August 20, 2026
TechCrunch reported on a new study finding that roughly a third of web pages published since ChatGPT's launch show signs of AI authorship. That's not a niche corner of the internet -- that's a third of everything new. For years the debate was whether AI-written content would be 'good enough.' That debate is basically over. The real question now is what happens to a web economy built on the assumption that content is scarce and expensive to produce, when suddenly it isn't.
For businesses, this cuts two ways. On one hand, teams can produce far more marketing copy, documentation, and support material than they could two years ago -- which is exactly the pitch behind tools in Marketing & Campaigns. On the other hand, when a third of the web looks the same, differentiation gets harder, not easier. Search engines and readers alike are going to get pickier about what's actually worth reading, and 'we published a lot of content' stops being a strategy on its own.
The same week that study landed, Google rolled out a new feature letting readers designate specific publishers as 'preferred sources' across Search, Discover, and Google News, per TechCrunch. The timing isn't a coincidence. AI search and AI-written content are both siphoning clicks away from the open web, and Google -- which built its empire on sending traffic to that web -- has a vested interest in keeping publishers alive enough to keep producing something worth surfacing.
I'm skeptical this alone reverses the trend. Asking readers to actively favorite a publisher requires a level of intent most people don't bring to a Google search. It's a nice gesture, but it puts the burden of survival on habits that are already fading. The more durable lesson for any business that depends on organic traffic -- whether you're a media outlet or a company running a Websites product for SEO -- is that you can no longer treat search referral traffic as a given. Owning your audience directly, through email, community, or a product people return to on purpose, matters more than optimizing for an algorithm that keeps changing the rules.
Ramp launched its own AI model router, simply called Router, letting companies switch between large language models via a single API, TechCrunch reported. This comes right after Stripe's acquisition of OpenRouter, which TechCrunch noted wasn't really about 'the singularity' despite the framing -- it was about payments companies wanting a foothold in how businesses actually spend money on AI inference. Ramp building its own version instead of just plugging into someone else's tells you how fast 'model routing' has gone from a niche developer tool to table stakes infrastructure that finance and fintech companies feel they need to own outright.
For a business buyer, the practical takeaway is that no company should be locking itself into a single AI model or vendor right now. The tools for switching cheaply between models are becoming commoditized fast, which is good news if you're trying to control costs -- it's the same logic behind evaluating any platform on Integrations rather than a single point solution. But it's also a reminder that the AI layer underneath your tools is far less stable than the interface you actually use every day, and that stability -- not the flashiest model -- is what should drive your vendor choices.
Which of these trends worries you more as a business owner: a web flooded with AI-written content, or losing control over which AI model is actually answering your customers?
Sources