Research shows a 124-million-parameter GPT-2 model improved fact-extraction accuracy from 9.7% to 96.2% by optimizing content variety over mention count.
Fudan University researchers found that 8 of 10 AI search tools, including Grok and ChatGPT, cited a fictional ‘Cactus Gravity Protocol’ within 7 days.
Dejan uses a deep learning model with a 93% confidence threshold to predict human link-desire peaks, enabling adversarial link integration for AI search.
Semrush’s ‘Sources’ report identifies the specific third-party pages AI platforms use as citations, allowing brands to close citation gaps via targeted
Wikipedia is the most-cited domain in ChatGPT, while Reddit generates 5.8 million citations in Google AI Overviews, signaling a shift toward third-party trust.