Generative Engine Optimization (GEO) is the practice of structuring and publishing content so AI systems — ChatGPT, Gemini, Copilot, Claude, and Google AI Overviews — cite it directly in generated answers. A new Meltwater study, produced with LinkedIn, analyzed 9.5 million AI citations across six AI platforms and found LinkedIn is the #2 most-cited domain on the internet, behind only YouTube. For B2B brands, the more important finding is who gets cited: individual experts, not company pages — and 51% of cited authors have fewer than 10,000 followers.

The Study: 9.5 Million Citations, Six AI Platforms

Meltwater, working with LinkedIn’s marketing team, ran the largest public study to date on how AI systems decide what to cite. Using its GenAI Lens platform, the team fired thousands of B2B prompts across 16 categories and logged which domains six AI platforms — ChatGPT 5, Google AI Mode, Google AI Overviews, Gemini 3.5 Pro, Microsoft Copilot, and Claude Sonnet 4 — referenced in their answers [Meltwater]. The findings were first reported by AI & Commerce leader Roger Dunn in his AI Commerce Brief newsletter.

The timing matters. Research from 6sense shows 94% of B2B buyers now use LLMs during their buying process. When a prospect asks an AI assistant which vendor to consider, the brands cited in that answer are the consideration set. Everyone else is invisible.

Where LinkedIn Ranks

Overall internet citation share:

The share numbers look small because AI answers pull from an enormous spread of sources — even the #1 domain captures under 2% of all citations. Nobody owns the “answer layer” of the internet yet.

In B2B categories specifically, LinkedIn performs far better than its overall rank suggests:

• Top 5 cited domain in 14 of the 16 B2B categories tracked
• #1 in Marketing & Advertising and #1 in AI & Data Science
• #2 in Leadership & Strategy, Sales & Revenue, Supply Chain, Consulting, and Financial Services
• #3 in E-commerce & Retail and Technology & SaaS
• Citation share grew 26% over the four-week research window

Zooming out, user-generated and third-party platforms — LinkedIn, Reddit, and YouTube — account for 47.5% of all AI citations, compared with 18.7% for company websites and 15% for peer review sites. Your owned site is one factor in the citation equation, not the whole equation.

Why People Beat Company Pages, 75 to 25

The most consequential finding in the study: 75% of LinkedIn citations come from individual member posts. Only 25% come from Company Pages.

Follower count is a weak predictor of citation likelihood:

• 51% of cited authors have fewer than 10,000 followers
• The 1K–10K follower band accounts for 40% of all citations
• That beats the 10K–100K follower band, which accounts for 38%

AI models appear to weight named authorship, demonstrated expertise, and topical specificity over audience size. A named individual writing with credentials, a title, and real examples gives the model a provenance signal a corporate page cannot replicate. Notably, the most-cited authors skew senior: CEOs, founders, VPs of Engineering, CPOs, and CTOs top the list of cited titles.

The Content Pattern That Gets Cited

Meltwater analyzed the top-cited LinkedIn articles and found a consistent structure:

By format, listicles made up 54% of top-cited content, side-by-side comparisons 50%, and decision guides 33%. Opinion pieces without supporting data barely register. The word-count sweet spot for a cited article is 1,500–2,500 words. And format matters more than production value: plain text posts and articles carry 83% of cited content, with text posts alone at 72% versus 11% for video.

Freshness also matters: 72% of cited content was original rather than reshared, and 48% was published within the previous three months. Citation visibility decays — it is a cadence to maintain, not an asset you publish once.

Caveats to Read the Data With

Four reasons to apply this data carefully rather than take it at face value:

• Conflict of interest. Meltwater sells the AI-citation tracking product this study justifies, and LinkedIn is the platform it recommends investing in.
• Small sample for the content recipe. The structural pattern above comes from analyzing 24 top-cited articles.
• Short measurement window. The 26% citation growth figure covers four weeks, not a confirmed long-term trend.
• Short content shelf life. With nearly half of cited content under three months old, GEO requires ongoing publishing, not a single campaign.

A GEO Checklist for B2B Brands

1. Identify your experts. Find the individuals on your team with genuine subject-matter credibility in your category — not just the brand account.
2. Audit real buyer questions. Collect 25–50 questions your prospects actually ask, in search engines and in AI tools.
3. Map questions to people. Assign each question to one internal expert who can answer it directly, by name.
4. Write decision-support content. Named competitors or tools, explicit criteria, real numbers — comparisons and “how to choose” guides over opinion pieces.
5. Publish on a recurring cadence. Meltwater’s suggested pace is two to three posts per week plus three to four articles per month per expert. A single post will not sustain citation visibility past a few months.

Frequently Asked Questions

Final Thoughts

For twenty years, the job of a brand was to be discoverable. In an AI-first world, the job is to be the answer. This study does not prove LinkedIn is the only place to build that visibility — but it does prove that structured, expert-authored, decision-support content is what AI models cite, wherever it lives. The brands that organize their internal expertise into a repeatable publishing program now will compound an advantage that is still largely uncrowded.