Google's AI Has an UNEXPECTED #1 Health Source (and What to Do About it)

If Google’s AI had to pick one source for health answers above all other options, what would you guess it’d be? Mayo Clinic? NHS? WebMD? That's what we assumed too. Then we pulled data from over 50,000 German health searches. AI Overviews showed up in 82% of them. And when the top cited source showed up, I genuinely thought our query filter broke. If you're doing any kind of YMYL content right now like health, finance, legal, or anything where trust matters, missing this is going to cost you visibility you don't even know you're losing. And I want to show you exactly what's happening — and what to do about it. Drumroll please…. aaand… YouTube is the most-cited source in Google's AI Overviews for health-related searches in Germany. By a lot. It gets almost 4.5% of all citations. That’s over 20,000 links in our dataset. YouTube gets cited 3.5 times more than NetDoktor,

which is Germany's biggest consumer health portal. Twice as much as MSD Manuals, an established medical reference used by professionals worldwide. Now here's the part that should make you stop. In classic blue-link results for the same queries, YouTube ranks 11th. But inside AI Overviews, it’s number one. And it's not just YouTube. NDR, a public broadcaster, is in second place. Sure, they do health content. But they also cover news, sports, and entertainment. Health is just one of the things they do. Meanwhile, government health portals, academic institutions, and hospital networks were nowhere near the top. And look, many of the most-cited videos came from legitimate hospital channels and licensed doctors. So this isn't "YouTube bad." The uncomfortable insight is this: AI doesn't reward who knows the most. It rewards those who are easiest to extract answers from. And right now, video is winning that game. So why exactly is video beating

decades of institutional credibility? There's a specific reason — and once you see it, you'll look at all your content differently. Think about how a typical medical YouTube video is structured. They usually describe the condition, talk about why it happens, and what you can do about it. Clean, direct, extractable. AI can pull from it in seconds. Now think about a typical hospital webpage on the same topic. Comprehensive. Accurate. Thorough. But it often opens with compliance disclaimers, dense background sections, and long paragraphs before it ever gets to a direct answer. Both sources might be equally credible. But one is structured for AI synthesis. The other requires interpretation. And here's the not-so-obvious part: If your expertise isn't structured in a way AI can cleanly extract, it becomes less visible regardless of how authoritative you are. That's a completely different problem — and a much more fixable one. Most YMYL creators are optimizing for human readers.

The ones winning AI citations are optimizing for both. That's the gap. Now here's where it gets more concerning. Because when we looked at the actual quality of sources AI pulls from across the board (not just YouTube), the numbers are hard to sit with. The Guardian ran an investigation and found AI giving medical advice that directly contradicted clinical consensus — including guidance that oncologists called actively dangerous for cancer patients. But individual bad outputs aren't the real story. The real story is what the system is built on. We categorized all 465 thousand citations AI Overviews used and grouped their sources into two buckets: those designed for medical reliability like institutions, academic journals, government agencies, and certified health portals — and everyone else. The result was hard to ignore. Only about 34% of citations came from what we’d consider more reliable sources. Roughly two-thirds came from places that were not built specifically for medical accuracy.

Let’s get specific here, because this reveals how the system actually weights sources. Academic research accounted for less than half a percent of citations. German government health institutions were at 0.39%. International health bodies like NIH or CDC were at 0.35%. Combined, that’s around one percent. Here's the non-obvious implication: the organizations that should dominate AI citations — as in those with clinical accountability and peer-reviewed standards — are essentially absent. That credibility gap is wide open for any publisher who structures their content correctly. Domain authority is outweighing medical expertise. A popular blog can outrank a hospital with 50 years of clinical experience. On top of that, only 36% of the URLs AI cites ever appear in Google's top 10 organic results for the same keywords. So, AI is creating a parallel visibility layer. And most YMYL publishers aren't even tracking it. Think about what that means right now.

Every query your audience searches could be triggering an AI answer. An AI answer that cites your competitor and never surfaces you. And you'd have no idea. Here's how to change that. First: track AI visibility separately from traditional rankings. Look at which queries trigger AI answers, which sources get cited, and how that changes over time. What ranks and what gets cited are no longer the same thing. If you're only monitoring traditional rankings, you're blind to where a significant portion of your audience is now encountering information. And here's what most people don't do: they don’t look at which competitors are getting cited for your target queries even when they don't rank in the top ten. That's your actual competition now. Second: demonstrate your expertise in multiple formats. Not because video is trendy. But because YouTube ranks 11th across regular blue links but 1st in AI citations — and that gap is telling you something. The most-cited YouTube videos in our dataset

weren't random influencers. The majority came from medical channels and licensed professionals. So, the play isn't "make viral health videos." It's "translate real expertise into formats AI can extract from." Publish the video on YouTube. Embed it on your site. Pull the transcript into a text version. Extract key Q&As as standalone pages. Basically, provide multiple entry points for AI to discover and cite you. And another thing almost nobody does: don't just repurpose content — re-structure it for each format. A video that reads like a narrated blog post won't perform. But a video built around a clear question and a direct answer will. Third: structure content around how people actually ask questions. This is the one most people skip — which is exactly why it works. AI favors direct answers to specific questions. Your comprehensive 2,000-word guide has value. But AI is more likely to pull from a 300-word section that directly answers: "what

foods should pancreatic cancer patients eat?" Lead with the direct answer. Then add the nuance. "For most people, X is true. However, if you have Y condition or take Z medication, that changes." Most YMYL writers do this backwards — they build up to the answer. Flip it. Answer first, context second. Fourth: make your credibility signals machine-readable. Add author credentials. Clear bylines. Medical review labels. These aren't cosmetic SEO additions — they're how AI assesses source quality. And video needs the same treatment. If a doctor appears on camera, state their credentials explicitly. On-screen text. Description. Don't assume context is obvious to a system pattern-matching across millions of pages. Here's what almost nobody does: add a dedicated credentials section near the top of every page — not buried in an author bio at the bottom. AI parses page structure sequentially. If credentials appear too late, they may

not factor in at all. And Fifth: monitor what actually gets cited and adjust. Publish, then track citation patterns. If AI pulls from your video but ignores your article on the same topic, that's a format signal. If AI cites your Q&A page but not your comprehensive guide, that's a structure signal. You don't have to guess. The data tells you exactly where to double down. So, whether we like it or not, rankings aren’t the full picture anymore. If you’re still optimizing only for blue links, you’re playing by the old rules. If you want to learn how to actually track and influence your AI visibility, I’ve linked the next videos you should watch in the description below. Be sure to check it out! And don’t forget to like the video and subscribe to our channel for more actionable search insights!