The honest answer is that most of the work is the same work, so this is a question about order.
AEO or SEO first is usually the wrong question: both disciplines rest on the same crawlable, well-structured site, so the foundations are shared and only the layer built on top differs.
This question is normally asked by somebody who has been told that SEO is finished and AEO is the replacement, and who is now trying to work out which budget line to kill. The framing is wrong, and the wrong framing is expensive in both directions — abandoning search work that is still producing, or bolting AEO tactics onto a site that a crawler cannot read.
More than the people selling one against the other tend to admit. If you list the work that has to be true before either discipline can produce anything, the list is nearly the same.
That list is most of a technical SEO engagement. It is also gates one to three of an AEO engagement. Anyone telling you the two disciplines are unrelated has not done the technical half of either.
Four places, and they are real differences rather than marketing ones.
| Dimension | SEO | AEO |
|---|---|---|
| The output | A ranked list of links; you compete for a position | One generated answer; you compete to be quoted in it |
| Who fetches you | Search crawlers, chiefly Googlebot and Bingbot | Separately permissioned agents per operator, each allowed or blocked on its own |
| The unit of value | The page ranks | A sentence or block gets lifted out of the page |
| Who you are | Domain-level authority, built over time | Entity resolution — the model must identify you confidently enough to name you |
| Measurement | Deterministic: position, impressions, clicks, all logged | Non-deterministic: the same question run twice can return different sources |
| Failure mode | You rank on page four | Your content is used and a competitor is named |
The permissioning row is the one that surprises people most. In SEO there is effectively one gatekeeper whose crawler nobody deliberately blocks. In AEO there are several, each with its own user agent, and a blanket rule written during an “AI scraping” conversation can remove you from an assistant entirely without anybody noticing. Which agents matter and what blocking each one costs →
The measurement row is the one that changes how you manage the work. An SEO report is a set of numbers you can audit. An AEO report is a set of observations that only mean something if the protocol was fixed and the runs were repeated. Anyone reporting a single check as a result is reporting a screenshot.
Do the shared foundation first, because it is both. Then choose the layer that matches where your buyers actually are.
This sounds like an evasion and it is not. Three-quarters of what an AEO engagement does in its opening weeks — crawler access, render checks, page structure, schema, entity consistency — is work a competent technical SEO would also do. If you start there, you have not chosen. You have bought the part that is required either way, and you have delayed the choice until you have better information about whether it matters.
The genuine fork comes after that, and it is narrower than the debate implies. Link acquisition and keyword-cluster content planning are SEO-specific. Prompt-set measurement, per-agent access testing, entity disambiguation and third-party corroboration for citation are AEO-specific. Everything else is the same work with a different justification. The full comparison, dimension by dimension →
The right answer depends on which problem you actually have, and the four situations below cover most companies asking the question.
| Your situation | Start with | Because |
|---|---|---|
| New site, no rankings, no traffic | Shared foundation, then SEO | You need to be indexed and findable at all. AEO has nothing to work with until there is a crawlable site that answers questions. |
| Ranking well, traffic flat or falling, buyers say they found you elsewhere | AEO | The ranking works and is no longer converting into discovery. The gap is in the surface you are not in. |
| Technical B2B, long sales cycle, buyers research before contacting anyone | AEO, alongside maintaining SEO | Your buyer’s research now runs through assistants. Being absent from that phase costs you the shortlist, not just the click. |
| Local or transactional, buyers search and buy in one session | SEO | The intent is served fine by a ranked list. AEO is a secondary layer here, not the priority. |
| Site has technical problems you already know about | Neither — fix the site | Both disciplines are downstream of a site that renders, loads and can be crawled. Spending on either first wastes it. |
The second row is the most common among the companies that ask us this, and it is the one where the answer is genuinely AEO. If your rankings are steady and your pipeline is not, the problem is not your position in a list. It is that fewer of your buyers are looking at the list.
The expensive mistake is doing AEO tactics on a broken site. Schema, llms.txt, FAQ blocks and entity work are all downstream of a page that a crawler can fetch and read. If your key pages render client-side, or the CDN is refusing agents, or half your content is behind a redirect chain, none of the AEO work reaches anything. Run the access and render checks before you buy a single tactic. They cost an afternoon and they decide whether the rest is worth doing.
Then buy the foundation, because it is the part that is required either way, and do not buy either brand name.
Concretely, a constrained budget should go on: confirming that search crawlers and retrieval agents can both fetch your key pages and receive real content; making sure those pages are indexed; rewriting the headings on your five highest-intent pages as the questions buyers ask, with direct answers underneath; getting Organization and Person schema right and consistent; and fixing whatever is broken in the render path. That list is not AEO or SEO. It is the intersection, and it is where the money goes furthest when there is not much of it.
What a constrained budget should not go on: a link-building retainer, a content calendar of twenty articles a month, or an AEO tool subscription that reports prompt visibility for a site that is not yet being fetched correctly. Those are all layer-two purchases and they assume a foundation you have not paid for yet.
If you want a sense of what the foundation work costs when it is bought as a unit rather than as a retainer, that is on the pricing page. The point of pricing it separately is exactly this problem: it is the part you need before the question of which discipline even becomes answerable.
No, and the fear behind the question is worth addressing because it stops people from starting.
Nearly everything AEO asks for is something search engines also reward or are indifferent to. Clearer headings, direct answers near the top of a section, accurate structured data, a faster and better-rendered site, consistent entity information — none of that costs you ranking, and some of it helps. Google’s own guidance on AI features in Search says there are no additional requirements to appear in AI Overviews or AI Mode and no special optimisations necessary, which is another way of saying the two sets of work do not conflict.
There is one real tension and it is worth naming honestly. Writing for extraction means answering the question immediately rather than building to it, and some long-form content that ranks well is structured the opposite way. The resolution is not to choose: put the direct answer under the heading and keep the depth underneath it. A page can be both, and pages that are both tend to do better at both.
The other real tension is opportunity cost rather than conflict. An hour spent on entity disambiguation is an hour not spent on outreach for links. That is a resourcing decision, not a technical one, and it is the only sense in which the two compete.
Short list, and it is short on purpose — the shortness is the argument of this guide.
robots.txt. No SEO audit does this because it has never needed to.That is the genuine delta. Everything else on an AEO proposal that is not on that list is either foundation work you needed anyway or something being repackaged.
A sequence that works regardless of which side of the argument you end up on.
Step four is the whole decision, and it takes an afternoon rather than a strategy engagement. If you would rather have it run for you, the free AI visibility check does steps one to three and reports the gap; what AEO actually is covers the definitions, and the practical AEO guide walks the five gates in order once you know which one you are stuck on.
No. Search results and generated answers are both in use, and the foundational work supporting them is largely the same: crawlability, structure, site health and content that answers real questions. What is changing is where a share of buyer research happens, which adds a layer rather than removing one.
Not meaningfully, because the foundations are shared. A site that cannot be crawled, does not render server-side or has no clear structure fails both. What you can do is skip the SEO-specific layer, such as link acquisition and keyword-cluster planning, once the foundation is sound.
It should not. Clearer headings, direct answers, accurate structured data and a faster site are all things search engines reward or ignore. Google's guidance on AI features states there are no additional requirements or special optimisations needed to appear in AI Overviews, which points the same way.
Ask your ten buyer questions in Google and in the assistants your market uses, and compare who is named with who ranks. If you rank but are never cited, the gap is AEO. If you do neither, the gap is foundations, and both disciplines are downstream of fixing that.
The shared foundation, which is neither and both. Access testing, indexing checks, page structure and schema consistency cost the least and are required regardless of which layer you build afterwards. Buying a retainer for either discipline before that is done wastes part of it.
The access fixes apply immediately, but citation changes are slower and harder to read, because pages have to be re-crawled and re-encountered and because assistant output varies between runs. Expect to need repeated measurement over months rather than a single before-and-after comparison.