The web's deal with Google is breaking

For twenty years, the open web ran on an unspoken exchange. Google crawled every public page it could find, added them to its search index, and in return sent the bulk of the referral traffic that funded independent publishing. The click was the price of the crawl. The latest Vergecast names the potential collapse of that exchange "Google Zero," a term The Verge's Nilay Patel coined for the moment Google's traffic to the open web actually dries up, and says the moment has arrived.

AI Overviews changed the click dynamic. The AI-generated answers that sit above the traditional ten blue links in Google Search now answer many queries without sending a click at all. According to a Define Media Group analysis of Google Search Console data across 64 sites, reported by Search Engine Land, organic search clicks fell 16% right after AI Overviews expanded in May 2025, never recovered, and by Q4 2025 were down 42% from the pre-AI Overviews baseline. Sample composition skews toward news-oriented publishers, so the magnitude may not generalize, but the direction is unambiguous.

Reddit is the highest-profile publisher testing a different response. The Verge reports that Reddit is publicly weighing whether to terminate its AI training deal with Google as the traffic flow runs the wrong way. The company's posture, as described in the reporting, is "interest in getting out of" the arrangement, not a confirmed exit. The move would matter because Reddit is one of the largest sources of human-written text on the open web, and Google has used it as both a training corpus and a search-indexed destination. Pulling out of the AI side while remaining in the search index is the new shape of the negotiation.

The lever most publishers are weighing is older and plainer than the AI negotiations: the robots.txt file, a plain-text file at the root of every website that tells crawlers which paths they may fetch. Blocking Google's crawler does not get a site un-indexed from yesterday's results overnight, but it does stop Google from refreshing its copy and from using the site as AI training material. Some publishers are reportedly considering exactly that step. "Considering" is the right verb: this is an emerging posture, not a coordinated industry action.

The exchange is breaking on Google's side, too. Search ads are sold against the click, and the same AI-mediated answers cutting into publishers' referral traffic are also reducing Google's ad revenue. The crawl-for-traffic bargain now has a smaller pie on both sides.

One honest counter sits inside the source: the deal was always asymmetric. Google made more from the exchange than most publishers did, and many of the sites pushing back hardest built their audiences on Google's free referrals in the first place. "Google Zero" may be the old asymmetry finally exposed, not a new grievance. The same data shows breaking-news traffic grew 103% from November 2024 through early 2026, evidence that the hit concentrates on evergreen search rather than the live, time-sensitive queries Google still routes well.

The decision space is open and not theoretical. Publishers can sign AI licensing deals, an established model where AI companies pay to train on a publisher's corpus. They can negotiate pay-per-crawl, the framework Cloudflare has been pushing. They can block Google's crawler in robots.txt and try to force a different deal. Or they can try to return to a direct-traffic model, where the audience comes back through email, apps, or social, and the search engine becomes less of a middleman. Each path has costs, and each will surface within the next year.

The next concrete signal to watch is the Reddit-Google AI training negotiation, and whether the 42% baseline stabilizes, recovers, or worsens once the AI Overviews expansion completes its rollout.