AEO/SEO
Ask ChatGPT, Perplexity, or Google’s AI Overview to recommend a designer, a tool, or a “best way to do X,” and it will name a handful of sources and quietly ignore everyone else. For the businesses that get named, that’s a new front door. For everyone else, it’s a door that’s closing — and most don’t even know it’s there.
So how do you end up on the named list? Here’s the short version, and it’s the opposite of what most of the “SEO is dead” crowd is selling:
To be quoted by an AI assistant, your page first has to be retrieved — and retrieval still runs largely on the search authority you build with SEO. Rank for the question, then make each answer self-contained enough that a machine can lift it without stitching paragraphs together. Schema and FAQs help at the margins. Extractability is the real tiebreaker.
This is a long one because the topic earns it. We’ll bust the myth, show you the mechanism almost nobody explains, and give you a checklist you can actually apply.
Two camps are loudly wrong right now. One says AEO and GEO (answer and generative engine optimization) have replaced SEO. The other has over-corrected into “ranking doesn’t matter anymore.” The truth sits between them, and getting it right is the whole game.
Start with why this matters at all. About 18% of Google searches now return an AI Overview, and when one appears, people click through to a website on just 8% of searches — versus 15% when there’s no summary. Only about 1% click the source the AI actually cited (Pew Research Center, 2025). Translation: the answer is increasingly the destination, and being the cited source inside that answer is the new visibility.
Now the myth. AEO is not a replacement for SEO; it’s a layer on top of it. AI answer engines don’t browse the whole web in real time — they pull from a candidate pool of pages they already trust, and that trust is built with the same signals that rank you in search. Pages on domains that rank in Google’s top 10 are far more likely to be cited across every major AI engine (Semrush, 2025). You can’t be the answer if you’re not in the room.
But here’s the nuance the over-correctors miss. The “ranking is the precondition” rule is strongest for Google AI Overviews and Perplexity, which lean heavily on top-ranked pages. Ahrefs found that in mid-2025, 76% of AI Overview citations came from top-10 pages; by early 2026 that had settled around 38% as Google’s systems began pulling from a wider set of related queries. ChatGPT and Gemini retrieve more loosely still — they’ll often cite a page ranking 21st or lower (around 90% of the time, per Semrush). That’s not proof ranking is irrelevant; it’s proof that overall authority matters more than your exact position for any one phrase. Either way, the move is the same: build the authority that gets you retrieved, then win on extractability. Strong web and UI/UX work is what earns that authority in the first place.
They run a pipeline, not a popularity contest. Most articles stop at “use schema and FAQs,” but the citation decision happens in five steps, and understanding them tells you exactly where to aim.
When someone asks a question, the assistant rarely searches for that exact sentence. It quietly expands the prompt into many related sub-questions and searches all of them. Ahrefs documented this fan-out directly. The practical takeaway: you’re not optimizing for one keyword, you’re trying to be a good answer to a whole cluster of related questions.
Each sub-question gets converted into a mathematical representation of its meaning, and the engine pulls pages whose content means the same thing — not pages that merely repeat the words. This is why keyword-stuffing does nothing here and why clearly written, on-topic content wins.
The engine merges the results from all those sub-questions. A page that shows up across many of the fanned-out searches outranks a page that’s #1 for just one phrasing. Breadth of coverage on a topic beats a single perfectly optimized sentence.
This is the step that surprises people. The engine doesn’t grade whole pages; it scores individual passages for how well each one answers the sub-question on its own. A page sitting at position #15 can get cited while the #1 result is skipped, simply because it had one cleanly liftable paragraph and the #1 didn’t.
Among everything retrieved, the assistant quotes the passage it can lift with the least risk — the one that answers the question completely, in one place, without needing the sentences around it. That’s the tiebreaker. That’s extractability.
Because ranking gets you retrieved, but it doesn’t make your answer easy to lift. Three things usually block an otherwise-strong page:
Ranking is necessary. It just isn’t sufficient. The fix is structural, which is good news — structure is something you can design.
The single biggest change you can make is to write self-contained answer blocks: a question as the heading, and a direct 40–60 word answer immediately beneath it. Google extracts passages in roughly that length band for snippets, and the same shape is what AI engines lift. Here’s the pattern in practice.
Before (ranks, rarely cited): > Heading: “Our approach to pricing” — “We’ve spent years refining how we scope projects, and over time we’ve learned that every client is different, which is why we take a flexible approach that starts with a conversation about your goals before we ever talk numbers…”
After (built to be lifted): > Heading: “How much does a brand identity cost?” — “A full brand identity typically runs $4,000–$12,000 depending on scope, timeline, and how many applications you need. A standalone logo starts around $1,000. Most projects take three to six weeks. We scope your stage and budget before any work begins.”
The “after” version answers the real question, in one place, with concrete numbers a machine can quote and a human can act on. Do that for every question your customers actually ask, and you’ve built a page made of citable parts.
Two formatting moves compound the effect. Lead with the answer, then explain — most cited passages come from the top of a section, not the bottom. And use lists and tables for anything comparative or sequential; structured data is far easier for an engine to extract cleanly than the same facts buried in prose. It’s the same discipline behind good UI/UX design: make the important thing the obvious thing.
Most AEO checklists are a pile of tactics with no explanation of why each one matters. Here’s a shorter list, with each item tied to the step it serves:
That last point is worth dwelling on, because it’s where most beautiful sites quietly disqualify themselves — and it’s a build decision, not a marketing one. (It’s exactly the kind of thing we’re rebuilding into the xVS Creations site right now: server-rendered pages so bots see real content, the crawler allowlist above, schema on every template, and 40–60 word answer blocks on each service page. This very post ships with its own FAQ and FAQ JSON-LD — we’d rather show the pattern than just describe it.)
They lean on a mix of high-authority reference sites, community platforms, and ordinary business sites. Across 680 million citations, Profound found Wikipedia dominates ChatGPT’s sources, while Reddit is the top-cited source in Perplexity and near the top for Google’s AI Overviews. That scares people into thinking only giant platforms get cited — but the same data shows ordinary .com business sites still receive the large majority of citations overall. Your own well-structured page absolutely can be quoted.
This is also where earned third-party signals matter. AI engines treat independent mentions — reviews, directories, reputable write-ups — as trust votes. A verified profile like xVS Creations’ 4.8-star Clutch rating is exactly the kind of corroborating signal these systems lean on, because someone other than us vouched for it. It’s the sort of third-party proof we point to across our work, too.
Which raises the Reddit question, because a popular shortcut is to manufacture those mentions — seed Reddit threads, use aged accounts, plant a recommendation where an AI is likely to read it. Does it work? Mechanically, yes: researchers showed that a single planted comment of about 13 words could get AI search tools to repeat a chosen brand in roughly half of tests. Some practitioners openly recommend it. We’d point out the other side honestly: it’s the same exploit now being used to push scam products and fake businesses into AI answers, platforms and AI providers are actively hunting for it, and a brand caught gaming it inherits that reputational stain. The line is clean — being genuinely useful in a thread you belong in is earned and durable; planting mentions is manufactured and fragile. Earn it.
Yes — as an investment, not an emergency. Be honest about the size: AI referrals are still a small slice of total traffic today, on the order of 1%. But the slope is steep. AI assistants drove over 1.1 billion referral visits in a single month in 2025, up roughly 357% year over year (Similarweb), and US retail traffic from AI sources jumped about 1,200% in roughly half a year (Adobe). It’s a small channel growing fast and converting well — exactly the kind you want to be early on, not late.
When you measure, track the right thing. Don’t just watch one keyword’s ranking; watch whether you’re being cited across the cluster of sub-questions the fan-out generates, and whether AI is sending referral visits at all. The goal isn’t a number on a rank tracker — it’s being the answer.
Does AEO replace SEO?
No. Answer engine optimization sits on top of SEO. AI assistants mostly cite pages they already trust from organic search, so ranking is the precondition to being retrieved. AEO is the extra layer of structure — clear answers, clean formatting, schema — that turns a ranking page into a quotable one.
How do AI assistants decide what to cite?
They expand your question into many related searches, retrieve pages that match the meaning, merge the results, then score individual passages for how cleanly each answers the question on its own. The passage that stands alone best gets quoted. Authority gets you retrieved; extractability gets you cited.
Why isn’t my top-ranking page getting cited by AI?
Usually because the answer is buried, depends on surrounding context, or is too vague to lift. Ranking gets the page into the candidate pool, but AI engines quote self-contained passages. Add a direct 40–60 word answer right under a question-shaped heading and your odds improve.
Do I need schema markup to get cited by AI?
It helps but isn’t magic. Pages that get cited disproportionately use Organization, Article, and FAQ schema, but that’s a correlation — schema helps machines read good content correctly; it won’t rescue weak content. Treat it as making your page legible, not as a growth hack.
Should I post on Reddit to get my brand cited by AI?
Being genuinely active in communities AI engines trust is legitimately valuable, since Reddit is one of the most-cited sources. Manufacturing mentions with seeded posts or aged accounts is a different thing — it works today but is increasingly detected, is the same trick fueling AI scam recommendations, and puts your reputation at risk. Earn the mention.
Is AI search traffic worth optimizing for yet?
Yes, as a forward investment. AI referrals are still around 1% of traffic, but they’re growing several hundred percent year over year and tend to convert better than average. The structural work that earns AI citations also improves your normal SEO, so you’re not betting on one channel.
The work that gets you cited by AI — server-rendered pages, clean structure, self-contained answers, schema-ready templates — is web design and build work, not a marketing add-on. If you’re building or rebuilding your site, we design it to be AI-citable from the ground up. Tell us about your project.