AI Search Reliability Under Scrutiny as Over 50% of Web Traffic Turns to Bots and Sites Fight Back

AI Search Faces a Perfect Storm of Reliability and Access Issues

AI search is fundamentally reshaping how people find information online, but emerging research and industry trends reveal a troubling paradox. While AI chatbots demonstrate superior performance in combating disinformation compared to traditional search engines, the reliability of AI-generated content remains questionable, and the economic model underpinning the web is collapsing. AI crawlers now represent over 50% of all web traffic according to Cloudflare, which manages 30% or more of the top 10,000 sites on the internet1

1

. This surge is forcing website owners to make difficult choices about blocking AI access entirely, creating a feedback loop that could degrade the quality of AI-generated search summaries for years to come.

The Citation Crisis in AI-Generated Search Summaries

A May 2025 study examining tens of thousands of Google AI Overview queries uncovered a significant problem: 11% of results were not supported by the citations provided2

2

. As researchers noted, this represents "a failure mode a reader who sees only the AIO summary has no way to detect." The trustworthiness of AI search results becomes even murkier when examining source credibility. A March 2025 study found that ChatGPT's GPT-5 model quoted reliable sources only 71.4% of the time, with source credibility determined using a media fact-checking database2

2

. These unsupported citations don't necessarily mean answers are wrong, but they indicate AI-powered web search tools are citing sources not recognized as authoritative roughly 3 out of 10 times.

How AI Chatbots Outperform Traditional Search Against Disinformation

Despite citation concerns, AI chatbots show surprising strength in combating foreign propaganda and false narratives. An NPR experiment conducted with NewsGuard tested popular AI chatbots including ChatGPT and Google Gemini against 30 questions based on false narratives pushed by China, Iran, and Russia between December 2025 and July 20263

3

. AI chatbots correctly debunked false narratives about three-quarters of the time, outperforming traditional search engines. Mike Caulfield, a digital literacy expert at the University of Washington, Bothell, said using AI chatbots with web search access is "a good way for users to start to investigate these issues"3

3

. When Russia shelled a historic Ukrainian monastery in June and Kremlin-aligned outlets falsely blamed Ukraine, all tested chatbots and Google AI Overview correctly identified the claim as disinformation3

3

.

AI-Generated Search Summaries Show Mixed Performance

While AI chatbots excel at debunking disinformation, AI-generated search summaries that appear at the top of search results present inconsistent results. These summaries debunked false narratives a majority of the time but at lower rates than AI chatbots, and they failed to challenge false narratives at higher rates than traditional search engines3

3

. Performance varied significantly between products: Google AI Overview debunked false narratives most of the time, while Bing summaries failed to debunk most of the time, and DuckDuckGo fell somewhere in between3

3

. Morgan Wack, a postdoctoral researcher at the University of Zurich studying digital political persuasion, emphasized that traditional search tools were never perfectly neutral, noting that "non-biased information was never really a state of affairs"3

3

.

The Economic Collapse Threatening Quality Information

For 30 years, the web operated on a social contract: sites allowed search engines free access in exchange for traffic-driving links. AI search is breaking this agreement. When users search for information, ChatGPT responses or Google AI Overview results include links as "optional extras" rather than primary destinations1

1

. AI crawlers examine sites more deeply and intensely than traditional web crawlers, creating costs for website providers without delivering the advertising revenue that comes from human traffic1

1

. This dynamic is accelerating toward "Google Zero"—the day when click-through traffic from Google drops to nothing1

1

.

Sites Are Blocking AI Access, Creating a Quality Crisis

Website owners are responding by blocking AI crawlers altogether. A Wikipedia study confirmed traffic in the English version dropped quickly after Google launched AI summaries in English, with the same pattern occurring in other languages as rollouts continued1

1

. Come September 15, Cloudflare sites will block AI crawlers by default on pages containing advertising, meaning up to 30% of the world's top sites will no longer appear in Google AI Overview summaries1

1

. Sites containing disinformation are far less likely to ban AI crawlers, creating a perverse situation where AI answers won't be informed by high-quality sources1

1

. One recent study found that already, around 1 in 6 sources used by AI search tools is itself an AI-generated website1

1

.

Model Collapse Threatens Future AI Search Quality

As AI models train on increasing amounts of AI-generated content, they risk model collapse—a phenomenon where output quality degrades over time1

1

. This creates a vicious cycle: high-quality sites block AI crawlers, forcing AI to train on lower-quality and AI-generated content, which further degrades the reliability of AI search results. The multi-stage process of AI-powered web search tools—parsing queries, retrieving relevant webpages, and packaging responses—can fail at multiple points by misunderstanding questions, choosing wrong information, or mixing sources incorrectly2

2

. OpenAI itself encourages ChatGPT web search users to check citations and sources and focus on authoritative information2

2

.

What This Means for Information Seekers

The quality of AI-generated search summaries is likely to decline in the short term as the economics of the web get sorted out. Search engines that depend less on AI may become more reliable, though finding ones that don't use AI crawlers is increasingly challenging1

1

. Users should maintain skepticism when assessing answers from AI search, particularly because AI chatbots are more likely to attempt answering queries even when little or no information exists online, presenting speculative responses with confidence2

2

. The convenience of AI search can be misleading—traditional searches make information gaps immediately visible, while AI tools may fill those gaps with unreliable content. Watch for continued site blockages, evolving "pay to crawl" compensation models, and potential regulatory interventions as the industry grapples with balancing innovation against information quality and economic sustainability.