On May 15 of this year, Google Search Central published guidance saying that answer engine optimization and generative engine optimization are not separate disciplines from SEO, and that they do not require llms.txt, special schema, or AI-specific content rewriting. Around the same time, a startup called Lorelight shut down after about seven months selling AI brand monitoring. Its founder wrote that the product had failed because the data did not change what customers did, since the same fundamentals that drive search visibility also drive AI visibility.
I think both of them are right, and I sell this service.
So let me make the concession completely, and then explain why I still think the question misses what changed.
The concession
Roughly 80% of AI search optimization is good fundamental SEO. Crawlable pages, clean structure, real authority, content that answers a question directly instead of circling it. We said this in our own guide to improving AI visibility three days ago, and we said it first in that piece rather than burying it, because the category is full of people who will not say it at all.
We have also lived it. The most expensive problem we found on our own site was a robots.txt misconfiguration, which is a 2004 problem. The second was that we could not tell whether Bing was crawling us, which is also a 2004 problem. Anyone selling you a new discipline is selling you the 20%.
Vendors resist this concession because it shrinks the addressable market. Google has its own reasons for making it, since a world where AI visibility is purchasable is a world where its results look bought. Both incentives are worth naming before anyone takes either statement as neutral. But the concession happens to be true, and arguing with it costs more credibility than it buys.
SEO was "just" marketing too
Here is the reframe I have landed on, and it is an opinion rather than a finding.
In 1998, optimizing for search engines was just marketing. It used the same fundamentals that had always applied: be findable, be credible, say clearly what you do, get other people to vouch for you. Nobody who understood marketing needed a new theory of persuasion to do SEO. The people who insisted it was a separate discipline were mostly selling something.
And yet the companies that treated it as marketing-as-usual, and waited to be convinced, lost roughly a decade of distribution to the ones that did not.
The discipline was not new. The distribution layer was. That is the distinction the "isn't this just SEO" question collapses, and collapsing it is expensive.
SEO was "just" marketing in the Google Era. AI optimization is marketing in the Agent Era.
The claim there is not that a new craft exists. It is that the layer between what you publish and who reads it has changed hands, and the new intermediary reads differently, cites differently, and is measured differently. In the first four months of 2026, 68% of US Google searches ended without a click, up from about 60% in 2024, and roughly 73% of B2B buyers now use AI tools during research. Whatever you call the work, a growing share of your buyers meet a summary of you before they meet you.
What genuinely is not SEO
Two things, and I would hold the line on exactly these two rather than claim more.
Entity resolution. SEO ranks pages. Answer engines retrieve entities, and before an engine can recommend you it has to be confident which company you are. This is where the failure is invisible, because an engine that is unsure does not return an error. It either skips you or describes somebody else, and both look identical from outside. Our own case is the clearest one I can give: we describe what we sell as a managed agent, and that phrase collides with hosted runtime products from Anthropic, Google, and IBM. We were not losing that term. We were invisible underneath it, and no volume of content about managed agents would have fixed it. The fix was changing the frame.
Nothing in a rank tracker surfaces that problem, because it is not a ranking problem.
Measurement. This is the part I have come to think is the real difference, and it gets almost no coverage.
Search gives you a rank. It is stable, it is ordinal, and if you check it twice you get the same answer. An AI answer is a distribution. On July 29 we asked ChatGPT the identical question three times in a row and got three substantially different vendor lists. Of roughly ten companies named across those three responses, exactly one appeared in all three. That is not a bug in our method. That is the medium.
Which means a single screenshot of an engine naming your competitor tells you almost nothing, and most of what circulates as AI visibility evidence is exactly that screenshot. Reporting an answer engine accurately requires repeated sampling, a fixed question set, per-engine reporting rather than one blended score, and stating how many times you sampled and on which surface you captured. None of that exists in standard SEO tooling, because standard SEO tooling was built for a deterministic system.
A useful way to hear the difference: in search, the question is where do I rank. In an agent context, the question is how often am I in the answer, and that is a frequency across repeated draws.
Current view, subject to change
I would drop this framing if either of two things happened.
If the API-versus-browser gap turns out to be large, a good deal of what the category currently reports, including some of ours, is measuring a surface real buyers never touch. We are running that test on ourselves and will publish the result whichever way it falls.
And if answer engines converge toward citing whatever already ranks, then the entity and measurement work collapses back into ordinary SEO, and the honest thing would be to say so and stop charging for it separately.
The evidence I would want against my own position is a company that ignored all of this for a year, kept doing competent SEO, and ended up cited at the same rate as one that did the entity work. I have not seen that study. I have not run it either, which is worth saying.
Final thoughts
Our own site has been running this work on itself since mid-June. Crawler traffic from AI agents has climbed the whole time. On August 2 we measured 32 cells across four engines on our category questions and were named in zero of them.
I publish that number because it is the honest state of a six-week-old effort, and because the alternative is the 900% lift claims this category is starting to fill up with. Ingestion precedes citation by weeks to months. Anyone promising you a date is guessing, and anyone showing you a lift chart without a sample size is showing you a screenshot.
Is it just SEO? Mostly, yes. So was SEO.
Regards,
Charles Stack