Category: Data & Experiments
Controlled experiments measuring how specific content variables affect AI citation rates. Each experiment isolates a single variable — structure, entity density, freshness, or platform behaviour — and tests it under documented conditions with replicable methodology. Results include raw citation data, statistical significance assessments, and practical recommendations based on observed effect sizes.
-
Do AI Detection Scores Affect Google Indexing? I Tested 47 Posts
Read more →: Do AI Detection Scores Affect Google Indexing? I Tested 47 PostsDo AI Detection Scores Affect Google Indexing? I Tested 47 Posts By Artur Ferreira · The GEO Lab Experiment E064. Pre-registered analysis of whether AI detection scores predict Google indexing,…
-
E108: I Ran a Stranger’s 50-Case Test Against My Instrument. The One Disagreement Was the Finding.
Read more →: E108: I Ran a Stranger’s 50-Case Test Against My Instrument. The One Disagreement Was the Finding.A blind run of citation-check against 50 external test cases. The results: 78/79 agreement where the instruments overlap, one construct boundary exposed, and a quantified map of what the instrument…
-
Do AI Overview Citations Change? I Ran the Same Query 10 Times and Got Cited 6
Read more →: Do AI Overview Citations Change? I Ran the Same Query 10 Times and Got Cited 6One query, one page, ten renders in a single afternoon. Google cited the page six times and dropped it three. E101 proves citation is a per-render draw, not a fixed…
-
Two Folds, One Corpus: Published Entity Folding Rules Tested Against connexion.me’s CRM Board
Read more →: Two Folds, One Corpus: Published Entity Folding Rules Tested Against connexion.me’s CRM BoardEntity folding decides whether a product reads as one row or twelve. Nine published rules applied to connexion.me’s 132 AI answers produce 96 canonicals against the board’s 95, and the…
-
GA4 Misses AI Referrals Entirely. Your Server Logs Show How Much.
Read more →: GA4 Misses AI Referrals Entirely. Your Server Logs Show How Much.GA4 captured fewer than 1 in 200 human page views on thegeolab.net and missed every Gemini AI referral. Server logs are the arbiter for AI referral counts.
-
llms.txt vs RSS, Measured: 78 Days of IP-Verified Log Analysis
Read more →: llms.txt vs RSS, Measured: 78 Days of IP-Verified Log AnalysisSeven verified AI-bot fetches of llms.txt in 78 days. The raw ratio to RSS is 1:9.5 and 52% of bot-claimed requests fail IP verification. First-person log analysis with CIDR and…
-
10 Research Papers That Define the GEO Field
Read more →: 10 Research Papers That Define the GEO FieldTen peer-reviewed papers that form the citable foundation for GEO. From the RAG architecture that makes AI search work to the first study that measured citation optimisation as a distinct…
-
What llms.txt Actually Does (And What It Doesn’t): A Measured Assessment
Read more →: What llms.txt Actually Does (And What It Doesn’t): A Measured AssessmentBy Artur Ferreira · The GEO Lab · Published: July 2026 What llms.txt Actually Does (And What It Doesn’t): A Measured Assessment The GEO Lab has an llms.txt. We still…
-
The ChatGPT Pre-Retrieval Gate: How Query Construction Determines Whether ChatGPT Searches at All
Read more →: The ChatGPT Pre-Retrieval Gate: How Query Construction Determines Whether ChatGPT Searches at AllBy Artur Ferreira · The GEO Lab · Published: June 2026 The ChatGPT Pre-Retrieval Gate: How Query Construction Determines Whether ChatGPT Searches at All You checked ChatGPT. Zero citations. The…
-
10 Probe Queries We Use to Test AI Citation Behaviour
Read more →: 10 Probe Queries We Use to Test AI Citation BehaviourTen distinct query types, each targeting a different layer of the retrieval pipeline — the working query taxonomy behind every GEO Lab experiment.