What the data actually says
Meltwater ran a substantial piece of work here. Its GenAI Lens platform analysed 9.5 million AI citations across 16 B2B categories, tracking which domains appeared in generated answers from ChatGPT 5, Google AI Mode, Google AI Overviews, Gemini 3.5 Pro, Copilot, and Claude Sonnet 4.
The headline finding: LinkedIn is the second most-cited domain in AI-generated B2B answers, behind only YouTube. It accounts for 0.53% of total citations, compared with YouTube at 1.52%. For context, Reddit sits at 0.44%, Capterra at 0.38%, and Medium at 0.21%. The three big user-generated platforms together (LinkedIn, Reddit, YouTube) account for 47.5% of all AI citations, versus just 18.7% from company websites.
That last figure is worth sitting with. Your company website, the thing your brand has spent years optimising for search, is getting less than half the AI citation share of social platforms combined.
The personal profile problem for brand marketers
Here's the part that should change how B2B social teams make the case internally. Approximately 75% of LinkedIn citations came from individual member profiles, with company pages accounting for the remaining 25%. That split holds across Meltwater's full dataset.
A separate study from Semrush, which analysed 89,000 unique LinkedIn URLs cited in AI search across 325,000 prompts, adds useful nuance: the individual-versus-company split varies by platform. On ChatGPT Search and Google AI Mode, individual creators account for around 59% of LinkedIn citations each. Perplexity shows the inverse, with company pages taking 59% there. So the model your buyer uses matters. But across the aggregate, personal profiles dominate.
The Meltwater data includes one more detail that cuts against the assumption that reach equals influence: more than half of all citations (51%) came from members with fewer than 10,000 followers. AI models aren't amplifying the loudest voices. They're pulling the most structurally useful ones.
Why LinkedIn specifically
The data reflects something structural about LinkedIn's content, not just its audience size. LinkedIn profile metadata attaches professional signals (title, company, industry) to each piece of content. That gives LLMs a compact credibility signal they can act on when retrieving sources for professional queries. LinkedIn also combines, as Meltwater puts it, professional identity, subject-matter expertise, and structured publishing in a single place, which is exactly what LLMs are optimising for when building B2B answers.
LinkedIn's own VP of Marketing, Davang Shah, responded to the findings directly: "Product and brand discovery doesn't happen in stages anymore. It starts with a question and ends with an AI assistant's answer."
Separate longitudinal data from Profound, an AI marketing intelligence firm, found that between December 2025 and mid-February 2026, LinkedIn more than doubled its domain rank on ChatGPT alone. LinkedIn's citation share also grew 26% across tracked models over the four-week window of Meltwater's research project. This isn't a plateau story. The trajectory is still upward.
What content is being cited
The structural patterns in Meltwater's data are specific enough to act on:
- LinkedIn articles and plain text posts make up 83% of all citations. Multimedia content barely registers.
- Every top-cited article in the study used bulleted or numbered lists. Clear headings appeared in 92% of the most-cited posts.
- Freshness matters: 48% of cited content was published within the previous three months.
- Original content accounts for almost all citations. Reshares do not.
A separate OtterlyAI analysis of 1.3 million LinkedIn AI citations collected between January and June 2026 found that LinkedIn Pulse articles are disproportionately cited: they make up 63% of cited content URLs but capture 72% of citations, averaging 8.5 citations per URL versus 5.9 for standard posts. Named individuals account for 87.8% of cited content URLs and 91.7% of citations.
One finding from OtterlyAI is worth flagging directly because it contradicts a lot of what social media managers are told: likes, comments, hashtags, and emojis all correlate near-zero with LinkedIn AI citations. Engagement optimisation and AI citation optimisation are different games.
What this means for B2B social strategy on Monday
This data reframes the executive and employee advocacy argument entirely. Until now, the case for getting your CMO or your head of sales active on LinkedIn rested on organic reach and thought leadership. That case was always slightly soft. Now there's a harder one: their personal profiles are the most likely vector through which your brand will appear in the AI answers your buyers are reading.
LinkedIn is simultaneously moving to reduce the visibility of low-value AI-generated content and generic posts in its feed, deprioritising content that appears overly automated or short on original perspective. The platform is signalling clearly that templated content will perform worse over time, not better.
The practical implication for teams managing LinkedIn brand pages isn't to abandon them. Perplexity citations, which favour company pages, are still meaningful, and a well-structured company page provides a content hub that employees can amplify. But the asymmetry is now documented: on ChatGPT and Google AI Mode, individual voices generate the citations. Brand pages support them.
The audit question for every B2B social team is no longer just "who follows our page" but "when a buyer asks ChatGPT about our category, whose voice is in the answer?"
Chris Hackney, Meltwater's Chief Product Officer, put the strategic shift plainly: "LLMs are now the first stop for decisions that used to take hours of research. If your brand isn't being cited, you're not in the consideration set."