AI Answers Are Killing Publishers
Here’s what’s happening: Your content appears in an AI answer box, the user gets a snappy reply—and never leaves Google. You don’t see the click-through. No pageview. No ad impression. No subscription signal. And no way to prove your piece was worth writing.
It sounds dramatic, but it isn’t speculation. A field experiment by Saharsh Agarwal and Ananya Sen, published in a 2026 paper titled The Impact of Google AI Overviews on Publisher Traffic and User Experience, recorded a 39.8% drop in outbound organic clicks and a 34.5% rise in zero-click searches when AI overviews appeared—without increasing overall search frequency or sponsored traffic.
That’s the headline. What matters, though, is what collapses with that hit to traffic: the whole open web bargain.
The Broken Bargain
Alex Chan, assistant professor at Harvard Business School, spells it out cleanly: the open web runs on a compact. Publishers create content; search engines and social platforms send users over; visits generate revenue—and, critically, they also generate informational quality signals.
Those signals—clicks, subscriptions, bookmarks, backlinks, time on page, return visits—are how platforms and readers figure out what deserves attention. They’re the invisible currency that ranks your piece above a vague competitor’s, invites a curious reader back, and prompts search engines to surface you again for related queries.
Generative AI breaks that compact. When a user asks how to audit an Office 365 compliance report, an AI answer might summarize your three-thousand-word deep-dive directly on the results page. The user walks away satisfied. You walk away with nothing.
Chan puts it bluntly: “An AI answer can use publisher content while keeping the user in the AI interface. Users may be better served in the short run. But the source may lose the visit, the revenue from the visit, and the source-level signal that the visit would have produced.”
You didn’t lose a customer. You lost an audit trail.
The Hidden Cost: Durable Attention Capital
It’s easy to fixate on lost ad impressions. But the real bleed is what Chan calls durable attention capital—the cumulative value that builds when readers stick around:
- Repeat visitors who become subscribers or paid users
- Backlinks that anchor your domain authority for years
- Bookmarks and browser history that surface you on future, unsolicited searches
- Brand familiarity—when readers recognize your logo and trust it implicitly
A one-off AI answer might be useful. But when every high-intent query gets answered without sending the user anywhere else, those durable signals atrophy. And once they do, even a perfect SEO score won’t bring them back.
Think of it like compound interest in reverse. Without daily deposits, your site’s authority compounds down to zero.
Why Band-Aid Fixes Don’t Work
The knee-jerk responses tend to fall into two buckets:
- Visitor replacement royalties—AI platforms pay publishers per skipped click, like a licensing fee.
- AI answer bans—simply block overviews on sensitive topics or certain domains.
Neither restores the ecosystem. Royalties treat the symptom (lost clicks) without addressing the root (broken signal flow). They also create alignment problems: who verifies how many clicks were truly diverted? Should a generic answer to “what is cloud security incident response” pay the same as one quoting your six-step playbook?
Bans, meanwhile, punt to regulation. They assume Google or Anthropic will suddenly stop optimizing engagement—and that regulators can keep up with model upgrades and market shifts. That’s a losing bet.
Chan is clear: “When an AI platform diverts the revenue and measurement events without replacing them, costly human information may fall below replacement.” That’s not a pricing problem. It’s a measurement crisis.
What Actually Needs Fixing
Chan’s fix isn’t about slowing AI. It’s about sharpening provenance.
He proposes four concrete levers:
- Provenance tagging—explicit signals about which pieces contributed to an AI summary, with a persistent link back.
- Diversity pricing—differing prices or prominence for AI answers based on whether they draw from multiple sources (high credibility) or a single piece (higher risk of hallucination).
- Exploration credits—incentives for users to click through and verify the AI’s answer, perhaps redeemable toward premium subscriptions.
- Informative audits—third-party verification that an AI answer accurately represents the original and fairly allocates attribution value.
None of this is trivial. Google already resists “provenance” as a technical burden, and platform owners are right to worry about user experience. But the alternative—watching a generation of publishers scale back security and compliance coverage because the business model no longer supports it—is worse.
Industry Playbooks Already Shifting
The quiet war has started:
- Cloudflare now blocks AI scraping bots on ad-supported pages, reasoning that automated harvesting without compensation erodes the ecosystem and exposes sites to additional attack surface.
- Google CEO Sundar Pichai disputes the “collapse” narrative, citing Pew Research and internal engagement metrics that suggest users remain loyal to destination sites.
- Analysts, citing the same field study Agarwal and Sen conducted, point to declining organic referral traffic from security topics—especially long-form guides and technical checklists—as early warning.
Here’s the tension, plain and simple: AI answers do help some users find faster. But they hurt the ecosystem that produces those answers. The open web has thrived on a flywheel: users visit, signal quality, creators get funded, more content ships. Break the signal—and you break the flywheel.
The Security Angle: Why Compliance Teams Should Care
Security & Compliance isn’t immune. In fact, it’s ground zero.
When a security analyst searches for “cloud security incident response playbook” or “security & compliance center office 365,” they often want a framework, not just an answer. They’ll cross-check multiple sources, bookmark templates, and return for the next update—because high-stakes environments demand repeatability and traceability.
AI overviews can summarise the first playbook. But they can’t point you to the one where the author included a 27-point post-mortem template that survived SOC 2 scrutiny last quarter. That’s the signal AI strips away.
Until platforms commit to richer provenance, compliance teams risk losing their most reliable sources of nuanced, audit-ready guidance—not because the content vanished, but because the path to it got quietly erased.
One Last Note on The Register’s Context
The Register’s 2026 coverage (where this article draws most of its facts) isn’t just background noise. Senior reporter Thomas Claburn nails the stakes: “Provenance, diversity prices, exploration credits, and informative audits are needed to restore the broader ecosystem: quantity, quality, diversity, source-level signals, and the conventional-search discovery channel that prevents self-reinforcing migration into AI answers.”
That’s a lot to unpack, but the core idea is clear: AI search isn’t replacing discovery. It’s replacing the ecosystem that discovers.
The fix won’t come from better AI. It’ll come from rebuilding the market to reward the human time that wrote the original answers in the first place.
