The big hypothesis · Running
A developer with no marketing training can learn marketing, SEO and GEO with AI as tutor and practice partner, and get this blog indexed, ranked, cited by AI answer engines, and known.
- By day 90: 10 to 15 pieces published, and search impressions showing in Search Console.
- Supported at 6 months if: at least 2 of the 6 answer engines I track cite this site for at least 1 non-branded question, and at least 1 site links here without being asked.
- Refuted at 12 months if: no tracked engine cites this site for any non-branded question.
Who I am
I’ve spent 20 years building software and data systems: full-stack apps, retail data warehouses, cloud platforms, and lately AI agents. I’m a hacker at heart, the kind who has to take a thing apart before I believe how it works. Most recently I led a global team shipping brand websites for one of the world’s largest food manufacturers, which is where this started.
I write here as GEO Geek, for now without my name. The About page explains why, and what my background is.
How I got here
The brand sites had a tight deadline, so we leaned on AI early, well before coding agents, to build, test and deploy faster. Along the way we built customer-facing chatbots and internal agents that called tools in a loop. I was building for marketers, not doing marketing.
Then the marketing and growth leaders, and eventually the board, asked a question I couldn’t answer: AI is changing how people find products, so how do we show up in it, and can we influence it? The answer I could give was about the engineering. The question was about marketing.
What I’ve learned so far
I thought the hard part would be technical: schema, crawlers, whatever format answer engines prefer. It isn’t, or not first. The hard part is knowing what you want to be found for.
Organisations want to be found for very different reasons, and very few can name the categories, let alone the actual questions, they want to show up for. Working out the goal and finding the right questions is as much of the job as the technical work. You can’t measure visibility until you’ve said visible for what.
So here are mine
Taking my own advice, these are the three things I want GEO Geek to be found for:
- Learning GEO as a developer. How someone who builds software picks up search and generative engine optimisation from zero.
- How AI answer engines choose their sources. Why ChatGPT, Perplexity and the rest cite one site and not another.
- GEO experiments with real data. Tests on a live site, with the numbers published, including the ones that go nowhere.
I expect this list to change as I learn. When it does, I’ll say so here rather than quietly edit it.
How this works
I don’t trust advice I can’t test, so this site is one big experiment and every post is evidence for or against the hypothesis above. Each experiment gets a baseline, one change, a measurement window and a single status: Planned, Running, Supported, Refuted or Inconclusive. I keep three things apart and label them: what I observed, what I think it means (the hypothesis), and what the data lets me conclude.
I measure by hand first, with a fixed set of questions asked of six answer engines, so I understand what the numbers mean before I trust any tool to produce them. The Methodology page has the details.
How AI is used here
Mostly Claude Code: it built this site and helps me write it. Plenty of other AI tools, paid and free, proprietary and open source, will turn up as subjects of experiments. The rule is simple. AI tutors, drafts and critiques. I decide, check and edit, and nothing goes live that I haven’t verified. Every post says what AI did in it, like the note at the top of this one.
What’s next
The first experiment, EXP-001, is a day-one baseline: where a new blog on a reused domain stands in Google, Bing and six AI answer engines before anyone has linked to it, and how that changes over its first 90 days with no promotion. I’ve written down my predictions before taking the first measurement, including one I expect to come true and would rather didn’t: that after 90 days, no engine cites this site for any of my three topics.
That’s coming next, on the experiments page.
Join in
If you’d like to watch the hypothesis play out, follow along by RSS. A way to suggest experiments is coming soon.