Dead internet and the future of brand discovery

What recent shifts from OpenAI, Anthropic, and Profound reveal about how brands get discovered and why human judgment and authentic human-led storytelling matters more than ever.

Over the past few weeks, something fundamental has changed about the internet, and most marketing organizations have not caught up yet. The front door looks the same. What has changed is who is walking through it.

For two decades, marketing has been built around human attention. The scarce cognitive energy of a consumer scanning a search results page, deciding which blue link to click, processing a headline, making a purchase decision.

Everything downstream of that -- content strategy, SEO, media planning, creative development, performance measurement -- has been engineered around the habits of the human user.

That user is being replaced. Not entirely, not yet. But faster than most CMO dashboards track.

When a consumer asks ChatGPT, Claude, or Gemini a question, they are sending an AI agent to crawl the internet on their behalf. That agent uses the internet very differently than a human does.

It reads more widely. It synthesises across dozens of sources. It has no patience. And no bias to exploit. And you do not control what it reads.

Three signals from the past week. All pointing to the same shift.

James Cadwallader on Sequoia's Training Data podcast.

Cadwallader is the founder of Profound, the Generative Engine Optimisation platform backed by Sequoia. His framing is the sharpest articulation of the shift: it is not that the front door of the internet has changed, it is that the person walking through the door has changed.

When he asked ChatGPT to recommend a showerhead, it used 65 webpages to construct an answer. The old SEO rule -- 95% of value sits in the top four blue links -- no longer holds. The agent reads the long tail. Every page is in play.

And every agent reads the internet differently: Gemini leans on YouTube, ChatGPT pulls from Reddit and LinkedIn, Claude increasingly uses real-time web. Different agents, different species. Not different channels. That changes how brands get discovered.

OpenAI's advertising shift.

OpenAI announced that ChatGPT ads are moving from cost-per-view to cost-per-click, with action-based pricing -- charging when a user actually makes a purchase -- under active exploration. Their stated revenue targets: $2.4B this year, $11B by 2027. If they hit 2027's number, ChatGPT will be a larger advertising business than Snap and Pinterest combined.

Cadwallader's prediction on what an ad unit looks like in this world: a marketer will not buy media. They will prompt the campaign itself. They will target women in Minnesota between 35 and 40, when they are talking about photography mention this but not that, use this knowledge base, deliver this much revenue. That is the unit. Not a banner. Not a keyword bid. An instruction to an intelligence that has consumed the entire public record and knows the individual consumer deeply.

Anthropic's design play.

Anthropic Labs launched Claude Design alongside Claude Opus 4.7. During onboarding, Claude Design reads a client's codebase and existing design files. It extracts a complete design system -- colours, typography, components, layout. Every project afterwards uses that system by default.

The message to Figma, Adobe, and Wix was unmistakable: their stock prices dropped sharply on the rumour alone. The tools that assemble and produce creative work are being rebuilt from the ground up with AI at the core -- not bolted on. The question for any marketing organisation is no longer whether AI changes your production stack. It is which parts of your stack still make sense in 18 months.

The slippery slope, in one company

MEDVi stands out as a warning. A $1.8B telehealth marketing operation, two employees, AI-generated doctors, deepfaked patient testimonials, 5,000 Meta ads, an FDA warning letter, and a celebratory New York Times profile that had to be publicly corrected within 48 hours.

MEDVi did not start with fabricated doctors. It started with AI-written product descriptions. Then AI-generated social proof. Then AI-generated personas. Then AI-generated video testimonials. Each step was a small, defensible efficiency gain on the last. By the end, the company's entire marketing surface was a house of mirrors -- and the economics were pulling it deeper in, not out.

MEDVi is not an outlier. MEDVi is a preview.

The dead internet, and why human judgment matters more

MEDVi sits inside a larger dynamic Cadwallader named directly: if AI agents do most of the reading, comparing, and increasingly the buying -- what happens to the web that trained them in the first place?

If humans stop clicking into websites because AI answers their questions directly, the advertising revenue that funds most of the internet collapses. If that revenue collapses, the incentive to publish disappears. All first-party reporting, original research, reviews, analysis, editorial writing -- all of it -- falls toward zero. And then AI has nothing new to learn from.

Cadwallader called this the dead internet outcome, and estimated it is possible within three years. Not guaranteed. But possible.

For brands, this poses a specific problem. In a world increasingly populated by AI-generated summaries of AI-generated summaries of AI-generated content -- how does any brand say anything distinctive? If superintelligence has already read everything that exists about your category, how do you tell it something it does not know?

Cadwallader's answer: humans are a fleshy API between reality and the internet. The only marketing that will cut through is marketing grounded in original insight -- the things only your customers, your employees, your operations, and your lived experience can surface. Everything else is noise added to a signal that already exists. And noise does not rank.

This is the reason authentic, human-led content marketing is going to matter more over the next three years, not less. Not because AI tools are not useful -- they are. But because the raw material that AI tools can legitimately remix has to come from somewhere.

The brands that have access to proprietary insight -- real customer conversations, operational patterns, expert judgment, and outcomes -- are the only brands agents will have anything new to say about. They win.

A commitment, not a pitch

The future of marketing, as Firstsource sees it, is not an army of agents producing infinite content at zero cost. It is a smaller amount of genuinely insightful, authentically human-informed, beautifully produced content -- amplified, personalised, and distributed by AI at scale.

The human judgment stays irreplaceable. The human craft stays irreplaceable. What changes is everything around it.

That is the brief Firstsource's marketing services team has set for itself. Scale what is distinctive about your brand without commoditising it. Personalise at volume without slipping toward the MEDVi end of the spectrum. Use AI to amplify human ingenuity, not to replace it with a convincing imitation.

Capability access is table stakes. Anyone can buy Claude, ChatGPT, Gemini, or Profound. The harder work is everything around it: the judgment about what should be said, the governance about how it is said, the memory about who has been told what, and the discipline to stop when a shortcut starts to feel too easy.

The internet's visitor has changed. Your brand's voice does not have to.