AI Search & B2B: Why Social Media is Invisible to LLMs

Hermann Bareis

...and what that means for your marketing strategy:

More and more buyers, engineers, and decision-makers are turning to ChatGPT, Perplexity, Gemini, or Claude for their research instead of classic search engines. This raises a critical question for companies: How do these systems actually know what to say about your products and technologies?

The answer is sobering for many: Large Language Models (LLMs) derive their knowledge overwhelmingly from the open web—meaning crawlable websites, industry portals, and press releases. Whatever sits behind a login screen simply does not exist for these systems.

The Invisible Border: Where ChatGPT & Co. Are Blind

For AI models to retrieve content in the moment of a query via live research (Retrieval-Augmented Generation, or RAG), the data must be freely accessible. This is exactly where most social media platforms fail.

In the West: Facebook groups, Instagram profiles, and private LinkedIn feeds sit behind a login. Furthermore, LinkedIn only permits crawling by explicitly approved search engines. Years of social media effort therefore contribute almost nothing to your direct visibility in Western AI answers.
The Special Case of China: The digital ecosystem there is built almost purely on closed app infrastructures. Content from WeChat feeds primarily into Tencent’s own models (Hunyuan/Yuanbao), while other market participants like ByteDance’s Doubao or DeepSeek are locked out. The platforms block one another. Anyone who wants to be found via AI in China must therefore be present in several app ecosystems simultaneously.

From Answer to Click: The Fine Difference in AI Citations

Modern AI search engines embed direct web links (citations) into their answers. How these work depends fundamentally on the market:

The West (Web-Based): A click leads to a normal website. The goal of GEO (Generative Engine Optimization) here is website traffic. Important: If your website relies too heavily on pure client-side rendering (e.g., React, Vue), it remains invisible to many AI crawlers. HTML-based, server-side rendered content is mandatory.
China (App-Based): Links function via deep-linking directly into the respective mobile Super Apps. Since platforms increasingly block AI-driven deep links as unwanted automation, content must be technically prepared so that it is not classified as bot traffic.

Two Worlds, Two Strategies: Website-First in the West, App Seeding in China

If your target audience increasingly uses AI for supplier research, the logic of your content investments shifts fundamentally:

In the West, "Website-First" applies: Your own website is the foundation of your AI visibility. Structured, deeply technical, and cleanly crawlable content determines whether the AI outputs you as a source.
In China, "Platform-Specific Seeding" applies: You must feed content directly into the ecosystems favored by the leading models (e.g., WeChat Official Accounts for Tencent, ByteDance network for Doubao).
Third-Party Sources as Global Amplifiers: Independent industry portals and trade media provide AI systems worldwide with the necessary proof of your expertise.

Social media does not lose its value for human reach and brand building—but for your visibility in AI answers in the West, it now plays almost exclusively an indirect role.

No More Gut Feeling: Make Your AI Visibility Measurable

The decisive question is: How visible is your company today really in global AI systems, and how often are you being cited?

This is exactly where Generative Engine Optimization (GEO) comes in. With our tool aiva, we at effective GmbH analyze your performance in globally relevant systems on a data-driven basis—from ChatGPT and Perplexity in the West to DeepSeek, Doubao, or Ernie in China. We determine your Share of Model, your Citation Frequency, and show you the content gaps compared to your international competitors.

Would you like to know how visible your company is in AI answers worldwide?

➡️ Learn more about aiva here
➡️ Download aiva sample analysis (PDF).
➡️ Request your custom analysis here!