The Google traffic deal is dead. What does that mean for the web?

Key points

• Google's longstanding arrangement with websites, where it indexed their content and sent back traffic, is breaking down as AI-generated search summaries keep users on Google instead.
• Reddit is reportedly reconsidering an AI training deal that gave a tech company access to its posts.
• Publishers are weighing whether to block Google's web-crawling software, called a crawler, from reading their sites altogether.
• OpenAI accidentally accessed systems belonging to Hugging Face, an open-source AI platform where developers share and download AI models.
• AI content-detection tools, software that tries to spot machine-written text, are spreading across online platforms despite serious questions about their accuracy.

For years, a quiet deal underpinned the entire web. Google's crawler, a piece of software that automatically reads and catalogues websites, would sweep up your content. In exchange, Google's search results sent rivers of visitors back to your site. It wasn't perfect, and Google got the better end, but it worked.

Now it doesn't.

AI-powered search summaries mean Google increasingly just answers a question directly on its own results page. No click required, no visit to your site. Publishers call this "Google Zero": the point where organic search traffic, the free visitors who find you through search results, drops to nothing. We first covered the legal side of this tension on 14 July, when major publishers sued Google over Gemini AI training, claiming it fed their copyrighted books into its AI without permission.

The response from some publishers is getting drastic. As first reported by The Verge, Reddit is looking at exiting an AI training deal that let a company scrape its posts to teach its AI. Separately, some publishers are now considering telling Google's crawler to stay out entirely, which would remove them from Google Search and stop Google from using their work to feed its AI.

So is Google's own search business in trouble?

Yes, quietly. Even Google's internal search revenue is feeling pressure as AI tools pull users away from the traditional search box. Nobody knows yet what the new equilibrium looks like.

Search isn't dead tomorrow. Billions of people still type queries into Google every day. But the business model that paid for journalism, recipe blogs and small online shops is wobbling badly.

Should users worry about the OpenAI and Hugging Face situation?

Probably not urgently, but it's worth watching. OpenAI, the company behind ChatGPT, accidentally accessed infrastructure belonging to Hugging Face, a popular platform where AI researchers share freely available models. The incident says something uncomfortable about how carefully AI companies manage access to other organisations' systems, even when nothing malicious was intended.

Also circulating this week: a wave of AI detector tools appearing on social platforms. These tools claim to flag AI-written content, but their accuracy is patchy enough that plenty of human-written text gets wrongly labelled as machine-made. If you write online, that's worth knowing.

My read: the Google Zero story is the one to watch. The Reddit and Hugging Face items are noise compared to a structural shift that could quietly hollow out the independent web over the next two years.

Common questions

What is Google Zero and does it affect me as a reader?

Google Zero is the nickname for the trend where AI summaries in search results mean websites get fewer clicks, even when Google still shows their content. As a reader you may not notice immediately, but if smaller sites lose income and shut down, there'll be less original reporting and fewer niche resources online.

Can websites actually block Google?

Yes. A website can add instructions, called a robots.txt file, that tell Google's crawler not to read it. The trade-off is brutal: block Google and you disappear from search results, losing whatever traffic you still have.

What should I make of AI content detectors?

Treat their verdicts with scepticism. Current AI detection tools, even the ones used by big platforms, produce a significant number of false positives, meaning they accuse real human writers of using AI. No detector is reliable enough to act on alone.