AEO or SEO First? Sequencing the Budget

Key takeaways

The premise of the question is mostly wrong. Google states that its AI features draw on core ranking systems, so the work that earns a ranking is largely the work that earns a citation.
Where they genuinely diverge is in format, not in substance. AEO rewards extractable answers, explicit attribution and clear entity definition on top of the same foundations.
So the sequencing answer for most businesses is: do the shared work, then add the AEO-specific layer. Funding a separate AEO programme before the shared work is building a roof first.
There is one real exception. If your buyers have already moved to assistants and your category is answered rather than searched, AEO goes first because the ranking is no longer what gets read.
The measurement problem is the honest reason to care. A citation inside an assistant answer leaves no referral trail, so you cannot wait for the data to tell you to start.
The research that exists points one way. The GEO paper found citations, quotations and statistics lifted source visibility by up to 40% in its test harness, which is the AEO-specific layer in one sentence.

The question arrives in budget meetings as a choice between two line items. It is usually not one, and the people selling it as one have an interest in a second line item.

Google own documentation says its AI features draw on the same core ranking systems as its search results. That single statement removes most of the apparent conflict.

What is left is a genuine and smaller difference, plus one case where the sequencing really does invert. Both are worth getting right, because the cost of the wrong order is a year.

Are AEO and SEO actually different work?

Short answer: Mostly not. Google states that its AI features draw on core ranking systems and that no separate markup opts a page in, so the foundations are shared. What differs is a layer on top: how extractable your answers are and how explicitly you attribute your claims.

That means the fear behind the question, that an SEO budget is now obsolete, is unfounded. The work transfers.

Shared foundations against the AEO-specific layer
Work	Earns rankings?	Earns citations?
Crawlable, server-rendered pages	Yes	Yes, and it is a precondition
Content matching real queries	Yes	Yes
Authority and references	Yes	Yes
Extractable answer blocks	Marginally	Yes, materially
Explicit attribution of claims	Marginally	Yes, materially
Clear entity definition	Marginally	Yes

Read the bottom three rows. Those are the AEO-specific layer, and they are additions to the first three rather than replacements for them.

So what is the AEO-specific layer?

Short answer: Three things: answers written to be lifted out of the page in forty to sixty words, claims attributed to a named source rather than asserted, and an unambiguous statement of what your organisation is and does. All three sit on top of ordinary good content.

The extractability point is the most practical. An assistant summarising a page needs a passage that answers the question without the surrounding context.

A page that only answers its question across four paragraphs is harder to lift than one that answers it in one and then explains. That is a formatting decision with a real effect.

The attribution point has research behind it rather than just intuition. The GEO paper measured citations, quotations and statistics lifting a source visibility by up to 40% in its test harness, which is a research result rather than a guarantee.

The entity point is the most neglected. Many organisations have no single page that plainly states what they are, which makes them hard to name confidently.

When should AEO genuinely come first?

Short answer: When your buyers have already moved. If the people who buy from you ask an assistant rather than running a search, then your ranking is no longer the thing being read, and improving it optimises a surface your market has left.

This is a judgement about your market rather than about the technology. It is true in some categories now and will be true in more later.

Your category is answered rather than browsed. Questions with a correct answer, where nobody wants ten links.
Your buyers are technical or early-adopting. Assistant use skews heavily by profession and age, and some audiences moved years ago.
Your branded search is rising while your clicks are flat. That pattern suggests people are hearing about you somewhere that sends no referrer.
You already rank and it is not converting. Position without clicks can mean the answer is being read above you.

The third signal is the most useful and the least known. Branded search growth with flat click volume is one of the few observable traces of assistant-driven awareness.

Why can you not just wait for the data?

Short answer: Because a citation inside an assistant answer leaves no referral trail. There is no row in your analytics for being named by ChatGPT, so the channel cannot announce itself the way organic search did, and a wait-and-see policy will see nothing.

This is the strongest practical argument for acting before the evidence is comfortable. The absence of data here is a measurement limitation rather than an absence of effect.

Pew found users clicking a result on 8% of visits where an AI summary appeared against 15% where none did, which tells you the surface is absorbing attention even where it sends no traffic.

The proxies available are branded search volume, direct enquiry growth and asking new customers how they found you. All three are imperfect and all three beat waiting.

What does this mean for a budget?

Short answer: Fund the shared foundations as a single line, then add the AEO layer as a modest increment rather than a second programme. For most businesses that layer is an editorial standard applied to work already happening, not a separate team with its own budget.

The editorial standard is cheap to adopt and expensive to retrofit. Every new page gets an extractable answer and attributed claims from the start.

The retrofit question is a separate decision. Applying the standard to a library of two hundred existing pages is a project, and it should be prioritised by which pages already have impressions.

Our own view, applied to our own site, is that the increment is small enough that treating AEO as a separate purchase is usually a sales construct rather than a budget reality.

Frequently asked questions

Is AEO replacing SEO?

No, and Google own documentation on AI features is the reason to think so: the features draw on core ranking systems. What is changing is how much traffic a ranking converts into, which is a measurement and expectation problem rather than a replacement.

Do I need separate markup to appear in AI answers?

No. Google states explicitly that no separate markup opts a page into AI features. Structured data still helps an engine read your page correctly, which is a different and real benefit.

Should I hire a separate AEO agency?

Only if your current agency cannot describe the extractability and attribution layer. For most businesses this is an editorial standard added to existing work, and splitting it across two suppliers creates a coordination cost for no gain.

How do I know if AEO is working?

Imperfectly. Branded search volume, direct enquiry growth and asking customers how they found you are the available proxies, because a citation inside an answer sends no referrer. Anyone claiming precise attribution here is overstating what is measurable.

Does AEO work for local businesses?

Yes, and the local case is simpler, because assistants answering local questions still lean on the same profile and page signals. The extractability layer applies equally; the foundations are the local work you would do anyway.

Sources and further reading

Want the sequencing decided from your own data?

Send us your branded search trend and your click trend. If they are diverging, the order inverts, and that is a judgement we would rather make from your numbers than from a category argument.