Until about 18 months ago, "SEO" meant exactly one thing: get a page to rank for a keyword, and a human will click the blue link. Everything we did — keyword research, on-page optimization, link earning, technical hygiene — was in service of that single feedback loop.
That loop is still there. It's just no longer the only loop that matters.
Today, a meaningful share of category-level queries never produce a blue-link click. They produce an answer — a synthesized response from ChatGPT, Claude, Perplexity, or Google's AI Overviews — where one or two sources get named, and the rest get summarized away. If you're not the source that gets named, you're invisible. The user got the answer they needed and never visited your site.
Which means brands now need to win on two fronts at once. We call this the two-SEO problem: page ranking on the classic Google surface, and claim ranking on the AI answer surface. They look similar from outside. Inside the content, the optimization patterns diverge sharply.
What changes between SEO and AEO
The shortest version: SEO rewards relevance + authority + technical health. A page wins because it's the most useful, most credible, most accessible answer for a query. The optimization unit is the page.
AEO is different. AEO rewards claim density + entity clarity + citation patterns. A page wins because the language model, when summarizing what's known about a topic, finds your specific factual claim, your specific entity name, and your specific source pattern more trustworthy than the alternatives. The optimization unit is the claim — a single sentence that asserts something concrete about the world, attached to an entity the model recognizes.
Practically, that means content that wins on AEO has three traits that most SEO content lacks:
- Attributable facts. Specific numbers, dates, named sources — not vague aspirations. "Conversions increased" loses. "Conversions increased 42% over six months, verified by Google Analytics 4 and a third-party audit by [firm]" wins.
- Entity scaffolding. The model needs to know who is making the claim. Schema.org markup, Wikidata entries, consistent author bylines, About-page entity data — all of it builds the entity graph the model uses to decide whether to trust you.
- Citation-friendly structure. Short paragraphs. Definitive sentences early. Direct answers before the long context. Headers that mirror the question shape. AEO content reads more like a reference card than a long-form essay.
The trap most agencies fall into
The trap is treating AEO as "SEO with more schema." Add some FAQ markup, drop in a few JSON-LD blocks, call it done. It won't move the needle, because the underlying content wasn't written to be cited — only to be ranked.
The reverse trap is just as bad: chase AEO so hard that the page becomes a list of bullet-pointed facts with no narrative, no voice, no reason for a human reader to engage. You get cited by ChatGPT and ignored by every human who clicks through. Worst of both worlds.
The brands winning right now treat the page as two documents in one: a narrative top that earns the human read, and a factual middle that earns the LLM citation. Same page, two audiences.
How we structure it
For every page we ship as part of an integrated SEO + AEO engagement, we run a checklist that looks roughly like this:
Above the fold: the human earn
- One clear H1 with the brand voice intact
- A 2-3 sentence lead that frames the stakes for a human reader
- No JSON-LD or technical detritus — this is for humans
Middle: the model cite
- One claim per paragraph, supported by a specific number, date, or source
- Headers in question form where appropriate ("How long does X take?", "What's the difference between X and Y?")
- Entity references linked to canonical About / Author / Service pages
- A FAQ block at the bottom, structured with FAQPage schema
Below the fold: the trust signal
- Author bio with credentials, linked to Person schema
- Last updated date, visible to readers
- Internal links to two or three deeper resources — both for human readers and for the entity graph
The measurement problem
Here's the uncomfortable part: AEO is harder to measure than SEO. Google Search Console gives you impressions and clicks. There's no equivalent for "times ChatGPT cited us in the last 30 days."
What we do measure: citation share per query category (manual + scripted sampling of LLM responses to commercial queries), branded search lift (the downstream signal that brand awareness from AI mentions is converting), and direct referral traffic from Perplexity, ChatGPT, and Gemini — all three now drive measurable visits to commercial pages.
None of that is as clean as classic SEO reporting. But the brands willing to invest before the measurement caught up are the ones embedding into the training data of every future model. That's a moat worth building, even when the dashboard is fuzzy.
Where to start
If you're already doing SEO competently and want to add AEO without burning everything down, start here:
- Audit your top 10 traffic pages. Add Schema.org Article + FAQ markup if missing. Add explicit author bylines linked to Person schema.
- Pick five commercial queries where you'd like to be the LLM citation. Manually sample what each model currently cites. Note the patterns.
- Rewrite one of those query-mapped pages with claim density in mind. Ship it. Re-sample LLM responses in 6-8 weeks.
- If it moved, scale the pattern. If it didn't, the entity foundations weren't strong enough — that's the next layer of work.
The new SEO is two SEOs. The good news is they reinforce each other: a page that earns AEO citations also tends to outperform on classic SEO, because the same signals (depth, authority, structured data) feed both surfaces. The bad news is that doing one without the other gets you a smaller share of an already-shrinking pie.
Two SEOs. Same playbook. Different scoreboards.