-dead genre is mostly selling something. GEO has nothing to work with if the foundations are missing.
- Crawlability still decides everything. An engine cannot represent a business whose site it cannot reach or parse. This remains the most common blocker we find.
- Authority still counts. Models weigh corroboration across sources, which is a close cousin of what links measured.
- Content quality still counts, differently. The bar moved from comprehensive to extractable. A clean, direct passage beats a long article that has to be summarised.
- Traditional results still drive traffic. Fewer clicks is not no clicks, and abandoning search because AI exists would be an expensive overcorrection.
The uncomfortable number
Both camps quote this one selectively, so here it is with its limits attached. In the United States, Pew Research Center tracked 68,879 Google searches by 900 adults and found people clicked a result in 8% of visits where an AI summary appeared, against 15% where none did. Only 1% clicked a link inside the summary itself.
Two honest caveats. That is United States data, and no equivalent Australian study has been published, so anyone quoting an Australian click-through decline is extrapolating. And it cuts both ways: it gets quoted as an argument for GEO, but it is plainly also a problem for organic traffic, and we are not going to pretend otherwise.
The counterweight is that the visits which do arrive appear to be worth more. Semrush found the average AI-search visitor was 4.4 times as valuable as the average organic visit on conversion rate, published 21 July 2025 across more than 500 topics. Directional rather than a promise. Both figures, with their sources and sample sizes, are on our AI search statistics page.
What this means for your reporting
This is where the change actually bites. A rank tracker cannot see a mention, so a business can be doing well in AI answers while its monthly report shows nothing, or losing ground there while the report looks fine.
The practical fix is to add one instrument rather than replace everything: a fixed set of the questions your buyers actually ask, run against several engines on a schedule, recording whether you are named and whether the description is right. Keep the search reporting as well. You want both, because they fail independently.
For the neighbouring comparisons see AEO versus GEO and SEO versus AEO.