AI search shifts media incentives toward cited expertise

As AI search systems increasingly summarize and cite online content, the center of gravity for media visibility is moving—from chasing clicks to being cited prominently. New research on AI citations finds LinkedIn and other platforms are treated as highly auth

For years, publishers built their businesses around a blunt metric: getting as many people as possible to land on a page.

Now that same bargain is being rewritten in the background. When AI search systems answer questions with summaries that cite sources, the competition is no longer only about website traffic. It’s about who gets pulled into the answer—prominently cited, neatly linked, and easy for the system to retrieve.

The fear that AI will “detour” attention away from media sites has been widespread. and the economics of scraping and compensation remain unresolved. But even before those details settle. one shift is becoming difficult to ignore: the battle for attention is slowly moving toward whose information is cited most prominently in an AI summary.

AI presence is not framed here as a replacement for website traffic. It’s described as a new proxy for relevance and authority.

A first test is underway. AI search engines have been around for well over a year, and their use is rising fast, creating enough data to begin judging how a “substance over clickbait” theory is playing out.

So far, the evidence points in that direction—though with significant caveats.

LinkedIn gets the bot bump

New studies analyzing data from millions of AI citations found consistent patterns in how AI systems treat different sources.

Research from Meltwater, a communications intelligence company, ranked LinkedIn as the second most-cited source overall in AI summaries, behind YouTube. A separate study from Semrush, a search-data analytics company, also placed LinkedIn at No. 2, closely behind Reddit.

The Meltwater data also suggested why LinkedIn may be a stronger indicator of substance than simple popularity. It found that individual members—rather than brand or company accounts—drove most of the citations.

It also found that structured content, including newsletters and posts, performed best. And more than half of the citations went to members with fewer than 10,000 followers.

Semrush’s findings added a further wrinkle: the most-cited LinkedIn posts had only modest engagement on the network itself. That is presented as evidence against a simple popularity model.

But structure can still beat substance

At the same time, there is evidence that AI systems may value structure in ways that can overpower raw factual value.

When looking at how large language models prioritize information, research connected to the Canadian AI company Cohere found that LLMs can miss key facts when an article lacks clear titles and headings.

A separate paper from Stanford University found that AI search systems strongly favor the beginning and the end of documents rather than the rest. The implication is direct: if the “meat” of a piece is highlighted only in the middle, it can often get missed.

Taken together, the reporting frames this as the core tension of the new environment. AI systems can be “gameable,” but in a different way than older search and social systems.

An article optimized for machines—featuring declarative introductions and conclusions. clear questions and answers. and consistent titling—has a pathway to being cited even if it is otherwise empty of substance. The incentive, as described here, becomes retrievable substance rather than necessarily the most insightful or information-dense content.

Making content legible to bots

This is where GEO (generative engine optimization) enters the picture. It’s described as an idea that can sometimes seem at odds with good writing, which often relies on clever titles, hooks, and the way narratives pull readers into topics.

Yet the same logic now appears practical: simply making content visible to AI engines isn’t enough. Content creators may need to “hand-hold” bots so the systems can find the good information inside.

The human edge in machine search

Even if AI citations don’t always track engagement the way social platforms do, the findings here still leave room for human influence.

Platforms like LinkedIn, YouTube, and Reddit are among the most highly ranked sources in these AI citation studies. And while the data is described as not always rewarding the most engaged posts. Semrush’s findings also indicate that frequent posting and having an established following can still help.

LinkedIn’s own internal guidance is described as aligning with that view: engagement still matters, just less directly than it did in the previous era.

There is also a subtle sense of progress in the way the story is told. Demoting raw engagement is framed as a step forward.

Still, the new route to visibility comes with a specific craft challenge. When journalists and content creators want AI systems to retrieve and cite their original reporting or insights. the reporting says the work needs to do several things: explain concepts clearly and quickly. include machine-friendly structures such as subheadings. and connect the dots by referencing other sources the AI is reading by name.

One paragraph can’t hide the trade-off

The promise of the next incentive system is that it could reward original facts, named sources, clear context, and demonstrated expertise more than outrage and raw engagement. The caution is that every new system also invites its own bad habits.

The winners in an AI-mediated future, the reporting suggests, shouldn’t simply be the loudest accounts or the best-formatted posts. They should be people who know something real and can demonstrate its worth to the right audience—both human and machine.

AI search media incentives AI citations LinkedIn Meltwater Semrush Reddit YouTube generative engine optimization GEO LLMs content structure journalistic credibility