AI Search Visibility Report - May 2026: The Rise of Authority Sources in AI Search

Contents

• YouTube has become non-negotiable AI infrastructure

• Wikipedia’s cemented its influence this month

• NIH's breakthrough signals a shift toward authority

• The model divide is growing wider

• Data journalism is emerging as an AI superpower

• Press releases are under growing pressure

• What brands should do next

• Conclusion: AI is becoming more selective about what it trusts

Last month we published our research into how LLM citation behavior changed between March and April 2026, and now we can share the same analysis for April to May. 

Using Meltwater GenAI Lens, we carried out this research on more than 8 million citations across eight major LLMs, and discovered an increasingly structured ecosystem of trust. While earlier studies showed distinct citation preferences across models, May's data suggests those preferences are becoming more entrenched, and the result is a rapidly emerging hierarchy of AI visibility.

At the top sit YouTube, Reddit, Wikipedia, National Institutes of Health (NIH), and Statista. Beneath them, a growing divide is appearing between models that prioritize institutional authority and those that prioritize social and community-driven content.

For PR, marketing, and communications leaders, the implication of this is that AI visibility is no longer a mere byproduct of digital presence, but rather is developing into a measurable reputation asset.

YouTube has become non-negotiable AI infrastructure

YouTube remains the leading digital estate for AI citation, generating 188,863 citations, growing 56.4% month-over-month, and widening its lead over every other source in our research.

YouTube is now the number one cited source across multiple AI systems, including Google AI Mode, Google AI Overviews, Gemini, and Perplexity.

This shows that AI models are reading transcripts of YouTube videos. The strongest-performing content isn't necessarily highly produced. It is well-structured, clearly explained, publicly accessible information that AI systems can parse, understand, and cite.

For brands, this increasingly makes YouTube a critical AI visibility channel.

Wikipedia’s cemented its influence this month

For several months, Wikipedia has gradually increased its influence across AI systems, and May’s data confirms this is no longer a temporary trend.

Wikipedia citations grew 55.2% month-over-month, reaching 83,191 citations, while visibility scores increased by 50%. 

More importantly, Wikipedia now performs strongly across multiple models. Unlike social platforms that may be favored by only a subset of models, Wikipedia appears to be becoming a shared trust layer across the broader AI ecosystem. The lesson for marketers is that a factual, neutral, well-maintained Wikipedia presence has become foundational to AI discoverability.

NIH's breakthrough signals a shift toward authority

The most structurally significant development in May was the rise of NIH, The National Institutes of Health generated 91,227 citations, increasing 41.3% month-over-month and overtaking both Wikipedia and Statista to become the third most-cited source overall. 

This remarkable development suggests multiple AI models are increasingly weighting scientific and institutional authority when answering questions, particularly in health, science, wellness, and evidence-based domains.

The model divide is growing wider

One of the clearest findings from May’s research is that AI models are becoming less alike, not more, as each platform is developing its own ecosystem of citation sources.

Most cited sources across all AI platforms 

Claude: the data analyst

Claude remains heavily anchored in:

• Statista
• NIH
• Wikipedia

The model continues to favor structured information, datasets, and evidence-backed content.

ChatGPT: the institutional authority

ChatGPT continues to rely primarily on:

• Wikipedia
• NIH

Its citation footprint remains highly concentrated around trusted institutional sources.

Google AI Mode: the reach engine

Google AI Mode generates the highest citation volumes overall and remains heavily influenced by:

• YouTube
• Reddit
• LinkedIn
• Facebook

Gemini: the balanced researcher

Gemini is becoming the most diversified model in the dataset, drawing heavily from:

• YouTube
• Reddit
• Wikipedia
• NIH
• Forbes

Perplexity: the research engine

Perplexity continues to lean heavily on:

• YouTube
• Wikipedia
• NerdWallet

The platform also maintains its distinct avoidance of Reddit.

What does this fragmentation mean? As multiple AI citation ecosystems develop in parallel, there’s no one-size-fits-all strategy for AI visibility - marketers need to balance the needs of several LLMs to ensure their brands perform well wherever customers might be searching. 

Data journalism is emerging as an AI superpower

One of the most interesting findings in May comes at the individual author level. Three of the ten most-cited journalists in the dataset come from Visual Capitalist, a popular site which focuses on publishing data-focused content.

This reinforces a clear trend in the kind of content that we know is most likely to get cited:

• Data-rich
• Visually structured
• Clearly sourced
• Easy for AI systems to reference

But more broadly, high quality, trusted journalism remains a bedrock of AI citation sources. While individual social media platforms often appear as the most frequently cited websites, the long tail of thousands of earned/news media sources cumulatively account for 39.5% of all citations in April, compared to just 7.2% for social media. 

Most commonly cited types of sources

Press releases are under growing pressure

The citation share of press releases dropped by 50% month on month: 

• 0.4% in April
• to 0.2% in May

This highlights that AI systems increasingly prefer third-party validation over self-published messaging. But don’t write-off the value of press releases, because they can still be useful signals for what messages are important to a brand, and can lend weight to the same messages when reiterated through earned media. So just because they don’t get directly cited, that doesn’t necessarily mean they aren’t contributing to your brand’s AI visibility. 

What brands should do next

1. Treat YouTube as core GEO infrastructure

• The platform's citation dominance is no longer experimental.
• Transcript quality, topic coverage, and search visibility should now be considered AI optimization priorities.

2. Invest in Wikipedia as a trust layer

• Wikipedia's growth is accelerating across multiple AI systems.
• Maintaining accurate, neutral, and complete entries is becoming essential.

3. Prioritize data-driven content

• The rise of NIH, Statista, and Visual Capitalist reinforces a clear message:
• Structured information performs exceptionally well in AI environments.

4. Align strategy to individual models

• Claude, ChatGPT, Gemini, Google AI Mode, Perplexity, and Grok are not drawing from identical information ecosystems.
• Optimization should increasingly be model-specific.

5. Continue investing in earned media

• Despite month-to-month fluctuations, earned coverage remains one of the most influential visibility drivers across AI systems.

Conclusion: AI is becoming more selective about what it trusts

The May 2026 data suggests that AI search is entering a new phase of maturity. Rather than converging around a single set of trusted sources, large language models are becoming more sophisticated in how they evaluate information and more specialized in the sources they rely upon. YouTube continues to dominate as the most-cited source in the ecosystem, while Wikipedia is increasingly functioning as a shared trust layer across multiple models. At the same time, NIH's rise into the top three most-cited sources signals a growing preference for authoritative, evidence-based information.

Perhaps most importantly, the differences between models are becoming more pronounced. Claude, ChatGPT, Gemini, Google AI Mode, Perplexity, and Grok are each developing distinct citation behaviors and source preferences, creating a more fragmented AI visibility landscape than many expected. For brands, this means success in generative search is no longer simply about publishing more content. It's about earning visibility within the sources and platforms that AI systems already trust. The organizations that recognize and adapt to this shift early will be best positioned to build authority, influence discovery, and maintain visibility as AI search continues to evolve.

FAQ

1. What was the biggest AI citation trend in May 2026?

The most significant development was NIH's rise into the top three most-cited sources overall, signaling a growing preference for authoritative scientific content across multiple AI models.

2. Why is YouTube cited so frequently by AI systems?

AI systems primarily consume transcripts, metadata, and structured explanations. Well-organized video content has become one of the easiest formats for models to interpret and cite.

3. Is Wikipedia becoming more important for AI visibility?

Yes. Wikipedia citations increased 55.2% month-over-month and now appear consistently across ChatGPT, Claude, Gemini, Perplexity, and Grok.

4. Are press releases losing influence in AI search?

Press releases remain useful for distribution and media engagement, but AI systems increasingly favor earned media, third-party validation, and structured authority sources over direct corporate messaging.