AEO, from the builder's seat. The stuff you actually change on the site, not the theory.
Try this before you read any further. Open ChatGPT, or Gemini, or Perplexity, and ask it the question your best customer would ask before they'd ever heard of you. Not "what is [your company]." Something like "what's the best tool for [the thing you do], for a mid-size team?"
Then watch what happens.
One of three things will. It recommends your competitors and never mentions you. Or it mentions you but describes you slightly wrong, the positioning you spent two years killing off. Or it nails it. Most founders I do this with get outcome one or two, and the room goes a little quiet.
That quiet is the whole reason AEO exists. So let me walk you through what's actually going on, and more usefully, what I'd change on your site to fix it. This is the builder's version, not the thought-leadership version.
What AEO actually is (without the buzzwords)
Answer Engine Optimisation is making your site legible to a machine that reads and answers, instead of one that ranks and links.
That's the whole shift. For twenty years, SEO was about being findable: show up in a list of ten blue links, earn the click. Now a big chunk of buyers never see the list. They ask an assistant a question and get one synthesised answer, and your job is no longer just to rank. It's to be the source that answer is built from, and to be described correctly when you are.
SEO gets you found. AEO gets you quoted, and quoted accurately. You need both, and the good news is they're built on the same foundation. More on that in a second.
Why this is worth your time right now
Because the behaviour already shifted, quietly, while everyone argued about whether it would.
Gartner reckons a quarter of search-style traffic moves to AI assistants by 2026. Google now shows an AI Overview on more than half of searches. And here's the part that matters for a SaaS founder: the people asking AI assistants about your category are doing exactly the high-intent research that used to land on your site. They're shortlisting. If the model builds that shortlist without you, you didn't lose a ranking. You never entered the room.
The flip side is the opportunity. This is new enough that most of your competitors haven't done the work either. The field is wide open in a way SEO hasn't been since about 2012.
The real reason AI gets your product wrong
It's almost never that you're invisible. It's that you're unclear.
We see the same thing on nearly every project, and it long predates AI: most sites don't have a visibility problem, they have a structure and clarity problem. Pages built around how the design looks instead of what the buyer's actually asking. Copy that lists features instead of answering questions. A product explained in language only the founding team understands.
A human visitor pushes through that fog because they're motivated. A language model doesn't push through anything. It reads what's literally on the page, and if the page doesn't plainly state what you do, who it's for, and how you compare, the model fills the gap with a guess. Usually that guess favours whoever wrote more clearly. Often that's a competitor.
So AEO isn't a trick you bolt on. It's mostly just being clear on purpose, in a way machines can parse. Here's where that gets concrete.
1. Write the answer first, then the page
The single highest-leverage change: lead with the answer.
Models lift well-formed, self-contained answers. If your page on "X vs Y" opens with three paragraphs of throat-clearing before it says anything quotable, there's nothing clean to extract. If it opens with a tight, standalone sentence that directly answers the question, you've handed the model a quote on a plate.
So under any question-shaped heading, the first sentence or two should answer it completely, on their own, with no "as we discussed above" or "it depends." Make it true, make it specific, make it stand alone if you ripped it out of the page and read it cold. Then elaborate underneath for the humans. (Notice that's how this very article is built. That's not an accident.)
2. Structure the page so a machine can take it apart
A model reads structure as meaning. Give it real structure.
That means actual semantic headings, not text that's just visually big. One idea per section. Question-shaped H2s where it fits, because that's literally how people phrase things to assistants. Short, self-contained chunks rather than one rolling wall of prose where every point depends on the one before it.
This is where the build matters, not just the copy, and it's the bit pure-content folks miss. Clean, semantic HTML, a real heading hierarchy, lists that are actually lists in the markup. On the NewsCatcher site, a chunk of the win was structural: we pulled a dense, single-page product into clear, separately-addressable sections, each answering one real question. Sessions went from about 20 seconds to over 3 minutes, because humans and crawlers could finally find the specific thing they came for. Same discipline serves both.
3. Give the machine schema, so it doesn't have to guess
Schema markup is you telling the machine what it's looking at instead of hoping it infers correctly.
It's invisible structured data in the page that says, in a format crawlers and models trust: this is a Product, this is its pricing, these are FAQs, this is the author, this is the organisation. For AEO the useful ones are Product, FAQ, Article, and Organization schema. It won't write your answers for you, but it removes ambiguity at exactly the moments ambiguity costs you, like when a model is deciding whether your "Starter plan" is a real pricing tier or just a heading.
In Webflow this lives directly in the per page settings. It's unglamorous and it's one of the most underused edges available right now.
4. Actually build the pages models reach for
Assistants don't quote your homepage hero. They quote the pages that answer specific questions, so build those on purpose.
The ones that earn citations again and again:
- Definition / "what is" pages for your category and the concepts around it.
- Comparison pages, the honest "X vs Y" kind, because comparison is exactly what someone shortlisting asks an assistant.
- Use-case pages, one real problem each. Buyers don't evaluate by feature list, they evaluate by "will this solve my specific thing." On NewsCatcher we stopped organising around job titles and rebuilt around real scenarios, each starting from a problem. That's the same shape an AI answer wants.
- A "build vs buy" page, if that's a real question in your category. We built one for NewsCatcher, a click-through that made the time-and-cost trade-off tangible. It answered a question buyers were already typing into assistants anyway.
- Clear pricing. Models constantly get pricing wrong because sites hide it. If you can state it plainly, you control how you're described.
You don't need all of these. You need the three or four that match the questions your buyers actually ask.
5. Say it in plain language
Models cite what they can understand, and so do non-technical buyers. Same fix for both.
Deep-tech products lose here constantly. On NewsCatcher we rewrote dense NLP concepts into plain English without stripping the technical meaning, because the buyers signing the cheque often weren't the engineers. A model has the same problem with jargon a busy VP does: if your page only makes sense to someone who already gets it, it won't be the thing that explains you to someone who doesn't.
Write the sentence you'd say out loud to a smart person who isn't in your field. That sentence is also the one that gets quoted.
6. Be the clear, current, primary source
AI engines lean toward content that's recently updated and clearly authored. Stale, anonymous pages get skipped.
Put real authorship on things. Keep your cornerstone pages current, and when something changes, update the page rather than leaving last year's version to be the thing a model learns from. You're trying to be the primary source on your own product and category, the page everyone else (and every model) ends up referencing. Nobody can out-authority you on what your own product does, if you actually write it down clearly.
The step everyone skips: test it
Here's what separates AEO that works from AEO that's a vibe. You check.
This isn't a fire-and-forget thing you do once and trust. It's a loop, and it's genuinely the part most people leave out: ask ChatGPT, Gemini, Claude, and Perplexity the real questions your buyers ask. Note how each one describes you, what it gets wrong, where it reaches for a competitor instead. Fix the page that caused it. Ask again in a couple of weeks once things re-crawl. It's the same testing loop we run on client projects, and it's the only way to know whether any of the above actually moved the needle, instead of guessing.
You can do this yourself, today, for free. Most founders never have.
What I'd actually do, in order
If you want to start this week, here's the sequence:
- Run the audit out loud. Ask all four assistants the five questions your buyers ask. Write down exactly how you're described. This is your baseline and it's usually sobering.
- List the pages that should exist. Map the real questions to definition, comparison, and use-case pages. Note which you're missing.
- Rewrite leading with the answer. On your most important pages, make the first sentence under each heading a clean, standalone answer.
- Fix the structure. Real semantic headings, one idea per section, question-shaped where it fits.
- Add schema. Product, FAQ, Article, Organization on the pages that matter.
- Cut the jargon. Rewrite anything that only makes sense to your own team.
- Re-test. Ask the same questions again in a couple of weeks. Watch the descriptions improve. Repeat on whatever's still wrong.
None of that needs a platform migration or a big budget. It needs clarity and a bit of discipline, which is the recurring theme of basically all of this.
The honest bottom line
AEO sounds like a new dark art, and it mostly isn't. It's the same thing good sites have always rewarded, clarity and structure, now with a second, less forgiving reader who can't squint past your fog the way a motivated human will.
The teams that win the next few years won't be the ones who gamed an algorithm. They'll be the ones whose product is so clearly explained that both a buyer and a machine can describe it correctly without help. That's an achievable bar. Most of your category hasn't cleared it yet.
If you'd rather not reverse-engineer all of this, it's a big part of what we do now, building the structure, the answer blocks, and the schema so AI systems explain your product correctly instead of reaching for a competitor. Happy to run the audit on your site and show you, plainly, how the assistants describe you today.
