Studies suggest LLMs favour earned media far more than any other sources.

This is welcome news for PR practitioners after a troubled few years. Elon Musk famously boasted that Tesla had no use for a PR team. After acquiring X, he urged everyone to become their own journalist. Meanwhile, a16z, after clashing with several tech journalists, encouraged startup founders to bypass traditional media altogether, and rely solely on social channels and owned platforms. PR was stuck defending its very existence, but now AI has finally put the whole discussion to rest. By prioritising earned media, AI revalidates and reaffirms PR’s essential status in the post-click world. Long live generative engine optimisation (GEO)!

Much of the discussion around GEO overlooks other important changes AI has wrought on media and PR. Considering all the new technical changes is undeniably important, but so is emphasising the continued importance of human relationships in earned media. Any PR pros that prioritise AI-friendly content while neglecting their ties with journalists, editors, and other human gatekeepers at media outlets risk losing sight of the bigger picture.

GEO may well be the new SEO. Like SEO before it, GEO’s goal is to maximise brand visibility. The mechanics, however, are different. GEO does not prioritise technical factors such as keyword density or backlinks, as SEO does. Unlike traditional Google searches, which separated news into distinct sections, earned media carries the greatest weight with LLMs. And earned media can't be gamed like traditional SEO. It still rests on relationships and credibility. Human-to-human trust, and what journalists find interesting, relevant, or newsworthy cannot be automated. It requires emotional intelligence, and a strong contextual and social understanding of reality. This is what makes PR instrumental to this shift.

Legacy media adapts to AI reality

Previously, Google algorithm changes meant enormous headaches for media outlets and blogs and dictated how headlines and websites were structured. Google was king. Now, AI is aiming for the throne, and its momentum suggests it’s only a matter of time. People increasingly prefer to search, find, and consume information without sifting through multiple websites. But if no one visits websites, how does the media monetise traffic? Short answer: they don’t. Welcome to the “post click era”. Even Google has adjusted to these new preferences by launching AI-powered results, reducing clicks and traffic to blogs and publications even further.

According to Press Gazette, since Google launched AI Overviews in May 2024, 42 out of the top 50 US websites by traffic saw traffic fall, and more than half of them saw traffic fall by more than 25%. Their data also shows that Forbes traffic fell by 52%. Other sources claim traffic declines in the ranges of 25-30% for the likes of The New York Post, CNN, and The Daily Mail, and 30-55% for Huffpost, The Sun, and Business Insider. These are shocking numbers.

This isn't the death blow to legacy media that some people imagine. From a business perspective, they are understanding and adapting to the situation better than they did with the onset of digital media. One thing that plays to their favour is the value of their content as determined by LLMs. Their content is key to AI companies for delivering up-to-date, trustworthy information to its users. The PR tech platform Muck Rack analysed over one million AI citations and found that when AI systems provide source links in their responses, 85% point to earned media, with outlets like Reuters, Axios, Time, Forbes, and the Financial Times appearing repeatedly.

Media companies are responding to this in different ways. Some are cutting deals with AI companies. The Financial Times, The Atlantic, and Vox Media signed deals with OpenAI. The New York Times did so with Amazon, for $20 million, despite strong growth with subscriptions and ads. Others are suing for copyright infringement or blocking access altogether, while some are still deciding. Not everyone is willing or able to make deals with the AI behemoths. However, most recognise their all-in relationship with traffic is over, and they're adjusting accordingly to avoid extinction.

These adjustments send the media industry in the opposite direction versus when they went from print to online and made everything available for free. They include live events, newsletters, video, and podcasts among publications that previously relied on article volume or clickbait. This strategy shores up valuable audience loyalty by creating strong relationships with readers, whom they are studying and understanding better to serve better, and keeps AI from farming everything immediately. It also enables them to set up paywalls and price increases. Publications from lifestyle brand Man of Many to Harvard Business Review are increasing prices. The former’s strategy is shifting from a reach of 2 million to a closer relationship with 100,000 people.

In this new era, it's all about the monetisation of value as opposed to the monetisation of traffic. Brian Morrissey, of the popular newsletter The Rebooting calls this a “total monetisation approach to make more money out of fewer visitors to their properties.”

Nicholas Thompson, CEO, The Atlantic, recently said in a podcast that the outcome he wants is not major payouts from lawsuits against AI companies, but to “work hand in glove and come up with a compensatory system that’s fair for everybody.” And new deals might be more flexible and easier to pursue for media companies of all sizes. Perplexity’s recent announcement of its new revenue-share for publishers could set a collaborative precedent.

Media is going multi-channel and PR must respond accordingly. If the media is betting on cultivating closer ties with audiences by providing greater and more diverse value, then that media is surely worth tapping, regardless of appearing on AI search or not. On the other hand, PR pros will need to understand how to help build reputations on LLMs, and that means understanding which publications partner with which AI companies.

Understanding GEO in practice

The stakes for LLM visibility are high. Decision-makers increasingly turn to ChatGPT, Claude, and Perplexity instead of Google throughout their research and buying journey. Absence from AI responses means missing critical touchpoints where choices are made.

However, PR pros need realistic expectations about what they can control. The Muck Rack study revealed that while 85% of results come from earned media, only 5% come from press releases on wires or company sites. This isn't a game you can win by gaming or flooding the system.

This data reveals a critical strategic priority: it's smarter to secure earned media coverage to begin with than to obsess over structuring press releases for AI consumption. Even perfectly optimised releases rarely get directly cited. AI systems prioritise the resulting journalistic coverage.

According to the Muck Rack study, AI visibility depends on three critical factors: recency (fresh content, especially for topical queries), query framing (advice-seeking prompts trigger 49% journalism citations vs 27% overall), and outlet authority (though performance varies significantly by industry and model).

AI systems value both mainstream credibility and subject-matter expertise. Industry expertise trumps general authority. While Reuters and Associated Press dominate overall citations, AI systems consistently reward niche-specific content. Only 15% of top sources appear across multiple industries. This means PR pros should prioritise authoritative trade publications and industry-specific outlets alongside mainstream media. This aligns with longstanding PR best practices and proves particularly valuable for startups, who are often too small to secure recurring coverage in top-tier publications.

Query types also fundamentally alter citation patterns. Advice-seeking and opinion-based prompts trigger dynamic citations from recent journalism, while encyclopaedic queries fall back on static training data like Wikipedia. Understanding how your target audience phrases questions becomes crucial for optimisation strategy.

Also noteworthy is that different AI models show dramatically different source preferences. Claude cites journalism 20x less than Gemini and 50x less than ChatGPT, while OpenAI models favour content from the last 12 months (56% vs only 36% for Anthropic).

For PR pros, these findings validate existing best practices and the wisdom of relationship-focused strategies. However, PR pros shouldn't assume these citation patterns will remain static. Just as Google's algorithm updates once sent SEO specialists scrambling, LLM preferences are already evolving. It’s a safe bet that AI models will continuously update their source selection criteria. The aforementioned ongoing and tentative alliances between media and AI companies are in flux. Other disruptive events are bound to take place. Grok recently announced plans to introduce ads into AI responses.

The same vigilance that marketing pros developed for Google algorithm changes will be needed for LLMs. As well as tracking the latest trends, companies will need to track their own visibility within LLMs in different situations. Research and close contact with clients will be key to identifying the nature of these situations, and what questions would-be readers will ask LLMs.

For startups seeking more direct and measurable LLM optimisation, owned content channels like blogs and social media remain better avenues for that technical approach, where you control both messaging and format. In case it wasn’t already clear, you shouldn’t limit GEO considerations to PR. Rather, every marketing function should understand how GEO applies within its remit.

The rise of independent voices

Traditional media adaptation is only part of the story. While legacy media adapts, another significant change is happening outside traditional newsrooms. Over the last decade, legacy media has steadily lost its monopoly on being what Nick van Osdol defines as "consensus intermediaries," a concept he introduced in his newsletter, Keep Cool. These are trusted sources who help shape what groups believe to be true. Fragmentation, atomisation, and audience siloing have redistributed that role, shifting credibility toward smaller outlets and individual voices. Some have journalistic backgrounds, others don’t.

This shift stems from tech platforms and tools that have made it easier to build dedicated followings, create content, and distribute it at scale without relying on legacy channels. More recently, tools like Substack, Patreon, and beehiiv have added a critical layer: monetisation. Creators can now earn sustainable income without needing massive audiences, making niche reach financially viable.

This shift has allowed both niche publications and experienced journalists, but also content creators outside of journalism, to build profitable, self-sustaining media businesses. With the right communication skills, increasingly powerful tech tools can handle much of the heavy lifting. And it’s only getting easier as these tools integrate AI, or new AI-first tools emerge. A decade of newsroom layoffs has simultaneously created a pool of experienced journalists with loyal audiences and deep expertise, many of whom are now monetising their knowledge directly through platforms like Substack. This is a dynamic combination.

In practice, this shift can be seen in the success of independent journalist Ken Klippenstein, now a go-to source on corporate and government whistleblowing, or 404 Media, which learned from VICE's missteps and recently broke the Tea App hack story. When Digital Frontier shut down, several of its staff launched Substacks to continue covering the same beats. Conversely, Resilience Media started on Substack and is now growing beyond it, having hired several senior journalists laid off by TechCrunch. Readers will follow trusted journalists who leave or are pushed out of traditional newsrooms.

Startups shouldn’t ignore new consensus intermediaries simply because their names don’t appear in the top ranking of AI citations. Due to their specialised nature, they occur less frequently, but in the cases they are relevant, they may be more relevant than any other sources. PR experts are now highlighting how niche publishers, podcasts and other ‘specialists’ are able to drive more retrievability within AI searches than top tier media.

How PR professionals are adapting

While AI reshapes the media landscape, PR professionals are simultaneously integrating these tools into their daily workflows. The reality, however, may be more modest than the industry hype suggests.

Industry observers describe a significant gap between perception and reality regarding current capabilities. Applications remain largely tactical, concentrated in content creation, keyword generation, and basic image production. A recent PRWeek survey of UK consultancies found that three-quarters use AI primarily for research, with one-third applying it to creative campaigns and others to administrative tasks. Very few use it for  sophisticated strategic functions like complex research, multi-agent workflows, or decision-support systems.

Interestingly, even AI companies approach their own PR with caution. Google's Gemini PR team reportedly avoids using AI for most of their work, believing certain communications functions are too sensitive for automation, too holy for humans not to be in total control. This perspective from AI developers themselves suggests the technology's limitations in strategic communication contexts.

The industry also faces growing concerns about AI's potential negative impacts, especially increased spam and decreased authenticity within PR-journalist relationships. In May, the Wall Street Journal’s Patrick Coffee wrote a piece on AI and the PR industry. He discovered how major PR agencies are leveraging AI and even creating their own tools, but also how AI is being used to create artificial "experts" designed to trick journalists and boost SEO rankings. One of Coffee’s PR expert sources concludes that comms "relies more heavily on the nuances of one-on-one engagement between real humans" and that automation risks interfering with authentic relationship-building.

In July, Press Gazette wrote about an AI tool that would send automated expert comments to journalists sending requests to HARO, using the “journorequest” hashtag. One PR expert’s take: if journalists wanted AI-generated answers, they could ask ChatGPT themselves.

Even content generation has its limitations. No editor at a respected publication wants AI-generated content even if the content reads well. Publications will remove articles if they identify them as AI-generated, as recently demonstrated by WIRED and Business Insider.

Not all AI tooling is conducive to spam. There are great AI-powered tools popping up that handle one or two more PR tasks better than before. Current limitations don't indicate a hard ceiling. As AI becomes more sophisticated and agentic, it may eventually penetrate more strategic areas of PR. For now, though, the tools enhance efficiency while reinforcing the centrality of human-led relationship-building, cultural understanding, and strategic thinking.

PR's strategic response

AI has fundamentally changed how people discover information and make decisions, elevating PR's importance as earned media directly influences what LLMs tell users. Yet the fundamentals remain unchanged: journalists control the final messaging, making relationship-building as crucial as ever.

At the same time, mass layoffs at traditional media have created a paradox. These outlets are both harder to reach (there are six PRs for every journalist, according to some estimates) and now more valuable than before with LLMs frequently citing their content. Meanwhile, the same layoffs, and new tools available to those laid off, have sparked an explosion of newsletters and independent media that provide genuine value to niche, engaged audiences, and may outperform top tiers in AI search in certain situations.

The transition of independent voices into leading consensus intermediaries continues the long-standing need to identify and cultivate relationships with emerging platforms, whether they're career journalists or industry experts building audiences on Substack. Similarly, legacy media's expansion into live events, newsletters, video content, and premium offerings represents new opportunities (and obligations) for PR pros. These trends demand that PR pros continue to deliver quality while widening their scope.

While certain tactics can be tweaked for AI visibility, they don't make or break earned media success. The earned media that LLMs prioritise as trusted sources remains the same high-quality coverage PR has always pursued. For direct LLM optimisation, owned content channels offer more control and potential.

The future of PR in the era of AI is, as it turns out, more human than ever.

For more startup news, check out the other articles on the website, and subscribe to the magazine for free. Listen to The Cereal Entrepreneur podcast for more interviews with entrepreneurs and big-hitters in the startup ecosystem.