E014: System Memory Longitudinal Citation Study (Perplexity, ChatGPT, Google AI Overviews), February to September 2026

This dataset is observational and wasn’t pre-registered. I deposited it after the series closed on 28 September 2026, together with the full deviation log. E014 checked once a month, from 28 February to 28 September 2026, whether Perplexity, ChatGPT and Google AI Overviews cited thegeolab.net for ten frozen queries. Five Tier 1 queries covered concepts defined on the site: Retrieval Probability, LLM readability, the GEO Stack, Extractability and System Memory. Five Tier 2 queries covered broader category questions. The study asked whether consistent publishing leads to citations accumulating over time, the mechanism The GEO Lab calls System Memory. Tier 1 citation held at 92% to 96% for three consecutive months, from M5 to M7, using the same instrument. Tier 2 stayed at 0% at every clean measurement. The combined rate rose from 6% at M4 to 48% at M5, but that jump coincided with a change to the collection instrument, so I’m not attributing it to accumulation.