Melissa Rosenthal, co-founder of the American owned-media firm Outlever and former VP Creative at BuzzFeed, published an analysis on LinkedIn at the start of May of a new study from Semrush. The study analysed 325,000 unique prompts across ChatGPT Search, Google AI Mode and Perplexity in January and February 2026, and identified 89,000 unique LinkedIn URLs that appeared as cited sources in AI search answers. According to Semrush, LinkedIn ended up as the second most cited domain across the three platforms, appearing on average in 11% of answers. On ChatGPT Search it was 14.3%, on Google AI Mode 13.5% and on Perplexity 5.3%. On ChatGPT and Google AI Mode, 59% of cited LinkedIn content comes from individual profiles rather than company pages.
What Semrush found
Semrush went through 12 major industry categories and measured how much LinkedIn content ends up in AI answers as citations, which forms of content generative models select, and who published it. The study “We Analyzed 89K LinkedIn URLs Cited in AI Search” ran in January and February 2026 and covered ChatGPT Search, Google AI Mode and Perplexity.
The split of citations between individual profiles and company pages proved to be a key variable. On ChatGPT Search and Google AI Mode, 59% of cited LinkedIn content comes from individual profiles. Perplexity shows the reverse ratio, with 59% of its LinkedIn citations pointing to company pages. On ChatGPT and Google AI Mode, the personal accounts of employees and founders predominate in citations. Company pages capture a smaller share of citations on these two platforms.
Length, originality and educational intent
According to Semrush, articles of between 500 and 2,000 words make up 50% to 66% of all LinkedIn citations in AI answers. Feed posts of 50 to 299 words cover the rest of the main share. Around 95% of cited content consists of original posts and articles, with only about 5% reshares. Educational and advisory content makes up 54% to 64% of citations. Promotional content, according to Semrush, barely appears in citations.
Semantic similarity between cited LinkedIn content and the AI model’s answer ranges between 0.57 and 0.60. According to Semrush, models often reproduce the whole structure and argument of the source text.
Reaction count and posting frequency of cited authors
The median number of reactions on a cited post is 15 to 25, according to Semrush. The median number of comments does not exceed 1. Around 75% of the authors whose content was cited published at least five times in the preceding four-week window. Roughly 50% of them have 2,000 or more followers.
Generative AI cites a consistent author with 2,000 followers who wrote five serious texts in the given month, with a median of 20 reactions. A viral post with five thousand reactions does not dominate the citations. State of Brand, in its analysis of the study, quotes the key conclusion: “Engagement numbers don't predict citation. Most cited posts had only 15 to 25 reactions. The signal that matters is consistency.” Rosenthal summarises that citability is determined not by the level of engagement but by consistency.
Ghost citation and brand mention
Rosenthal’s analysis also covers what happens when AI takes content but does not name the brand. According to State of Brand, a publication owned by Outlever, which published the analysis of the study on 23 April 2026, the citation rate in such a case falls from 53% to 10.6%. State of Brand states verbatim: “When AI cites content without mentioning the brand in the response text, the citation rate drops from 53% to 10.6%.” In other words, if AI cites content without mentioning the brand in the text of the answer, the citation rate falls from 53% to 10.6%. If the author names their company or product directly in the article text, AI is more likely to add the brand name to the citation. If not, AI uses the content and sends the potential customer to a competitor.
Ranking of citation sources by platform
LinkedIn competes for citations with Wikipedia, Reddit, YouTube and traditional publishers. According to an earlier three-month Semrush study covering July to October 2025, which measured more than 100 million citations, the per-platform rankings differ. On ChatGPT the top five are Reddit, Wikipedia, Medium, Forbes and LinkedIn. On Google AI Mode the top five are LinkedIn, YouTube, Reddit, Google and Google Blog. On Perplexity they are Reddit, LinkedIn, NIH, Microsoft and Google.
The ranking varies by platform. Google AI Mode gives LinkedIn the first position and Perplexity the second, while ChatGPT places it fifth, behind Reddit, Wikipedia, Medium and Forbes. For European brand teams, the difference is whether they optimise content for Google AI Mode, which reaches users through Google Search, or for ChatGPT and Perplexity, where different domains carry different weight in citations.
The European regulatory layer
Article 50 of the EU AI Act takes effect for all systems on 2 August 2026. Article 50(1) requires providers to inform users that they are communicating with an AI system, unless this is obvious from the context. Article 50(2) requires providers of generative AI systems to ensure their outputs are “marked in a machine-readable format and detectable as artificially generated or manipulated”. For generative AI systems placed on the market before 2 August 2026, a transitional period to 2 December 2026 applies to meeting this requirement, based on the AI Omnibus agreement of May 2026.
The Code of Practice on marking and labelling of AI-generated content, which specifies these obligations, is in a draft open for public consultation until 3 June 2026. Article 50 addresses the marking and detection of AI output, not attribution directly to a cited source. The question of whether an AI system in the EU will have to name the specific brand or the author’s LinkedIn profile from which it took content remains open in the draft guidelines, and its concrete impact will depend on interpretation and enforcement.
Outlever and Semrush build on the same data
Outlever, which published Rosenthal’s analysis, builds its business on the thesis that every company needs its own brand newsroom. Semrush, which collected the data, builds on the thesis that SEO is turning into AI visibility. In between sits every CMO who has a LinkedIn ad campaign budget and no plan for who in their team will write five articles next month. The EU AI Act grace period until autumn 2026 gives European companies less than three months to decide who will publish under their own name.
Comments