Why People Do Not Click Sources After AI Summaries

As generative AI chatbots and large language models (LLMs) become increasingly popular for answering health-related questions, a notable shift has emerged in the way people seek and consume medical information. Instead of using traditional https://bizzmarkblog.com/why-cochrane-asks-authors-to-acknowledge-technologies-used/ search engines that present multiple sources and direct links, many users now receive concise AI-generated summaries, often without following up on the cited references. This trend has raised critical discussions about trust, accuracy, and the role of source verification in the digital health information ecosystem.

You ever wonder why historically, when people had questions about health, they turned to search engines like google or bing. These platforms returned lists of links to webpages, academic articles, and trusted databases, such as those from The Cochrane Collaboration or the Cochrane Library. Users could then choose which sources to review, helping them form their own understanding supported by original studies or systematic reviews.

With the advent of generative AI chatbots powered by LLMs, such as ChatGPT and similar tools integrated into search engines or standalone apps, the experience has changed dramatically. Instead of pages of links, users receive succinct answers, sometimes styled as plain language summaries, that paraphrase or synthesize existing knowledge.

This transformation has benefits. AI assistants can save time and effort by distilling extensive medical evidence into digestible content. They can also help bridge the literacy gap by producing responses in accessible language, clarifying complex concepts for lay audiences.

Despite improved usability, trust remains the critical bottleneck preventing many users from clicking through to original sources after reading an AI-generated summary. There are several reasons for this hesitation:

In the context of health, this is particularly problematic. Medical decisions should ideally be based on reliable evidence, such as systematic reviews regularly curated by organizations like The Cochrane Collaboration. The Cochrane Library offers rigorously peer-reviewed summaries and meta-analyses, but if users skip this step, they may miss nuances or relevant caveats.

Moreover, health information is especially sensitive to inaccuracies or misinterpretations, where misplaced trust can lead to real-world harms.

One fundamental feature—and limitation—of LLMs is that they generate text based on statistical patterns learned from vast datasets. They do not possess understanding or fact-checking capabilities inherent to human experts or specialized databases. This means:

This has implications for the trustworthiness of AI assistant outputs. A generated summary may incorporate accurate information but also inadvertently include errors, oversimplifications, or misleading statements without signaling these issues to the user.

One client recently told me learned this lesson the hard way.. “Hallucination” is a technical term describing when an AI model generates false or fabricated information with confidence. In the health domain, one particularly concerning manifestation is invented citations—where an AI chatbot produces references or DOIs that look legitimate but do not actually exist or do not support the stated claims.

This phenomenon undermines user confidence and complicates source verification. For example, if an AI summary cites a Cochrane review but the link or DOI fails to resolve legitimately in the Cochrane Library, users may become skeptical of all AI-generated content. Unfortunately, hallucinated references are relatively common because the model “fills in” gaps using learned patterns without rigorous fact-checking.

Summarizing the key barriers to source clicking:

Quality evidence synthesis groups such as The Cochrane Collaboration have long emphasized the importance of transparent, accessible, and full reporting of systematic reviews to reduce misinformation and empower patients and clinicians alike.

The Cochrane Library is a vital resource designed to provide plain language summaries alongside detailed technical reports. This layered approach encourages readers to consult the summary for quick understanding, but also to dig deeper when needed, enhancing informed decision-making.

Integrating trustworthy sources like Cochrane into AI chatbot platforms—complete with functioning links and clearly labeled evidence levels—could help improve user trust and promote source checking behavior.

Addressing the reluctance to click sources Additional hints after AI summaries requires multiple complementary strategies:

The shift from search engine summaries to AI-generated replies offers exciting possibilities to democratize health knowledge with plain language summaries. However, trust remains the linchpin for safe and effective use. Large language models do not “understand truth” and are prone to hallucinations, especially in the form of fabricated citations. This contributes to why users often do not click through to authoritative sources like the Cochrane Library after reading AI summaries.

To realize the full promise of AI in health information, it is essential to build transparency, improve user education, and strengthen ties between AI developers and trusted evidence synthesis providers such as The Cochrane Collaboration. Only then can we help ensure that users not only receive plain language answers but also feel confident exploring and verifying the scientific foundations behind them.

What would change your mind about the value of AI-generated health summaries without clickable, trustworthy references? Let’s keep questioning and improving—because in health, only evidence should lead.