Google is cannibalizing the web to feed AI

OPINION At Google’s I/O 2026 developer conference last week, the company’s Search VP Liz Reid celebrated the rapid growth of AI Mode, which Google refers to as its end-to-end AI search experience. External observers refer to it as simply as the end of search. “We’re seeing phenomenal growth with AI mode queries more than doubling every quarter since launch,” said Reid, noting that AI Mode now has more than one billion monthly users. She continued, “But what’s even more remarkable, you’re asking your real questions in all their super specific and detailed glory, knowing search can really tackle them.” Reid is absolutely right! We put this real question to AI Mode, “Why does Google Search suck now?” And you won’t believe the answer! (Actually, you won’t be surprised.) The Chocolate Factory’s AI search substitute offered this completely trustworthy reply: “You are definitely not imagining it. Users, tech critics, and researchers have documented a measurable decline in Google Search quality. The core issue is that the search engine no longer feels like a tool designed to find the best corner of the web; instead, it feels like a vehicle designed to keep you on Google-owned properties or clicking on monetized links. “The degradation of Google Search stems from a mix of aggressive monetization, an ongoing arms race with web spammers, and the disruptive introduction of AI features.” Despite the absence of explicit blame assignment and the obsequious initial sentence, this all seems reasonable. However, in the low-rent area of the results page, there was some fine print: “AI can make mistakes, so double-check responses.” In other words, Google disclaims responsibility for AI Mode while providing a way for you, the user, to double-check its responses. If you look carefully at the results, you may see citation links back to the websites laundered to generate the unvouched-for text. The top AI Mode paragraph represents the remains of four morsels of online text swallowed and digested by Google’s site crawler: “This is Why Google Search is Dying* and How to Search Instead” from Tadeusz Szewczyk’s seo2 blog, “Why Has Google Search Become So Annoyingly Bad? (and some useful workarounds),” a post to Medium by Terry Hutchins, a Reddit thread titled, “Is Google Search actually getting worse, or is it just me?” by Severe_Aardvark_3109, and “Some Simple Economics of the Google Antitrust Case” by Alex Tabarrok. The second paragraph is calling from within Google’s house. It’s sourced from a YouTube video by AlexFalcone and a different Reddit thread, “Am I going crazy or are search engine results becoming less and less accurate?” by Ok-Extent-7596. Google generously presents these links, in the form of inline citations or source chips – clickable, numbered footnotes that have been embedded with AI Mode emissions – as a nod to shouty traditionalists who value information provenance. Not that it helps these websites much. As noted recently by SEO biz Ahrefs, “Google’s AI Overviews now result in a 58 percent lower average clickthrough rate for top-ranking pages, up from 34.5 percent just eight months ago.” (Google last year said the opposite.) AI Overviews differ from AI Mode. AI Overviews have been integrated into the Search experience and appear in response to certain kinds of queries, generally atop traditional search results. AI Mode “expands on the benefits of AI Overviews with more advanced reasoning and multimodal capabilities,” as Google puts it in its explainer [PDF]. It’s accessible through a tab icon on the right-hand side of Chrome’s omnibox when a new tab is opened and through an “AI Mode” button within the Google.com search box. They may or may not provide similar answers and may or may not cite the same sources. For example, when we put the question “Why does Google Search suck now?” to AI Overviews, here’s what we got: “Google Search often feels worse today because it prioritizes profits over precision. The search results page is heavily cluttered with AI-generated summaries, sponsored ads, and search-engine-optimized (SEO) spam, which forces you to dig past multiple links to find the actual information you want.” Google’s AI again cites Szewczyk’s seo2 blog as a reference, along with two new sources, “These Results Illustrate Why Google Search Is So Awful in 2025,” from Make Use Of, and “Enough is enough: I ditched Google’s broken search engine and boosted my productivity,” from Android Authority. Just for the sake of completeness, here’s AI Overview answering, “Why is Google Search great?” “Google Search is great because it combines unmatched speed, advanced AI capabilities, and a massive index of information. It does more than just provide links, allowing you to easily plan trips, research local businesses in your immediate San Francisco vicinity, and understand complex topics with minimal effort.” Among the five citation links for that result, three reference Google blog posts. The privacy intrusion – having the browser surface location information unbidden – is just a reminder about the consequences of browser personalization. This is the new (horrible) normal Like it or not, there’s going to be more AI in search going forward. Reid described how Google’s redesigned Search box now includes deeper AI integration. “We’re making it even easier to continue the conversation with Search, bringing AI Overviews and AI Mode into one seamless AI search experience, so you can flow effortlessly from your question to your response on the main search results page to follow-ups in AI mode,” she said. Why not just let users decide which sites they trust? The anticipated result, at least among those outside of Google, is that search results links to external websites will be further deemphasized. The problem with this approach is that it means Google makes its AI into a traffic routing layer, as opposed to presenting a set of site and document referral options that users can select. It wants to direct the flow of traffic instead of just providing a map. The endgame, presumably, is capturing some portion of automated transactions made by software agents through mechanisms like Google’s Universal Cart, Agent Payments Protocol (AP2), and Universal Commerce Protocol (UCP). When Google itself cautions, “AI can make mistakes, so double-check responses,” why not just let users decide which sites they trust for themselves, rather than burying source information in footnote links and directing those using its AI service to handle the fact-checking? AI models can be quite helpful for coding questions. They’re a vast improvement over man pages for looking up esoteric command line flags and options. But they should be kept separate from search. If they must be entrusted with automated browsing, their decisions or recommendations should be subject to user discretion and review. Search assumes source visibility and invites searchers to consider whether they trust sites enough to visit them. It demands mental engagement and decision-making to move from a list of results to query resolution. That’s a role people should play. AI presents an answer, one that’s so easy to accept it comes with a reminder to double-check. But it’s an answer without responsibility or liability. It’s not the right answer for the web. ®

المصدر: The Register