告诉HN:Perplexity在处理世俗查询时默认使用宗教来源。

1作者: fallinditch19 天前原帖
我最近在一次Perplexity聊天会话中遇到了一个奇怪的检索模式,这表明他们的RAG可能受到严重偏向或被小众SEO操控。<p>当时我正在研究一个与宗教完全无关的主题(技术/文化分析),但模型却提供了一系列自信但显然错误的答案。当我审查引用时,发现10个来源中有8个是宗教出版物——比如biblehub、耶稣基督教会和各种宗派博客——这些都与我所寻求的客观事实毫无关联。<p>这让我感觉他们索引中的“域名权威”权重正在被利用。我想现在有一些机构专门提供“基督教Perplexity SEO”服务,以帮助基于信仰的组织“衡量AI扫描器的可信度”。<p>在这种情况和对大型语言模型(LLM)调优中“多样化视角”的普遍推动下,世俗的、客观的查询似乎正受到冲击。<p>最近有没有其他人注意到他们的AI搜索结果中突然涌现出大量基于信仰或智库的来源?<p>[1] https://himfirstmedia.com/perplexity-seo/
查看原文
I recently ran into a bizarre retrieval pattern while on a Perplexity chat session that suggests their RAG might be heavily skewed or actively gamed by niche SEO.<p>I was researching a topic completely unrelated to religion (technical&#x2F;cultural analysis) when the model provided a series of confident but obviously incorrect answers. When I audited the citations, 8 out of 10 sources were religious publications - think biblehub, think churchofjesuschrist, and various denominational blogs - none of which were relevant to the objective facts I was seeking.<p>It feels like the &quot;Domain Authority&quot; weights in their index are being exploited. I guess it&#x27;s a niche for a few agencies now explicitly offering &quot;Christian Perplexity SEO&quot; [1] to help faith-based organizations &quot;measure trustworthiness&quot; in AI scanners.<p>Between that and the general push for &quot;diversity of perspective&quot; in LLM tuning, it seems secular, objective queries are being caught in the crossfire.<p>Has anyone else noticed a sudden influx of faith-based or think-tank sources in their AI Search results lately?<p>[1] https:&#x2F;&#x2F;himfirstmedia.com&#x2F;perplexity-seo&#x2F;