A quick note before any of this: AEO and GEO are maybe three years old as real disciplines. I don't think anyone, including me, has this fully figured out yet. What I have is a working process, the same one I'd use for any new marketing problem: learn something, try it, watch what actually happens, and change course when the evidence says to. Good marketing fundamentals still apply. The buyer is still the buyer, even with a language model standing between us.
Why AEO clicked for me
With SEO, you publish a page and then you wait, usually a few months before it ranks and traffic begins. AEO didn't work that way for me. Get the content right, and within days, sometimes a week, you'd see it: showing up in Google's AI Overviews, showing up in ChatGPT's answers, cited accurately for the things you actually worked hard toward. Not keywords. The whole story of what your brand is about.
Where I think people get this wrong
I'll be upfront that I don't have full visibility into everyone else's work in this space, so take this as my view, not a survey of the field. But a few things stand out. Anyone who tells you this is basically just SEO, or anyone who isn't actually measuring themselves against it, is going to lose ground. And a lot of people aren't paying attention to how search itself has changed. Look at what you typed into Google a year ago versus what you type now: longer, more specific, more like an actual question with context in it, because on some level, people have figured out that giving an AI more context gets you a better answer back. That shift in how people ask is something we need to capture as marketers.
How I actually work
Take TiDB. Someone asks ChatGPT about database solutions. Or a developer building an AI app needs a database. Or their AI agent needs to connect to one. The goal is to come up as an option there, and then make the case that we're the right choice. Not the right choice for everything. The right choice for the specific places where we can actually shine, based on what we're genuinely capable of.
Once you know where you want to show up, that becomes the brand story. From there, it's looking at prompt behavior: brainstorming the actual prompts, not alone, but with product, engineering, and the rest of the team. Then lining up a content calendar around the gaps, the prompts where we're not getting cited yet. Content and design teams build from there. This is the exact process behind the TiDB work — the full case study walks through what actually moved.
And separate from the content itself: the website has to render properly, for humans and for AI systems, agents and generative models both, so AI isn't blindsided.
A technical must-have
Get your schema right, on every page, not just a few. This is the foundation the content sits on. Without it, none of the content work matters much, because the systems reading your site can't parse it properly in the first place.
Where to focus your effort
Different AI models source their answers differently, and there's no single approach that works for all of them the same way. Part of the job is figuring out where your actual audience is spending time and making a deliberate, if imperfect, call on which models are worth prioritizing. It's not an exact science yet.
How to actually measure it
Measurement is still catching up. Analytics platforms are evolving to give better data here, but attribution for AI-driven traffic still isn't fully accurate. Worth watching both organic and direct traffic now, not just organic the way we did for SEO, since AI-referred visits don't always show up where you'd expect.
What this looks like in practice
That's roughly the shape of it: get the technical foundation right, find the specific places a product can actually win, build content around the gaps, and keep checking whether it's landing where the answers actually get generated. It's slower to explain than to just show, which is most of why the TiDB case study exists.
If you're trying to solve this for your own product, that's the kind of problem I like being handed.