What the GEO Lab Is and How It Works

We run real A/B experiments on what makes AI engines cite a brand — and we publish the results, including the ones that fail. Here's what the GEO Lab is, why it exists, and how it keeps us honest.

Most GEO Advice Is a Guess

Search the web for "how to get cited by ChatGPT" and you'll find a lot of confident advice. Add schema. Use headers. Include statistics. Some of it is probably right. But almost none of it comes with evidence. It's people guessing, then stating the guess as fact.

We didn't want to be another voice guessing. So we built the GEO Lab — a place to actually test these ideas and find out what's true.

This post explains what it is, in plain terms.

What the GEO Lab Actually Does

The GEO Lab runs controlled experiments. The idea is borrowed from science, and it's simple:

Make a guess (a hypothesis). For example: "Adding a Key Takeaways section to an article makes AI engines more likely to cite it."
Create two versions. Version A has the Key Takeaways section. Version B is identical except it doesn't. Only one thing changes between them — that's the rule that makes the test fair.
Ask the AI engines. We run the same set of real questions across ChatGPT, Gemini, Claude, and Perplexity, and record which version gets cited more.
Do the maths. We check whether the difference is real or just luck, using a proper statistical test.
Publish the result — whatever it is.

That last step is the important one.

Why We Publish the Failures Too

Here's something most companies would never do: when an experiment shows that an idea doesn't work, we publish that too.

That's not us being noble. A "this doesn't matter" result is genuinely valuable. It saves you from wasting time on something that sounds clever but changes nothing. In science, this is called a null result, and a good one is worth as much as a positive finding.

If we only ever published the experiments that made GEO look easy and our product look magic, you'd have no reason to believe any of it. Publishing the failures is what makes the wins believable.

The Rules That Keep It Honest

The Lab runs on a few strict rules:

One change per experiment. If two things change at once, you can't tell which one mattered.
Write the guess down first. We record the hypothesis and what would count as success before running anything — so we can't move the goalposts after seeing the data.
Show the raw data. Every claim traces back to an actual file of recorded AI responses.
Enough trials to be sure. We don't claim a result from three lucky answers. We run enough trials for the maths to mean something.

These rules are boring. That's the point. Boring process is what separates real findings from marketing.

How This Connects to the Product

When the Lab proves something works, that finding flows into L8EntSpace. The content-generation pipeline can apply the tactics that experiments have actually validated — not folk wisdom, but tested levers.

So the Lab isn't a side project. It's the research engine that keeps the product grounded in reality.

Key Takeaways

The GEO Lab runs controlled A/B experiments to find out what actually makes AI engines cite a brand.
Each experiment changes only one thing, writes the hypothesis down first, and shows its raw data.
We publish null results — the experiments that fail — because that's what makes the successes trustworthy.
Validated findings flow into the L8EntSpace product as real, tested tactics.

What's Next

We're running our first batch of experiments now, and we'll publish each result here as it comes in — wins and nulls alike. There's also a YouTube channel coming where we'll walk through the experiments on camera.

Follow along, and you'll learn what works at the same time we do.