I've been experimenting with GEO instead of traditional SEO for my startup. Here's what actually worked (and what surprised me).
Quick background: I run a startup. We were doing the standard SEO playbook — keywords, backlinks, on-page stuff. It worked okay. About two weeks ago I shifted our focus to what people are calling GEO, and the traffic pattern changed in a way I didn't expect. We picked up around 2K new users, and when I dug into the referral data, roughly 40% came through ChatGPT-connected searches. I can share analytics if people want proof — not trying to flex, genuinely was surprised.
So what is the difference between SEO & GEO?
SEO is keyword-driven. You match what people type into Google.
GEO is a different game. It's about understanding how people query AI tools — which is way more conversational and intent-driven than a Google search. Nobody types "best project management tool 2026" into ChatGPT. They say something like "I'm managing a 5-person remote team and need something lightweight, what should I use?"
The mental model that clicked for me: every product is basically competing in a speech contest inside the AI model. The one the model trusts most gets recommended. Keywords don't win that contest — authority and clarity do.
The thing most people miss
About 90% of ChatGPT and Gemini users are on the free tier. Free-tier users rely heavily on web search — meaning the AI pulls from live web results to generate answers.
Here's the key: you can't really control what the model "knows" from training data. But you can influence what gets pulled in during web-connected search. And from what I've seen over a few months of testing, these sources tend to carry the most weight:
- Official websites and blogs (well-structured, clear answers)
- Wikipedia (if you can get a legitimate mention)
- Reddit (ironic, I know — you're reading this on one of the highest-authority sources for AI search)
- Citations from established media
That's basically the pecking order. If your content lives in those places with real authority signals, it gets fed into the AI's answer.
What I actually changed
I stopped thinking in keywords and started mapping intent chains — what does someone actually ask an AI when they have the problem my product solves? Then I worked backwards:
- Wrote blog content that directly answers those conversational queries
- Started participating genuinely on Reddit in relevant subs (not spamming, actual answers)
- Got our product mentioned in a couple of niche publications
- Made sure our site structure was clean enough for web crawlers to pull the right context
Nothing fancy. No secret tool. Mostly just shifting the mental model from "rank on Google" to "be the answer ChatGPT gives."
Honest caveat
This seems to work well for startups with focused products and clear use cases. I've talked to people at bigger companies trying to do GEO and it's a totally different beast — tons of different intents, legacy content everywhere, brand complexity. Not a simple plug-and-play.
But if you're running something small and focused? The window is wide open right now. Almost nobody is optimizing for this yet.
Happy to answer questions or share more specifics. Still figuring this out myself.