What are the sources of AI visibility?

What are the sources of AI visibility?
R. Nagy András
He is the owner and managing director of PRBK, a board member of the Worldcom PR Group EMEA, and a member of the Ethics Committee of the Hungarian Public Relations Association (MPRSZ). With over 25 years of professional experience in communications, he is a regular speaker at both national and international industry conferences and events.

Short answer: According to joint research by Worldcom PR Group and Semrush (June 2026), AI assistants (ChatGPT Search, Gemini, Google AI Mode, Perplexity) primarily cite social platforms — Reddit, YouTube, LinkedIn — rather than official brand content in their responses, though traditional media also rank prominently. 40% of the global top 15 cited domains are social-type sources. This is fundamentally rewriting what “credible communication” and AI visibility means today.

What should you know about the research?

Worldcom PR Group, the global PR network of which PRBK has been a member for over 20 years, partnered with Semrush, a provider specializing in keyword research and competitor analysis, to analyze AI-based search data across a total of 34 weeks between October 2025 and May 2026. The analysis covered eight markets (including Brazil, Germany, the USA, and Singapore) and four AI engines. The goal was to determine where AI engines “draw from” when asked about a brand or market.

What are the sources of AI visibility?
How do social media platforms influence AI visibilty?

The 10 most surprising findings

  1. Reddit beats Wikipedia. Reddit is the most-cited domain globally, with a 31.5% prompt share — more than four times that of Wikipedia (7.4%).
  2. YouTube nearly catches up to Reddit. With a 29.1% share, video content carries the same weight as “evidence” for AI as text-based social content.
  3. 40% of the top 15 domains are social platforms. Reddit, YouTube, X, LinkedIn, Facebook, Instagram — these aren’t just incidental mentions but primary reference points.
  4. There’s no unified AI-source hierarchy. While in the Google era a good SEO strategy worked across every search engine, here each AI engine draws from different sources, requiring engine-specific strategies.
  5. LinkedIn is practically nonexistent for Gemini. 0.0% citation rate across the entire study period — while for Perplexity it climbed to 19.0% (up from an initial 4.8%).
  6. Wikipedia’s dominance collapsed in ChatGPT Search. Its citation rate averaged 19.2%, but fell back to 4.6% in the final week. This suggests that even the “most trusted” source can fluctuate sharply on certain platforms.
  7. Reddit surged in ChatGPT Search. A +31.3 percentage-point jump over the study period (from 12.4% to 43.7%).
  8. Gemini cites sources half as often as the other engines. It only cites in about 62% of prompts, compared to 95–96% for the other engines, making AI visibility harder to measure there.
  9. NIH.gov outranks Forbes. In the global top 15, the American health institute (6.5%) holds a higher position than the iconic business publication (4.9%). Institutional credibility appears to be stronger evidence.
  10. Local sources carry the same weight as global platforms. In Brazil, gov.br and globo.com appear as evidence; in Singapore, local providers; in the UAE, insurance comparison sites. This makes clear that global content alone isn’t enough — local reference points are also needed.

What does this mean in practice?

Brand visibility today no longer depends on what’s published on a brand’s own channels, but on what the surrounding source ecosystem says about them. This system of sources includes institutional websites just as much as traditional media outlets and social media platforms. The strategic response isn’t to “conquer” every domain, but to map the relevant sources by AI engine, market, and category — and deliberately build presence within them.

If you need help developing your GEO strategy, or just want to talk about the topic, reach out to us at rnagy@probako.hu.

Similar articles

Storytelling & AI 

How is AI transforming traditional storytelling? What are the potential dangers of overusing artificial intelligence from a storytelling perspective? What

More