Answer Engine Optimization: how growth changes when AI answers first
Growth used to start with a search box. Someone typed a few words, Google returned ten links, and the game I learned was ranking in that list. People don’t find things that way anymore. They ask Claude, ChatGPT or Perplexity a full question and they get one answer. I do it myself, most days I ask an assistant first and open a search page second.
I’ve spent the past year hands-on with this shift, building tooling and content around it, and this is what I’ve learned so far.
What is AEO?
Answer Engine Optimization (AEO) is the work of getting your content cited inside AI answers. Traditional SEO gets you ranked on a results page and hopes for the click. AEO gets you quoted by Claude, ChatGPT, Perplexity and Google’s AI Overviews when someone asks a question you can answer.
The questions I watch people ask these engines are different from search queries too. The average chat prompt runs about 25 words, and the average Google search about 6, a gap Graphite’s Ethan Smith calls out on Lenny’s Podcast. Nobody types a long, specific, conversational question into a search bar, but they ask assistants exactly that, so the tail of demand you can capture is much longer than it was in classic search.
Why this matters now
Two numbers frame the whole thing for me. Industry studies put AI Overviews in roughly 45% of Google searches now, and when one appears, clicks to websites drop by as much as 58%. The old playbook I was trained on, rank the page, win the click, convert on your site, is losing its middle step.
I don’t read that as search dying, though. Smith’s take is that Google’s slice of the pie stays about the same while the pie gets bigger. The new demand is showing up in chat, and that part is still wide open.
At the same time, Graphite’s data shows visitors who arrive from LLMs are about 6x more likely to convert. Fewer clicks, better clicks. The people who still land on your site have already read the summary and they’ve come anyway, so they show up warm.
What I actually did about it
My first move was measurement. I built AEO analysis tooling to track how a site and its blog surfaced in AI answers, so I could see which questions it showed up in and where competitors got cited instead. Before I built it I couldn’t see any of that, I was guessing.
My second move was chasing real demand. I pulled Google Search API data to find what people were actually asking, then I turned that demand into a content pipeline across blog, YouTube and Reddit. Same questions, three surfaces, and the answer engines read all of them.
My third move was community. I built an internal platform that finds Reddit threads worth joining. That one felt strange to me at first, writing code to find conversations, but that’s how these engines behave: research shows brands are 6.5x more likely to get cited through third-party sources than through their own domain, and Reddit is one of the places AI engines read most. Showing up honestly in the right threads became a repeatable acquisition channel. And the volume you need is small. Smith says a handful of useful comments from a real account, name and employer stated, can be enough.
What actually gets cited
A Princeton study on generative engine optimization (KDD 2024) measured what makes content more visible in AI answers. Citing sources boosted visibility about 40%. Adding statistics, about 37%. Keyword stuffing lowered visibility by about 10%.
There’s also a mechanic I had to unlearn from SEO. In classic search you fight for one blue link. An AI answer summarizes many citations at once, and Smith’s observation is that the brand an engine names first is usually the one mentioned most often across its sources. So rank in one place matters less than presence in many, which is exactly why the off-site work above pays off.
The pattern I take from all of it: answer engines lift passages, not pages. Lead with the direct answer, keep it self-contained, put real numbers next to real sources, and use headings that sound like the questions people actually ask. I shaped this article that way on purpose.
There’s one more finding I keep coming back to. Graphite found that 82% of the articles ChatGPT and Perplexity cite are human-written. So if you’re wondering whether pure AI content wins in AI search, the data says mostly no. Hybrid wins: AI for structure and speed, a human for the substance that’s actually lived.
Where I’d start
One thing that surprised me: this channel is fast. SEO makes a small company wait years to build domain authority, but one Reddit thread or one YouTube video can get you cited in an AI answer almost immediately. If I were starting AEO work at any company today:
- Ask Claude, ChatGPT and Perplexity the ten questions your buyers ask, and note who gets cited, so you know your baseline before touching anything.
- Rework your most important pages to answer first: a clear definition up top, stats with sources, headings that match real questions.
- Go where the engines read. Reddit, review sites, industry publications. Your own blog is only one voice in your favor.
- Write the content yourself, or at least keep a real person’s experience in it, since human-written work is most of what’s getting cited.
I don’t know how stable any of this is. Citation patterns shift whenever the model providers change how retrieval works, and what wins this quarter may need remeasuring next quarter, which is why I’d build the measurement before I built the content.
Search used to end with a list of links. Now it ends with an answer, and growth is the work of getting into it.
Sources: Graphite (graphite.io/aeo), Ethan Smith on Lenny’s Podcast (“The ultimate guide to AEO”, Sep 2025), the Princeton GEO study (Aggarwal et al., KDD 2024), Graphite’s research on human-written vs AI-generated content in search (Oct 2025).