YouTube now accounts for just over 3% of all links cited by Google AI Overview — a huge share for a single domain in such a long‑tail ecosystem. When users see AI Overviews at the top of Google, there is a very good chance at least one of the citations comes from a YouTube video.
In this post, we use Spotlight’s data to show how big that YouTube number really is in context, and how it compares with other “everyday” platforms like Reddit, Wikipedia, LinkedIn, Facebook, TikTok, and Instagram across six major models: Google AI Overview, ChatGPT, Gemini, Google AI Mode, Perplexity, and Grok.
All numbers come from Spotlight’s internal analysis of 5.8 million links cited in AI answers over roughly the last month, across 20 countries. For each model, we measure how many times it cited a given domain and what percent that domain represents out of all links the model used in that period. This is not web‑wide traffic data; it’s a window into how models behave when they answer real user prompts. Percentages are rounded to two decimals.
For background on how AI answers are generated and evaluated, see resources from OpenAI, Google Gemini, and Microsoft Copilot, as well as neutral references like Wikipedia’s overview of large language models.
How much does Google AI Overview rely on YouTube?
Across all the models we studied, YouTube shows up again and again — but nowhere as strongly as in Google AI Overviews, where it crosses the 3% mark by itself. Below, we break down the most‑cited domains (from our focus list) for each model, starting with Google AI Overview.
How much does Google AI Overview rely on YouTube, Reddit, and friends?
For Google AI Overview, the last month of data shows:
- YouTube: about 3.29% of all links cited
- Reddit: about 0.96%
- Facebook: about 0.59%
- Instagram: about 0.50%
- LinkedIn: about 0.48%
- TikTok: about 0.24%
- Wikipedia: about 0.22%
Even though YouTube sits near 3%, that still means over 96% of links come from other domains. Google AI Overview leans visibly on video, forums, and social sources, but it also mixes in a wide range of brand and editorial sites.
How often does ChatGPT cite Wikipedia, Reddit, and social platforms?
ChatGPT is often perceived as “Wikipedia‑heavy,” and Spotlight’s citation data supports that impression — but with useful nuance. In our last‑month window:
- Wikipedia: about 1.49% of all links
- Reddit: about 0.77%
- LinkedIn: about 0.35%
- YouTube: about 0.17%
- Facebook: about 0.12%
- Instagram: about 0.03%
These numbers show that ChatGPT does favor Wikipedia and Reddit, but still keeps each of them under 2% of overall citations. From a visibility standpoint, it is powerful to be cited by these domains, yet they are only a small part of the full link ecosystem powering ChatGPT’s answers.
How does Gemini distribute citations across social and editorial sites?
For Google’s Gemini, the picture is again a low single‑digit share for each focus domain:
- Reddit: about 0.58%
- YouTube: about 0.47% of all links
- Wikipedia: about 0.29%
Gemini’s behavior suggests a slightly stronger tilt toward reference and editorial sources (like Medium and Forbes) compared with heavy consumer social platforms. Still, even its favorite domains individually account for well under 1% of all links.
How much does Google AI Mode rely on popular platforms?
Google AI Mode is another Google surface that combines search and generative results. Over the last month, its citation percentages for our focus domains look like this:
- YouTube: about 1.84% of all links
- Instagram: about 0.88%
- Facebook: about 0.74%
- LinkedIn: about 0.52%
- Reddit: about 0.48%
- TikTok: about 0.40%
- Wikipedia: about 0.26%
Compared with Gemini, Google AI Mode leans more heavily on video and social platforms like YouTube, Facebook, and Instagram. That makes sense for consumer‑style questions, but again, no single domain comes close to dominating its citation mix.
How heavily does Perplexity cite Reddit, YouTube, and Wikipedia?
Perplexity is known for its research‑style answers and explicit citations. In our last‑month data:
- Reddit: about 4.09% of all links
- YouTube: about 2.58%
- LinkedIn: about 1.08%
- Wikipedia: about 0.85%
- Instagram: about 0.19%
- Facebook: about 0.15%
- TikTok: about 0.06%
Perplexity clearly leans harder on Reddit and YouTube than most models in this comparison. From a brand perspective, that means participating in Reddit communities and YouTube content can have an outsized impact on how Perplexity explains topics to users.
How much do Grok’s citations come from social platforms?
Grok, X’s conversational model, has some of the highest social‑platform shares in this analysis:
- Reddit: about 3.84% of all links
- YouTube: about 3.08%
- Facebook: about 1.22%
- Instagram: about 0.80%
- LinkedIn: about 0.44%
- TikTok: about 0.27%
- Wikipedia: about 0.20%
This pattern highlights Grok’s strong connection to social and user‑generated content. Brands that invest in Reddit conversations, YouTube videos, and Instagram content are more likely to be part of the sources Grok surfaces and cites.
What does this mean for brands trying to improve AI visibility?
Across models, a few themes stand out:
- No single domain dominates: Even the strongest players like Reddit, YouTube, and Wikipedia usually sit below 5% of all citations for a model. Taken together, the nine platforms in this post still account for less than 5% of all 5.8 million citations we measured.
- Social platforms matter a lot for some models: Perplexity and Grok, in particular, give noticeable weight to Reddit and YouTube.
- Editorial and reference sites stay influential: Wikipedia, Medium, and Forbes keep showing up as trusted reference points, especially for research‑style queries.
- Each model has its own “favorite mix”: Gemini leans a bit more into editorial content, Google AI Mode into social and video, and other models into different blends of social, reference, and brand content.
For brands, this means you should not focus on a single channel. Instead, think in terms of citation ecosystems: official sites, high‑quality blog content, YouTube, Reddit communities, LinkedIn thought leadership, and even appearances in trusted third‑party articles (such as Forbes or industry publications).
Tools like Spotlight (get-spotlight.com) reveal where the other 95%+ of citations come from for each model, help you track where your brand is currently mentioned, and suggest new content to create so you can appear in more of the questions your customers actually ask AI.
Frequently asked questions about AI citation percentages
How is “citation percent” for an AI model calculated?
In this analysis, citation percent is the share of all links a model cited in the last month that belong to a specific domain. For example, if a model cited 100,000 links and 3,000 of them came from YouTube, YouTube’s citation percent for that period would be 3%. Spotlight calculates this per model using raw link‑level data collected from real AI answers.
Why do Reddit and YouTube show up so often in AI citations?
Reddit and YouTube combine high topical depth with strong user engagement, which makes them attractive sources for many AI models. They cover everything from product reviews and tutorials to niche technical discussions. Models like Perplexity and Grok, which prioritize rich, up‑to‑date context, tend to reward that breadth with a higher share of citations.
Does getting cited by Wikipedia, Forbes, or LinkedIn help my brand in AI answers?
Yes, but often indirectly. When respected sites like Wikipedia, Forbes, or LinkedIn discuss your brand, those pages can become trusted reference sources that AI models cite. Even if the model does not link to your own website, being positively featured in those articles shapes the narrative users see when they ask about your product category or brand.
How can I increase my chances of being cited by AI models?
The most reliable path is to create high‑quality, credible content in places models already love to cite. That means strong, well‑structured pages on your own site, but also helpful YouTube videos, transparent Reddit participation, and expert‑driven LinkedIn or Medium posts. With Spotlight, you can see exactly which prompts, models, and sources matter most for your brand and generate new content tailored to those gaps.
Do these citation percentages change over time?
Yes. Models evolve, ranking systems shift, and user behavior changes. A domain that represents 3–4% of citations today could rise or fall as models retrain or adjust how they browse the web. That is why ongoing visibility monitoring — not just a one‑time audit — is essential if you care about long‑term AI presence.
How does Spotlight collect and analyze AI citation data?
Spotlight regularly sends large sets of real prompts to leading AI models and records their answers, including every link each model cites. It then groups those links by domain, model, and topic, and combines that with data on brand mentions and sentiment. This lets you see not only which domains models rely on, but also whether those sources are helping or hurting your brand’s perception.
This article was written by Spotlight’s content generator.