AEO basics · · By Gerald De Oleo

AEO vs SEO: Ranking #1 No Longer Means You Get Chosen

The short answer: SEO optimizes for being ranked on a page of links. AEO — answer engine optimization — optimizes for being selected, cited, and described correctly inside an answer an AI writes.

SEO isn’t obsolete, and AEO isn’t a new name for it. It’s also not the same as GEO — generative engine optimization, though the terms overlap. The real difference between SEO and AEO is where the decision happens. In SEO, the buyer reads the results and decides. In AEO, a model reads the results and decides what the buyer sees.

Everything else follows from that.

What SEO was built to measure

SEO has a clean unit of success: a position on a results page. Everything the discipline measures points at it — rankings, impressions, click-through rate, sessions.

The logic was simple. If you rank, the buyer sees you. If they see you, some share click. Once they click, your page makes the case.

That logic carries an assumption: the buyer does the synthesis. They open three tabs, compare, and form their own picture of who you are.

That assumption is what changed.

What changed: the synthesis moved

When an AI answer appears, the model does the comparison before the buyer sees anything. The buyer reads a paragraph that already contains a conclusion.

Pew Research tracked the browsing of 900 U.S. adults in March 2025 and measured what happens when Google shows an AI summary. People clicked a traditional result on 8% of those visits, compared with 15% when no summary appeared. They clicked a source cited inside the summary on just 1% of visits. And they ended their browsing session entirely 26% of the time, versus 16% without one.

Two details in that data matter for B2B.

Question-shaped searches trigger summaries. In the same study, 60% of searches beginning with who, what, when or why produced an AI summary. Buyer research is almost entirely question-shaped: what’s the best CRM for a 20-person agency, how do I automate trade confirmations, which platform handles SOC 2 evidence collection.

Being cited doesn’t bring the buyer to you. At a 1% click rate on cited sources, the summary is the impression. Whatever the model wrote about you is the version of your company the buyer leaves with.

That’s the part SEO has no metric for — and it’s the part where understanding how generative engines assemble their answers becomes essential.

Rankings still help — less predictably than you’d think

The natural response is: fine, but if I rank, the AI will cite me. That used to be close to true. It’s getting less true, and nobody agrees by how much.

Same question, three answers, ranging from a sixth to three quarters.

The honest reading isn’t any single number. It’s that ranking is an input to citation, not a proxy for it. You can rank second and not be cited. You can be cited from page four.

Part of the reason is that AI systems don’t only search your exact query. They generate related sub-queries and pull sources from those too — so the page that earns the citation is often one ranking for a question next to the one you optimized for. This retrieval behavior is one of the key mechanisms that GEO research has studied in depth.

Where SEO’s instruments go blind

I break AI visibility into four stages. SEO covers the first well, the second partially, and the last two not at all.

1. Retrieval — does the model pull your content at all? This is where SEO fundamentals do most of the work: crawlability, indexing, speed, clean HTML. If a system that grounds its answers in search results can’t reach your page, nothing downstream matters. Strong overlap with SEO.

2. Citation — does it use your content as a source? Partial overlap. Ranking helps, as above. But citation also depends on whether your page answers the question directly, whether it’s specific enough to quote, and whether other sources say the same thing. GEO research shows that verifiable claims, statistics with sources and clear prose consistently outperform keyword-stuffed content at this stage.

3. Interpretation — does it understand what your content says? No SEO equivalent. A page can rank and be cited while the model extracts the wrong claim from it — the integration you mention in passing instead of the product you actually sell.

4. Expression — how does it describe your company in the final answer? No SEO equivalent, and it’s where the expensive failures live.

I recently looked at a B2B SaaS company that appears in seven third-party roundups. Cited in every one, every description technically accurate. All seven place it in the category it has since moved away from. The buyer searching for what the company sells today never hears about it — and the buyers who do hear about it get the old pitch.

By presence metrics, that company looks healthy.

Side by side

SEO AEO
Unit of success A position on a results page Being selected and described in an answer
Who does the synthesis The buyer, across tabs The model, before the buyer sees anything
Core question Do we rank for this query? What does the model say about us when asked?
What shapes the outcome Your pages and the links pointing to them Your pages plus what the rest of the web says about you
Main metrics Rankings, clicks, CTR Retrieval, citation, interpretation, expression
Typical failure Not found Found, cited, and described wrong
Unit of research Keywords Questions and prompts

For a comparison with GEO — how it differs from AEO and where the two overlap — see AEO vs GEO: Same Thing or Different Disciplines?.

What doesn’t change

AEO often gets sold as if SEO is dead. That isn’t what the data says, and it isn’t how the work goes in practice.

AI systems that cite live sources retrieve them from a search index. If your site is slow, thin, badly structured or blocked, you fail at stage one, and no amount of AEO fixes it. Technical SEO, genuinely useful content and a coherent site architecture are prerequisites, not legacy.

To be direct about it: a fair amount of what gets pitched as AEO is SEO done well, with a different reader in mind. The genuinely new part is narrower than the hype suggests. It’s also more important.

What actually changes in practice

Your company needs one consistent description, everywhere. In SEO, an outdated directory listing is a minor consistency issue. In AEO it can become the answer. When a company repositions, the old description doesn’t disappear — it competes with the new one, and it’s often winning on volume. I’ve looked at companies with six indexed sources describing the product they used to sell and two describing what they sell now.

Third-party sources matter more — but the engines disagree on how much. One study of 22,295 AI answers to B2B prompts found ChatGPT drew 69% of its citations from brand websites and product pages, while Perplexity and Google AI Mode spread theirs roughly evenly across brand sites, comparison lists and editorial content. Other research reports the opposite for ChatGPT — a strong preference for third-party coverage. The practical conclusion holds either way: your own site has to be the clearest source on what you do, and the lists and reviews in your category have to agree with it.

Content is written to be quoted, not only to be found. Direct answers at the top of sections. Clear definitions. Specific numbers with sources. First-party data nobody else can publish. A model can’t cite a claim it has to dig for. The GEO research paper tested exactly this — and found that adding citations, statistics and quotations improved visibility by 30–40% in controlled experiments, while keyword stuffing performed worse than doing nothing.

The unit of research moves from keywords to questions. Buyers ask models full questions with context — company size, industry, constraints. The content that gets used answers those questions specifically, not the head term.

Measurement moves from positions to descriptions. You still track rankings. You add a second instrument: what the models actually say about you.

How to measure both

Keep your SEO reporting. It still tells you whether you’re reachable.

Then add a prompt set:

  1. Write 20–30 questions your buyer would ask before they’ve heard of you. Not “what is [your company]” — that introduces the brand artificially and tells you nothing.
  2. Run them monthly across ChatGPT, Perplexity, Gemini and Google’s AI results. The engines behave differently, so track them separately.
  3. Log four things for every answer: were you retrieved, were you cited, was the claim interpreted correctly, and were you described in the right category.
  4. Note who appears when you don’t — and which sources they’re being cited from.

That last step is usually the most useful. It tells you exactly which pages are shaping the answer you should own.

FAQ

Is AEO replacing SEO? No. AEO depends on SEO for retrieval. It adds a layer SEO was never built to measure: what the model says about you once it’s found you.

Does ranking on Google help me appear in ChatGPT and AI Overviews? It helps, but it doesn’t decide the outcome. Published estimates of the overlap between top-10 rankings and AI citations vary widely by study, engine and method. Treat ranking as an input, not a guarantee.

What’s the difference between AEO and GEO? They overlap heavily, and the industry uses both terms loosely. I wrote a full breakdown in AEO vs GEO: Same Thing or Different Disciplines?. The short version: GEO focuses on how generative engines assemble answers. AEO focuses on whether you end up as the answer — and are described correctly.

Do I need a separate team for AEO? Usually not. You need someone who owns the second measurement layer — the prompt set — and who can trace what the models say back to the specific pages, listings and descriptions that need fixing.

How long does AEO take to show results? It depends on the engine and on how much outdated information already exists about your company. Fixes to your own site and listings tend to register sooner than earning new third-party coverage. Anyone promising a specific timeline for a specific engine is guessing.

How we work

We don’t start with a pitch. We start with data.

Before any call, I run a free Shortlist Check: three or four of your buyers’ questions in one engine, to see whether you appear and how you’re described. The audit is the full version: every question across ChatGPT, Perplexity, Gemini and Google’s AI results, through the same four-stage check described above — retrieval, citation, interpretation, expression. That’s where you see exactly where you appear, where you don’t, and what the models say about you when they do.

On the call, we walk through findings, not slides. From there, the work follows the layers: fix what’s wrong in how you’re described, make your content worth citing, and measure the results per engine, monthly.

If you want to understand how generative engines decide which sources to use — and what the research says actually works — I break that down in AEO vs GEO: Same Thing or Different Disciplines?.

Book a call — the free Shortlist Check is done before we talk.