Your Google ranking used to be the proxy for visibility everywhere. If you held a top-10 position, you could reasonably assume that any system pulling from search results — including AI — would find and cite you.
That assumption just broke.
A large-scale study analyzing 863,000 keywords and 4 million AI Overview URLs found that only 38% of pages cited in Google's AI Overviews also rank in the top 10 for the same query. Seven months earlier, that overlap was 76%.
That's not a gradual erosion. It's a structural decoupling — and it changes how marketing teams need to think about discoverability.
AI platforms don't cite like search engines rank
The reason for the decoupling is straightforward once you see it: AI platforms have their own citation logic, and it doesn't mirror traditional search ranking signals.
A separate analysis of citation patterns across ChatGPT, Google AI Overviews, and Perplexity — covering over a million data points — reveals just how different each platform's sourcing behavior is:
ChatGPT leans encyclopedic. Wikipedia accounts for 47.9% of its top-10 citations. Brands receive 44.7% of total citations, but mostly as mentions without clickable links. If you're a brand hoping ChatGPT will send traffic, the data is sobering — it cites you but rarely links to you.
Perplexity is community-driven. Reddit captures 46.7% of its top-10 citations, followed by YouTube at 13.9%. Brand domains receive less than 29% of citations. Perplexity's retrieval system rewards authentic, discussion-based content over polished marketing pages.
Google AI Overviews favor brands — but selectively. Brands receive 59.8% of citations in AI Overviews, the highest of any platform. But Reddit still captures 21%, YouTube 18.8%, and Quora 14.3%. And AI Overviews only appear in roughly 33% of searches, meaning two-thirds of queries still follow traditional result patterns.
The takeaway: there is no single optimization strategy that works across all AI surfaces. Each platform has fundamentally different preferences for what it cites and how.
The technical barrier most brands don't know about
Here's the statistic that should alarm every marketing team: 73% of websites have technical barriers that prevent AI crawlers from accessing their content.
These aren't obscure issues. They're robots.txt blocks that were set up years ago to manage traditional crawlers, CDN restrictions that inadvertently block AI user agents, and JavaScript-heavy pages that AI systems can't render.
Most companies don't even know they're invisible to AI. They're investing in content strategy, keyword optimization, and thought leadership — while their technical infrastructure silently prevents AI systems from reading any of it.
The fix isn't complicated, but it requires a specific audit: check your robots.txt for AI crawler directives, test whether your pages render without JavaScript, verify that your CDN doesn't rate-limit or block AI-associated user agents, and ensure your structured data is complete and accessible.
What actually gets cited
The research also reveals what content formats AI systems prefer to cite:
Listicles dominate. Across AI Mode, ChatGPT, and Perplexity, listicles account for 21.9% of citations. Articles follow at 16.7%, and product pages at 13.7%. AI models retrieve content that is structured, scannable, and easy to extract discrete claims from.
Commercial domains (.com) account for 80.4% of all citations. Non-profit (.org) sites capture 11.3%. The newer .io and .ai domains are emerging but still small at 1.7% and 1.1% respectively.
Community platforms get disproportionate trust. Across all three platforms studied, 52.5% of citations go to community sources (Reddit, Quora, forums) versus 47.5% to brand domains. AI systems treat community discussion as a signal of authentic, unbiased information — and they cite it accordingly.
For marketers, this means your content needs to exist in two forms: the polished brand content on your own domain, and authentic presence in community spaces where AI models are actually looking.
The new discoverability framework
Traditional SEO was built on a simple chain: create content, optimize it for ranking signals, earn a position on page 1, capture traffic. AI citation works differently.
The chain now looks like this: create content that is structured for machine retrieval, ensure it's technically accessible to AI crawlers, distribute it across the platforms each AI system trusts, and format it so that key claims can be extracted at the sentence level.
Specifically:
Audit for AI crawlability, not just SEO crawlability. The bots are different. GoogleBot and GPTBot and PerplexityBot have different user agents and different rendering capabilities. Your SEO crawl audit won't catch AI-specific blocks.
Build content for sentence-level extraction. AI models don't cite pages — they cite claims. Dense, factual sentences with explicit entity relationships outperform traditional keyword-optimized paragraphs. When an AI system needs to answer a question, it looks for the sentence that most directly answers it. Make sure that sentence exists in your content.
Invest in community presence. If Reddit accounts for 46.7% of Perplexity's top citations and 21% of Google AI Overview citations, your brand's absence from those platforms is a visibility gap — not a strategic choice. This doesn't mean marketing on Reddit. It means being a genuine, helpful participant in the communities where your expertise is relevant.
Diversify content format. Different platforms prefer different formats. Google AI Overviews cite product pages and brand content more readily. ChatGPT prefers encyclopedic and authoritative sources. Perplexity trusts discussion threads. One content format won't perform across all three.
What this means strategically
The 38% number is the one that should change how you plan.
It means that even if your SEO strategy is excellent — top-10 rankings across your target keywords — you are invisible to AI systems for 62% of the queries where you should be cited.
That gap will widen. As AI models improve their retrieval and develop more sophisticated citation logic, the correlation between search ranking and AI citation will continue to decline. The systems are optimizing for different things.
Marketing teams that treat AI visibility as an extension of their SEO program will be caught in this gap. The teams that build a separate AI citation strategy — focused on crawlability, content structure, community presence, and platform-specific optimization — will capture the attention that AI-mediated search is redirecting.
The old question was: "Do we rank for this keyword?" The new question is: "Does every major AI system know we're the answer — and can it actually access our content to prove it?"
*itscool.ai helps brands build discoverability strategies that work across every AI surface — not just Google. If 73% of sites are invisible to AI crawlers, the first question is whether yours is one of them. Let's find out.*