Quick answer: GEO Lab is our running experiment series. We pick target queries, publish content optimized with specific GEO tactics, then check whether AI engines — ChatGPT, Perplexity, Google AI Overviews, and Claude — actually cite us. Then we publish the raw results: what got cited, how long it took, and what seemed to make the difference. No theory-only advice — just what we can observe firsthand.
Most GEO advice (ours included) is built on reasoning about how AI engines work. That’s useful — but the most convincing evidence is a real test. GEO Lab is where we put our own playbook on the line and report what happens, citation by citation.
🧪 Experiment #1 (live now): does llms.txt actually earn AI citations? Google just confirmed it doesn’t use llms.txt — but the AI answer engines haven’t said either way. So we added a real llms.txt to this site, captured timestamped baselines showing zero citations across ChatGPT, Perplexity, Gemini, and Google AI Overviews, and we’re re-checking in a few weeks. Report with before/after evidence coming soon.
What we actually do
- Pick a target query cluster — a specific question (or set of questions) people ask AI engines.
- Publish GEO-optimized content for it, using a stated tactic (answer-first structure, schema, freshness signals, and so on).
- Wait for indexing — usually one to three weeks.
- Check for citations — we ask ChatGPT, Perplexity, Google AI Overviews, and Claude the target questions and record whether GeoParrot is referenced, and how.
- Publish the result as a GEO Lab Report — including the misses, not just the wins.
Why we’re doing this in public
- Original data beats opinion. Firsthand results are exactly what AI engines — and readers — find credible.
- Transparency builds trust. We’ll show what didn’t work, not just the highlight reel.
- It keeps us honest. If a tactic doesn’t move the needle, you’ll see it here.
What to expect
Regular GEO Lab Reports, each tied to a specific query and tactic, with the dates, the engines checked, and the outcome. Over time these add up to a practical, evidence-based picture of what actually earns AI citations in 2026 — and what’s just folklore.
Frequently asked questions
How do you check if an AI engine cites you?
We ask each engine the target question and look at whether GeoParrot appears as a cited source or linked reference, capturing the result for the record.
Will you share tactics that fail?
Yes. Negative results are part of the value — knowing what doesn’t work saves everyone time.
The bottom line
GEO Lab turns GeoParrot from a blog that explains GEO into one that tests it. First reports are on the way — follow along to see what AI search actually rewards.
Leave a Reply