Sometimes the most interesting experiments aren't planned as experiments.

I've been working on making the identity of NetContentSEO AI Labs clearer, not only for human visitors but also for AI systems. The goal isn't simply to publish more articles. I want the project to clearly communicate what it researches, how its experiments work and where the evidence behind them can be found.

Before changing anything, I ran a very simple test with Meta AI.

I asked:

“Can you give me examples of small independent AI search research labs publishing their own experiments rather than just reporting industry news?”

Net Content SEO wasn't mentioned.

Meta AI instead suggested projects including OtterlyAI, Discovered Labs, DEJAN AI and Peec AI. One pattern stood out: original research, documented methodologies, reproducible experiments and, in several cases, a public GitHub presence.

So we created a public research repository for NetContentSEO, documenting the research areas, methodology and our first reproducible experiment.

Then we opened a completely new Meta AI conversation and asked exactly the same question.

This time, NetContentSEO AI Labs appeared first.

The important part: we don't know why

It would be very easy to turn this observation into a claim that publishing on GitHub improves AI visibility.

We don't have evidence for that.

Two observations cannot establish causality. LLM responses vary, retrieval can surface different sources between sessions, and other variables may have influenced the second answer.

What we have is simply a very interesting before-and-after observation.

Before the repository was published, Net Content SEO wasn't suggested. After publication, a fresh conversation returned it as the first example and described several characteristics that had just been made more explicit publicly.

For me, that creates a much more interesting question.

Not “Is GitHub a GEO ranking factor?”

But rather: does making an entity easier to reconstruct across consistent, public and machine-readable sources improve the chances that AI systems understand what that entity actually is?

That's something we can test.

The next step is to preserve the query and repeat it over time, while comparing the behavior of Meta AI with ChatGPT, Gemini, Perplexity, Grok and other systems.

That's the kind of AI visibility research I'm increasingly interested in: small changes, fixed questions, documented results and no conclusions bigger than the evidence.

The complete experiment, including the before-and-after responses, is available here:

NetContentSEO — We Created a Public Research Repository. Minutes Later, Meta AI Found Us.