GEO, AEO, LLMO: Mostly SEO With a New Invoice

Generative Engine Optimization. Answer Engine Optimization. Large Language Model Optimization. Three acronyms, arriving with the confidence of a new discipline, frequently accompanied by a retainer priced as though it were one.

Some of it is real. Most of it is search engine optimization that has been given a fresh name so it can be sold again at a higher price. Telling the two apart is worth doing before you sign anything.

Start with the mechanism

When an assistant answers a question about current things, it is generally not reciting from memory. It retrieves — searches an index of the web, pulls in pages, and writes an answer grounded in what it found.

Read that again, because the whole argument sits in it. The system is searching an index of the web. Something crawled those pages. Something decided which were relevant and worth surfacing. The retrieval layer is doing a job that looks an awful lot like the job a search engine has always done.

So the question "how do I show up in AI answers" collapses substantially into "how do I get retrieved," and that question has a twenty-year-old answer set.

The parts that are just the fundamentals

Your content has to be reachable in the HTML, because a crawler that does not run JavaScript cannot use what it cannot read. Your pages have to be fast enough to be fetched without timing out. Your structure has to make it clear what a page is about. Your business has to be described accurately and consistently in the places machines look. And other sites need to reference you, because a retrieval system's estimate of who matters is built substantially from what the rest of the web says.

Crawlable, quick, clear, accurate, referenced. If that list looks familiar, it is because it has been the list for a very long time. Anyone selling you these five as a new discipline is selling you a rename.

The parts that are genuinely different

There are real differences, and there are fewer of them than the marketing suggests. In our experience four matter.

Answer-shaped writing. A page organized to answer specific questions plainly is easier to lift a sentence from than a page organized as a brochure. This is not a trick; it is the same clarity that helps a human skim, which is why it also tends to help conversion.

Being quotable in one sentence. An assistant summarizing your service will take a compact, concrete statement over a paragraph of positioning language. If nowhere on your site says plainly what you do, where, and for whom, the model will construct something — and it may construct it wrong.

Off-site mentions carry more weight than they used to. Not links specifically. Being talked about, listed, reviewed and named in places that get retrieved. What the web says about you is increasingly the raw material of what an assistant says about you.

There is frequently no click. Someone can learn about you, form an opinion and decide, entirely inside an answer. That changes what success looks like, and it means being described accurately is as important as being described at all.

Four things. Real, worth doing, not a new profession.

What to ask a vendor

If someone proposes an AI search engagement, three questions sort it out quickly.

Which parts of this are the same work as SEO? An honest answer is "most of it." A vendor who claims none of it overlaps either does not understand the mechanism or is counting on you not to.

How will we know if it worked? Watch for a stable position number. As we have written elsewhere, a generated answer is not an ordered list, and a single rank figure is smoothing away the actual behavior. Frequency of mention across many sampled prompts is the honest version.

What here is proven, and what is a bet? Everyone working in this area is making some bets. The distinguishing feature of somebody worth hiring is that they will tell you which are which.

Our own position, since it is fair to ask

We sell this work. So: it is largely the fundamentals, done properly, plus the four things above. We think the fundamentals are worth paying for because most sites genuinely do not have them, not because we have discovered a lever nobody else knows about. When we do something speculative — and we do — we say so at the time rather than in a footnote afterwards.

If that sounds less exciting than a new discipline, it should. The work that reliably moves this is not exciting. It is a fast site with your actual words in the HTML, pages that answer real questions, facts about your business that are correct everywhere, and enough presence elsewhere that the web has an opinion of you.

Next steps

Take any AI search proposal you have been sent and mark each line item as fundamentals, genuinely new, or unproven. Most of them sort cleanly, and the exercise usually answers the pricing question for you.

If you would like ours marked up that way before you spend anything, that is how we do it — or start with the free audit and see whether the fundamentals are the problem first. Frequently they are.