OtterlyAI's LinkedIn AI citations study analyzed 1,310,455 citations across 384,205 unique LinkedIn URLs collected between January and June 2026, spanning ChatGPT, Google AI Overviews, AI Mode, Perplexity, Copilot, and Gemini. LinkedIn's share of social media AI citations rose from 7.8% in January to 11.7% in May — and one platform accounts for nearly none of it: Gemini cites LinkedIn at 0.2%, while Perplexity leads at 43.3% and Google AI Overviews at 22.2%.

The content format split is stark. Pulse articles represent 63% of cited LinkedIn content but capture 72.2% of citations — averaging 8.5 citations per URL versus 5.9 for posts and 3.0 for profile pages. Named individual authors receive 91.7% of citations despite representing 87.8% of URLs. Company pages average just 5.5 citations per URL. And perhaps most useful for anyone gaming engagement metrics: likes, comments, emojis, and hashtags all correlate near zero with LinkedIn AI citation frequency.

Global generative AI traffic share has shifted dramatically in 12 months. ChatGPT fell from 76.4% to 52.7% — a loss of 23.7 percentage points. Gemini rose from 8.9% to 27.3%, now the clear second platform. Claude expanded from 1.6% to 8.9%, with the fastest month-over-month growth rate in the latest period (+2.9pp versus Gemini's +1.8pp). Platform-specific GEO strategies are no longer optional: optimizing only for ChatGPT now reaches just half the AI search market.

"The LinkedIn data clarifies something important: AI systems don't care about engagement signals at all. Likes, comments, emojis - zero correlation. What they do care about is whether a named human made a specific, substantive claim in a format that's easy to extract and attribute. That's a Pulse article from an individual, not a company post with 800 reactions. The same logic applies everywhere - the meta-analysis finding that brand mentions outperform backlinks 3 to 1 fits perfectly. AI systems are trained on text that references brands. Backlinks are a web-graph signal; they don't appear in the training text. Mentions do. If your GEO strategy is still built around link acquisition rather than mention generation and substantive content from named people, you're optimizing for the wrong signal entirely."