What happens to AEO when the sources AI needs can no longer afford to publish?

There is a contradiction at the heart of AI search. Answer engines need credible information: journalism, expert analysis, research, reviews, original reporting and authoritative websites from which to construct useful answers. Yet those same systems can increasingly consume that information, summarise it and satisfy a user's question without sending the user to the source that created it. That creates an uncomfortable question for the emerging AEO industry: what happens when the sources AI needs can no longer afford to produce the information AI needs?

<h2>The click problem is becoming measurable</h2>

The click problem is no longer merely theoretical. A recent Pew Research Center study of browsing behaviour found that users clicked a traditional search-result link on 8% of visits where a Google AI Overview appeared, compared with 15% of visits without one. The same study found that only about 1% of visits to pages with an AI Overview resulted in a click to one of the cited sources within the overview. The study used observational data, so it cannot establish that AI Overviews caused the differences. But the association is clear enough to sharpen a fundamental distinction for AEO: being cited and being visited are becoming two different outcomes.

<h2>Publishers are already feeling the pressure</h2>

The commercial pressure on publishers is visible across multiple markets. The Wall Street Journal reported that People Inc.'s traffic from Google decreased 40% year on year in the second quarter of 2026. Reporting on wider publisher responses, Nieman Lab noted that some publications saw search-traffic declines of more than 40% between June 2025 and June 2026, according to Semrush data, while others have begun considering whether continuing to grant Google crawler access remains commercially viable. Those results will vary by publisher, audience and content mix, but the broader direction is difficult to ignore.

In France, the Alliance of General Information Press asked the country's competition authority to act over Google's AI-generated article summaries on August 11. Reuters reported that the trade group argued the summaries deprive newspapers and magazines of readers, while communications regulator Arcom estimated that AI-generated summaries had contributed to traffic declines of between 33% and 38% for the publishers' sites. The group also argued that the new use of press content had been deployed without prior authorisation or dedicated remuneration.

<h2>The evidence supply chain</h2>

AI search can be understood as an information supply chain: original information, source, retrieval, AI interpretation, answer, consumer. AEO has largely concentrated on the middle of that chain, asking how brands become part of what gets retrieved and interpreted. But the whole system depends on the first two stages continuing to function. Someone still has to conduct the interview, test the product, analyse the clinical trial, investigate the company, compile the statistics, write the review and verify the claim. AI can synthesise that information efficiently. It cannot sustainably replace original information when nobody has an economic incentive to create it.

<h2>Why this is an AEO issue, not just a publisher issue</h2>

Most AEO conversations begin with the brand: how to get cited, which sources influence an answer, where competitors appear, and what content should be created. The source itself is another participant in that ecosystem. If authoritative publishers restrict crawlers, move more material behind paywalls, demand licensing fees or reduce the production of original reporting, the information environment available to answer engines changes. That affects every brand trying to optimise within it. A model trained or grounded in a thinner, more fragmented evidence base will produce different answers from one with broad access to current, independently reported information.

<h2>Licensing may become part of the infrastructure</h2>

One likely response is already emerging. Publishers may increasingly treat access to their information as something AI companies must license rather than freely retrieve. Others may deploy crawler controls, restrict access, or develop structured commercial feeds for AI systems. At scale, that could divide the source environment between open, licensed, restricted and proprietary information. For AEO, the question would no longer simply be which sources a platform cites. It would become which sources each platform can access at all. The answer may differ materially among ChatGPT, Gemini, Claude, Perplexity and future agentic search products.

<h2>Citation value needs to be rethought</h2>

AEO platforms increasingly treat citation frequency as a core performance indicator. The publisher-economics problem shows why that is incomplete. A source can be frequently retrieved, frequently cited, highly influential in shaping an answer, rarely clicked and economically uncompensated. Those are not interchangeable states. As the category matures, citation measurement should distinguish between discovery value for the publisher, evidentiary value for the answer engine, and commercial value for the brand that is represented. Counting citations without understanding those roles risks overstating what a citation actually means.

<h2>AEO Updates Takeaway</h2>

AI search needs high-quality sources, and high-quality sources need sustainable economics. AI answers can reduce the traffic that historically helped fund those sources. That is the paradox. The publisher-versus-AI dispute is therefore not a side issue for AEO. It is an infrastructure issue, because the long-term quality of answer engines depends on the long-term quality of the evidence environment beneath them. Brands should continue to improve how they are represented in that environment, but the industry also needs to confront the question of who will keep funding the information that AI systems depend on.

<h3>References</h3>

[1] Pew Research Center authors, <a href='https://arxiv.org/pdf/2608.04831' target='_blank' rel='noopener noreferrer'>Investigating Click Behaviors On Google Search Result Pages That Produce an AI Overview</a>, August 2026.

[2] Reuters, <a href='https://www.reuters.com/world/french-media-asks-french-anti-trust-watchdog-act-googles-ai-2026-08-11/' target='_blank' rel='noopener noreferrer'>French press body asks competition watchdog to take action over Google AI</a>, August 11, 2026.

[3] Nieman Lab, <a href='https://www.niemanlab.org/2026/07/search-traffic-has-declined-so-much-that-some-publishers-are-considering-opting-out-of-google-entirely/' target='_blank' rel='noopener noreferrer'>Search traffic has declined so much that some publishers are considering opting out of Google entirely</a>, July 22, 2026.