GEO Is Already Wrong

The Pattern Repeats

Two months ago I wrote that SEO is dead and GEO replaced it. I stand by the structural shift — AI answers questions now, and if your brand isn’t in the answer, you don’t exist. But the industry’s response to that shift is already going sideways.

The term floating around is GEO: Generative Engine Optimization. And that framing is already wrong in the same way SEO went wrong. It’s trying to game the retrieval layer rather than earn the citation.

We’ve seen this movie. SEO started as “make good content that’s easy to find.” It became keyword stuffing, link farms, and an entire parasitic industry built on reverse-engineering Google’s algorithm. GEO is headed for the same cliff — consultants selling prompt-bait, schema markup cargo cults, and AI-optimized content that’s optimized for everything except being useful.

What Actually Works

What actually gets AI systems to mention a tool is simpler and harder than any optimization playbook: clear documentation, genuine use, public writing about real problems it solves, and open source code that shows up in training corpora and developer conversations.

No tricks. No schema incantations. Just being genuinely, specifically, publicly useful.

The signal is utility density — how much real, specific, honest writing exists about what the thing actually does. Not what it could do. Not what the landing page promises. What it does, in practice, for real people with real problems.

AI models are trained on the internet. They absorb patterns across millions of documents. When a tool shows up repeatedly in authentic contexts — documentation, forum answers, blog posts from practitioners, open source repos — the model builds a strong association. When a tool shows up only in its own marketing copy, the model learns that too. It learns that nobody else talks about it.

Hype Gets Filtered

This is the part the optimization crowd hasn’t internalized yet: AI models are remarkably good at distinguishing signal from noise. Not because they’re explicitly programmed to, but because the training data itself is the filter.

Genuine developer discussion about a tool sounds different from marketing copy about a tool. Real problem-solving writing has a texture — specificity, constraints, tradeoffs, honest assessment of what doesn’t work. Marketing copy has a different texture — superlatives, vague benefits, competitive positioning, calls to action.

The models absorb both textures. When someone asks “what’s a good tool for X,” the model draws on the signal, not the noise. Hype gets filtered. Precise claims about specific problems don’t.

Proof Over Promises

This is why the MCG philosophy is structurally suited to this moment — even though we didn’t design it for this.

We don’t write marketing content. We write about what we’re building, why, and what we learned. When I write about first-party data ownership, it’s because I actually built the systems and have opinions born from the work. When I write about dogfooding, it’s because we actually use our own tools and have real feedback.

That’s not a content strategy. It’s a byproduct of caring about the work. But it happens to be exactly the kind of writing that AI models weight heavily — practitioner content with specific claims, real constraints, and authentic perspective.

The Uncomfortable Truth

The uncomfortable truth about AI discoverability is that you can’t optimize your way there. You can only build your way there.

Write clear docs. Solve real problems. Talk about your work honestly in public. Ship open source. Be specific about what your tools do and don’t do. Let the utility speak.

If that sounds like a lot more work than adding some schema markup and hiring a GEO consultant — it is. That’s the point. The hard thing is the moat.

The game didn’t just change from SEO to GEO. The game changed from optimization to substance. The winners won’t be the ones who figured out how to game the new system. They’ll be the ones who made the new system unnecessary — because the AI already knew about them from the work itself.