Artificial Intelligence has become a part of our daily lives. Whether we are using ChatGPT, Perplexity or Google’s AI Overviews, many of us rely on these tools to find quick, accurate answers. 

But have you ever paused to wonder where do these AI assistants actually get their information?

That is the question Ahrefs’ Brand Radar set out to answer in June 2025.

Toggle

Free SEO Audit: Uncover Hidden SEO Opportunities Before Your Competitors Do

Gain early access to a tailored SEO audit that reveals untapped SEO opportunities and gaps in your website.

Analyze My Site

In a detailed report, Ahrefs analyzed over 78.6 million AI interactions across the three most popular AI systems. 

The data revealed a consistent pattern: platforms like Wikipedia, YouTube, Reddit and a few key news sites are cited most frequently and are silently shaping the information AI delivers.

These findings give us a glimpse into the digital backbone of AI-generated content. It is not just about what questions people ask but it is about which sources AI trusts to answer them.

So now the real question is: In a future where AI becomes the first stop for knowledge, how can your website become the one it turns to?

What Is This Study by Ahrefs  and Why Does It Matter?

The research was powered by Ahrefs’ Brand Radar, a tool that monitors citation visibility across AI-generated responses. 

It does not just tell you how often a brand or domain is cited but it lets you compare across platforms and queries like a competitive intelligence dashboard.

For June 2025, Brand Radar analyzed:

• 76.7M Google AI Overviews
• 957K ChatGPT prompts
• 953.5K Perplexity prompts

That is 78.6 million AI-generated interactions in just one month.

“Brand Radar is less like a rank tracker and more like a visibility explorer,” says Ahrefs, underscoring how this tool provides unparalleled insight into what AI assistants actually surface.

Which Is the Most Trusted Domain Across the AI models?

Across all platforms analyzed, Wikipedia emerged as the most mentioned domain. However, not every AI uses it equally:

ChatGPT seems more reliant on structured encyclopedic data, whereas Google AI Overviews distribute citations more evenly among user forums and niche sites.This dominance shows that open-source, community-moderated, and structured knowledge wins when it comes to AI trust.

ChatGPT leans on Wikipedia the most, likely due to its structured data and language model training sources. Wikipedia is not only frequently cited but is often used as a first point of fact validation.

YouTube Dominates in Perplexity and Google AI Overviews

What about multimedia content? Surprisingly, YouTube earns the #2 spot overall, especially in Perplexity:

• Perplexity: 16.1%
• Google AI Overviews: 9.5%
• ChatGPT: Doesn’t even crack the top 10

The absence in ChatGPT is telling OpenAI’s system is not built to parse video content natively. Meanwhile, Perplexity and Google AI Overviews increasingly surface how-to guides, tutorials and reviews from YouTube to enrich responses.

For instance, when someone asks, “How to tie a bowtie?”, Perplexity might prefer showing a video tutorial, while ChatGPT would list step-by-step text instructions. 

Is Google’s AI showing a preference for user-generated forums?

UGC (User Generated Content) plays an important role in Google’s AI ecosystem. 

These platforms shine in product reviews, lived experiences and community feedback and Google’s AI taps into that real-world richness especially for trending or ambiguous questions like: 

“What are real users saying about semaglutide side effects?”

“Best backpack for solo hiking—Reddit threads?”

This suggests Google is indexing and summarizing dynamic UGC faster and more effectively than other AIs.

ChatGPT Relies on Authoritative News Sources

OpenAI’s ChatGPT demonstrates a clear preference for news-oriented, credible and timestamped sources. 

Key mentions:

• Reuters
• Apple News (Associated Press)
• AS.com

These publications make up 2.6% to 4% of ChatGPT’s citation landscape. That may not seem huge until you consider the platform’s total data spread.

ChatGPT likely surfaces such sources when queries involve: 

• Real-time developments (e.g., elections, stock updates)
• Global policy (e.g., WHO regulations)
• Breaking stories

By favoring institutional reporting, ChatGPT positions itself as a reliable knowledge agent, though it may lean more conservative in sourcing than its peers.

Health Content: Institutional vs. Local Preferences

Another sharp divide appears in medical and health queries.

• Google AI Overviews: Mayo Clinic, Cleveland Clinic
• Perplexity: Tuasaude (Brazil), Alodokter (Indonesia)
• ChatGPT: None in top 10

This paints a clear picture that Google sticks to U.S.-based institutional authorities, while Perplexity localizes answers based on regional healthcare providers.

This has real implications for health content creators. If you want to be cited by Perplexity in Brazil? Be Tuasaude. If you are a U.S.-based publication? Google is your target. It is an emerging case for geo-targeted medical content optimization in AI SEO.

What is the Difference Between Mention Share vs. Impression Share

Let’s go deeper that not all mentions are equal. The value of a citation depends on how popular the search term is.

 And here Enter:

• Mention Share = How often a domain is cited
• Impression Share = Citation * keyword popularity
• Potential Reach = Estimated exposure based on Google’s search volume data

Let’s take an example: A mention on “AI tools for marketing” (50K searches/month) is far more valuable than one on “history of typewriter ribbons” (500/month), even if both cite Wikipedia.

This is where YouTube acts most; it may not have as many mentions as Wikipedia, but its citation footprint often aligns with high-volume, high-intent queries like “best CRM software 2025” or “how to fix iPhone overheating.”

Similarly, Reddit threads on trending issues can outperform static sites in visibility due to high discussion volumes and indexing speed.

Top 10 Cited Domains – Side-by-Side Comparison

Here’s a consolidated view of the most cited domains by each assistant in June 2025:

AI Assistants Are the New Discovery Engines

AI assistants are quickly becoming invisible gatekeepers of information. Your brand or website might be part of the most influential responses people read and you would not even know it.

But now you do. So the next time you build content, ask: “Would an AI assistant want to quote this?”

Because the brands that dominate AI citation share will be the ones that define digital trust in the years to come.