Large language models and deep research agents supply citation URLs to support their claims, yet the reliability of these citations has not been systematically measured. We address six research questions about citation URL validity using 10 models and agents on DRBench (53,090 URLs) and 3 models on ExpertQA (168,021 URLs across 32 academic fields). We find that 3--13\% of citation URLs are hallucinated -- they have no record in the Wayback Machine and likely never existed -- while 5--18\% are non-resolving overall. Deep research agents generate substantially more citations per query than search-augmented LLMs but hallucinate URLs at higher rates. Domain effects are pronounced: non-resolving rates range from 5.4\% (Business) to 11.4\% (Theology), with per-model effects even larger. Decomposing failures reveals that some models fabricate every non-resolving URL, while others show substantial link-rot fractions indicating genuine retrieval. As a solution, we release urlhealth, an open-source tool for URL liveness checking and stale-vs-hallucinated classification using the Wayback Machine. In agentic self-correction experiments, models equipped with urlhealth reduce non-resolving citation URLs by $6\textrm{--}79\times$ to under 1\%, though effectiveness depends on the model's tool-use competence. The tool and all data are publicly available. Our characterization findings, failure taxonomy, and open-source tooling establish that citation URL validity is both measurable at scale and correctable in practice.

翻译：大型语言模型与深度研究代理通过提供引用URL来支撑其论断，然而这些引用的可靠性尚未得到系统化评估。我们针对引用URL有效性提出六项研究问题，采用DRBench基准（53,090条URL）上的10个模型与代理，以及ExpertQA基准（涵盖32个学术领域的168,021条URL）上的3个模型开展研究。研究发现3%-13%的引用URL属于幻觉——它们未在Wayback Machine中留下记录且很可能从未真实存在——而总体有5%-18%的引用无法解析。相较于检索增强型大语言模型，深度研究代理每个查询生成的引用数量显著更多，但URL幻觉比例更高。领域效应尤为显著：无法解析率从5.4%（商学）到11.4%（神学）不等，各模型的效应值甚至更大。通过剖析失败案例发现，部分模型完全虚构所有无法解析的URL，而另一些模型则呈现大量链接失效比例，这表明存在真实检索过程。作为解决方案，我们发布了urlhealth开源工具，该工具利用Wayback Machine进行URL存活性检测，并可将失效链接与幻觉链接分类。在代理自我修正实验中，配置了urlhealth的模型将无法解析的引用URL降低了6至79倍，使其降至1%以下，但具体效果取决于模型的工具使用能力。该工具及所有数据均已公开发布。我们的特征刻画发现、失败分类体系及开源工具共同证明：引用URL有效性既可在规模化层面测量，亦可在实践中实现修正。