I spent the better part of last decade perfecting on-page optimization. Title tags, meta descriptions, header hierarchies, keyword density, internal linking structures. The stuff that made pages rank. I was good at it. And now I am watching a fundamental shift happen that renders much of that playbook secondary at best.
Here is the uncomfortable truth that the data is screaming: when it comes to AI search, what others say about you matters dramatically more than what you say about yourself. The signals that drive traditional rankings are not the signals that drive AI citations. And if you are still focused primarily on your own website when building visibility strategy, you are optimizing for a game that is already being replaced.
The Numbers That Forced Me to Rethink Everything
Muck Rack analyzed over one million AI prompts recently. The finding that stopped me cold: 85.5% of AI citations come from earned media sources. Not brand websites. Not owned content. Third-party coverage.
Let me sit with that for a moment. We are not talking about a slight preference here. We are talking about earned media being cited roughly 5x more frequently than brand-owned content by AI engines.
BrightEdge data adds another dimension. Their analysis shows that 34% of AI citations come from PR-driven coverage, with an additional 10% from social channels. That is nearly half of all AI citations coming from off-site sources that most SEO teams barely touch.
The implications are significant. You could have the most perfectly optimized website in your industry, with impeccable technical SEO, flawless content structure, and keyword coverage that would make any traditional SEO strategist weep with joy. And AI systems might still barely acknowledge you exist, because they are drawing their answers from what trade publications, journalists, analysts, and industry voices are saying about the space.
Understanding Why AI Systems Think Differently
Traditional search engines were designed to crawl, index, and rank web pages. The unit of optimization was always the page. Make the page better, and it ranks better. Simple causality.
AI systems do not think in pages. They think in knowledge. They are trained on massive datasets that include the entire observable internet, and they form understanding based on patterns of consensus, authority, and corroboration across sources.
When a traditional search engine encounters your product page, it asks: "Is this page relevant and authoritative for this query?"
When an AI system encounters a question about your product category, it asks something more like: "What does the collective internet say about solutions in this space? Who gets mentioned? What consensus exists about capabilities, features, and market positioning?"
This is a fundamentally different question. And it demands a fundamentally different optimization strategy.
The AI does not see your perfect product page as the primary source of truth. It sees your product page as one voice among many, and probably a biased one at that. Of course your website says your product is great. What does everyone else say?
The Corroboration Problem
Think about how humans evaluate claims. If someone walks up to you and says "I am the best accountant in the city," you are skeptical. But if three different business owners independently tell you they use the same accountant and she is fantastic, you take notice.
AI systems have internalized this same skepticism toward self-promotion. They are pattern matchers at their core, and one of the patterns they have learned is that claims made by the subject of those claims are less reliable than claims made by third parties.
This creates what I call the corroboration problem. You can optimize your website until it is technically perfect, but if the AI training data does not include multiple independent sources confirming or discussing your brand in relevant contexts, your on-site optimization is screaming into a void.
The training data already contains whatever it contains. Your website is one data point. The collective internet discussion of your brand, your competitors, and your industry constitutes thousands of data points. Which signal do you think carries more weight in the aggregate model?
What AI Models Actually See
Here is something that confuses people: optimizing for AI is not the same as optimizing for search engines, even when those search engines now include AI features.
Google AI Overviews pull from a combination of indexed pages and generative synthesis. But the underlying LLMs that power ChatGPT, Claude, Perplexity, and others were trained on fixed datasets. What they "know" about your brand was determined by what was in those training datasets.
This means there are two distinct visibility challenges now. The first is real-time AI features in search engines, which can be influenced by recent content and crawling. The second is foundational AI model knowledge, which is determined by training data that is already fixed.
For the second category, what matters is what was being said about your brand when the model was trained. If you were not getting mentioned in the conversations, publications, and contexts that made it into training data, the model simply does not know much about you. No amount of on-page optimization after the fact changes that.
The practical implication: you need to be consistently present in the conversations that matter, not just when you are launching something. Always-on visibility in earned media is what gets you into training datasets. Sporadic coverage misses windows.
The New Hierarchy of Signals
Based on what the research shows and what I have observed working with brands trying to crack AI visibility, here is how I would rank the signal hierarchy for LLM citation likelihood.
At the top sits earned coverage in high-authority publications. When TechCrunch, industry trade publications, or respected analysts write about you, that content carries enormous weight. It is in the training data. It is corroborated by the publication's own authority. It is the exact signal AI systems are designed to trust.
Just below that comes expert commentary and quotes. When journalists quote your executives as sources in their coverage, your brand gets associated with expertise and authority on specific topics. This contextual association is powerful. The AI learns that your brand is referenced when experts discuss specific domains.
Third in the hierarchy is independent reviews and analysis. When third parties test, compare, and analyze your products or services, that analysis creates citation-worthy content that AI systems treat as more objective than your own claims.
Fourth is social proof at scale. When your brand is discussed organically across forums, social platforms, and community spaces, that creates a volume of contextual mentions that shape AI understanding.
And only fifth, after all of that, comes your owned content. Your website, your blog, your documentation. It matters, certainly. But it is not the primary driver of AI citation behavior.
Why Traditional SEO Still Matters (Just Differently)
I want to be clear that I am not saying technical SEO is dead or that you should neglect your website. That would be overcorrecting in a dangerous way.
Traditional SEO remains foundational for multiple reasons. Google still dominates search volume, and rankings still drive traffic. Your website needs to convert the traffic that comes, which means on-page optimization still matters for user experience. And AI systems that do real-time crawling will still evaluate your content directly in some contexts.
But the relationship between activities has changed. On-page optimization used to be the primary game, with off-site work supporting it through link building. Now the balance is flipping. Off-site reputation work feeds directly into AI visibility, with on-page optimization serving as the foundation that supports everything else.
Think of it like building a house. Traditional SEO is the foundation. You need it. But earned media is the house itself. The foundation without the house is not somewhere you can live.
The Practical Shift in Resource Allocation
Most marketing teams I talk to allocate maybe 10-15% of their SEO budget to digital PR and earned media activities. The rest goes to content creation, technical optimization, and link building.
Based on where the signals are moving, I would argue that allocation should be closer to inverted. If AI citation is becoming a primary visibility channel alongside traditional rankings, earned media deserves more than a supporting role.
This does not mean abandoning content creation. It means rethinking the purpose of content creation. The question is not just "will this rank?" The question becomes "will this get covered, discussed, cited, and referenced by others?" If you create something worth talking about, others will talk about it. If you create pure SEO content optimized for keywords but not for newsworthiness, you are playing only half the game.
The practical implication is that every piece of strategic content should have a PR angle. Why would a journalist care? What is the newsworthy element? What makes this discussable? If you cannot answer those questions, you might rank, but you probably will not get cited by AI systems.
Conversational Query Coverage vs. Head Terms
Here is another dimension where the shift matters: the nature of queries themselves.
Traditional SEO optimized for head terms. High-volume keywords. "Best CRM software." "Project management tools." "Marketing automation platform."
AI search skews toward conversational queries. People do not ask AI for keywords. They ask questions, describe situations, seek advice. "I am a small business owner who needs to keep track of customer conversations and follow up on sales opportunities. What should I use?"
This query type is longer, more contextual, and more specific. And AI answers to these queries tend to draw on sources that discuss the context, not just the keywords. A comparison article that discusses CRM options in the context of small business needs is more likely to be cited than a perfectly optimized product page that hits all the keywords but does not address the situational context.
Earned media coverage naturally tends toward this contextual, situational framing. Journalists tell stories. They explain use cases. They describe who benefits and why. That narrative structure maps better to conversational queries than typical on-page SEO content does.
The Always-On Imperative
One of the biggest mindset shifts is moving from campaign-based PR to always-on presence.
The traditional model was to do PR around launches, funding announcements, major milestones. Create a spike of coverage, then go quiet until the next announcement.
The problem is that AI training happens on a continuous timeline that does not care about your announcement calendar. If you are only visible during launch windows, you are missing massive stretches of time when training data is being assembled. And once a model is trained, your launch announcement from last month does not matter. You are not in that model's knowledge base.
What you need is consistent presence. Fresh mentions in high-authority sources on an ongoing basis. Not because any single mention is going to transform your AI visibility, but because the accumulation of mentions across time creates the density of signal that AI systems need to form strong associations.
This requires a fundamentally different relationship with PR. Not campaigns, but continuous generation of newsworthy angles. Ongoing executive commentary. Regular contributions to industry conversations. A steady drumbeat of reasons for third parties to discuss your brand.
Trust as the Underlying Currency
Underneath all of this is a single concept: trust. AI systems are fundamentally trust-assessment machines. They are trained to distinguish reliable information from unreliable information, authoritative sources from questionable ones, corroborated claims from unverified assertions.
Earned media is trusted because it passed through an editorial filter. Someone who was not paid by you decided your brand was worth discussing. That editorial validation is a trust signal that no amount of owned content can replicate.
This is why the shift toward earned media is not a temporary blip or a gaming opportunity. It reflects something deeper about how AI systems are designed to function. They are built to weight independent corroboration over self-promotion. That is not a bug to work around. It is a feature that shapes the entire landscape.
Brands that understand this will invest in becoming genuinely newsworthy, genuinely expert, genuinely worth discussing. Brands that do not understand it will keep optimizing pages and wondering why AI systems barely know they exist.
What This Means for Different Business Types
The implications vary by business type and market position.
For established brands with existing media relationships, this is mostly good news. You probably already have a foundation of earned coverage. The task is to increase volume and consistency, ensuring continuous presence rather than sporadic coverage.
For newer or smaller brands, this is more challenging. Breaking into earned media is harder than publishing blog posts. It requires actual news, actual expertise, actual differentiation worth writing about. You cannot buy your way into training data the way you could buy links in the old paradigm.
For B2B companies, the dynamics favor thought leadership. Industry publications are always looking for expert perspectives. Technical expertise translates into quotable commentary. The path to earned coverage goes through demonstrating genuine authority.
For consumer brands, the emphasis shifts toward making your products and customer experiences interesting enough to discuss. What is the story? What would make someone write about you? If you cannot answer that, you have a brand problem, not just an SEO problem.
The Integration of PR and SEO
Historically, PR and SEO have operated as separate disciplines with occasional overlap. PR teams focused on relationships and coverage. SEO teams focused on rankings and traffic. They might coordinate on link building, but their goals and metrics were distinct.
That separation is becoming untenable. The same coverage that builds brand reputation now directly drives AI visibility. The same on-page optimization that used to be the whole game now serves as infrastructure for off-site reputation efforts. The disciplines have to integrate.
What this looks like practically: SEO teams need to understand media dynamics and newsworthiness. PR teams need to understand how AI systems parse and weight coverage. Joint planning needs to happen, with shared metrics that capture both traditional rankings and AI citation signals.
The teams that figure out this integration fastest will have significant advantages. Most organizations still have these functions siloed. The opportunity is there for those willing to bridge the gap.
The Uncomfortable Conclusion
I have been doing SEO work for a long time, and there is something uncomfortable about writing this. A lot of what I have mastered is becoming secondary. On-page optimization, keyword research, technical audits. All still valuable. All still necessary. But no longer sufficient, and no longer primary.
The game has changed. AI search is not traditional search with a fresh coat of paint. It is a fundamentally different paradigm that weights different signals and rewards different behaviors.
Earned media is the new SEO. Third-party citations beat on-page optimization for AI visibility. What others say about you matters more than what you say about yourself.
The sooner you restructure your approach around these realities, the better positioned you will be. The data is clear. The shift is happening. The only question is whether you will adapt your strategy accordingly or keep optimizing for a game that is being replaced.
I know what I am doing differently. The question is whether you will too.