What happens to web traffic when AI sits between people and publishers?
Opinion
If AI is to become a lasting layer in how people access information, it needs to coexist with the structures that produce that information in the first place, writes Seedtag’s Vitor Bellote.
During the last Super Bowl, an advert for Anthropic’s Claude, framed as a disagreement between AI systems, hinted that the competition for user attention is starting to centre on the interfaces people turn to first, rather than the destinations (publishers) they eventually reach. That moment has passed, but the underlying change has not.
The way people find information online has changed substantially in the past year. A question that once started with a search engine now often begins inside an AI interface. Instead of having to sort through a list of links, the response arrives instantly, shaped, summarised, and sometimes complete enough that the journey ends there. Other times, it leads somewhere else, perhaps to a publisher, perhaps to another AI. The path is less defined, and far less visible.
The question that the Claude ad initially raised remains open. If AI becomes the primary way people access information, how does the wider ecosystem continue to function?
A redistribution of attention, with uneven effects
Recent data has started to reflect what many in the industry have been sensing for some time. Referral traffic from traditional search is softening in certain areas, particularly for general informational queries. At the same time, engagement with AI tools continues to grow.
This doesn’t mean that interest in content has declined. It has simply shifted position. Early-stage discovery is increasingly absorbed before users reach a publisher site. The visit, when it happens, tends to come later and with more intent.
The effect is uneven. Large, well-known publishers still attract direct audiences, but smaller publishers, especially those relying on search visibility, are feeling the change more sharply. In some cases, content is used to inform AI-generated responses without generating a corresponding visit.
A structural imbalance is then created. Value is extracted at a single point in the journey, with limited opportunity for publishers to recover lost traffic or engagement.
There is a broader implication here. If this pattern continues without adjustment, the incentive to produce high-quality, original content begins to weaken. Over time, that affects the depth and reliability of the information available across the open web.
Why protecting publisher value is becoming central
The open web has always relied on a simple exchange. Publishers create content and audiences engage with it. Advertising supports the model that makes this possible.
AI introduces a new layer into that exchange, sitting between creation and consumption and shaping how content is accessed and experienced. That position demands responsibility.
There is growing momentum behind the idea that publishers should be able to recover value from the role their content plays in these systems. This is beginning to take shape through licensing agreements, partnerships, and frameworks that recognise the contribution of original reporting and specialist knowledge.
The direction of travel is important. A model that supports publishers not only protects individual businesses, but sustains the quality of the broader ecosystem. Without that foundation, the outputs generated by AI systems become less reliable over time.
This is particularly relevant when considering the long tail of publishing. Beyond major media brands sits a wide network of niche publishers. These are often closer to specific communities, topics, and moments of intent. They provide detail that broader coverage cannot always capture.
If these publishers lose visibility or viability, the ecosystem becomes flatter, advertisers have fewer opportunities to align with precise, high-intent environments, and users encounter a narrower range of perspectives.
A different approach to using content within AI systems
Not all approaches to AI operate in the same way. Some rely heavily on broad, unstructured data collection, which can create tension around ownership, attribution, and value exchange.
An alternative model is beginning to emerge. One built on direct relationships with publishers, in which content is used within agreed-upon frameworks rather than extracted without a clear return. In this approach, data comes from partners in the ecosystem, supported by explicit agreements that ensure value flows back rather than being diverted.
This has a subtle but important effect. It allows AI systems to operate without reducing traffic to the publishers that contribute to them. In some cases, it helps maintain visibility and creates routes back to original sources, reinforcing the role those publishers play.
It also aligns more closely with the broader need to support journalism and high-quality content. If AI is to become a lasting layer in how people access information, it needs to coexist with the structures that produce that information in the first place.
Rethinking measurement in a fragmented journey
As these dynamics play out, measurement becomes harder to pin down. The traditional markers of performance were built around a more linear web. A user clicked, arrived, engaged, and moved on.
Now, parts of that journey take place in environments that do not always pass through standard analytics. A user might ask a question, refine it, compare responses across platforms, and only then visit a site. Each step shapes the outcome, though not every step is visible.
This calls for a broader view of how attention is formed and where influence sits. It also places greater weight on environments where context and intent are clear. When the path is fragmented, the quality of the moment matters more than the volume of impressions.
For advertisers, this shifts the focus. Reaching people in trusted, content-rich environments carries more significance. These are the places where attention is more deliberate, and where messaging can connect more naturally with what the user is already engaged in.
A shared interest in keeping the web open
The current transition is about how different layers of the internet interact.
AI will continue to shape how people discover and process information. That much is clear. The challenge is ensuring that this evolution does not come at the expense of the open web that underpins it.
There is a shared interest here. Publishers need sustainable models to continue producing content. Advertisers want to be in environments that offer quality, context, and measurable outcomes. AI systems rely on a steady supply of reliable and diverse information.
For that balance to hold, the systems shaping access to information need to reinforce, rather than diminish, the publishers behind it.
Attention hasn’t disappeared. It has become more fluid, moving between interfaces, platforms, and publishers in ways that are still being understood. The task now is to shape that movement in a way that keeps value flowing back to the sources that make it possible in the first place.
Vitor Bellote is the VP of Publishers, EMEA & LATAM at Seedtag