一位用户详细讲述了他们使用Google的AI搜索来调查关于电影《蜘蛛侠：明日不再》的一个事实声明的经历。该AI最初提供了虚构的背景故事以及关于维基百科编辑过程的不正确信息，包括错误地声称维基百科的编辑无法被追踪。尽管AI存在不准确和幻觉，用户还是能够识别出虚假声明，将其从维基百科中移除，并强调了AI在事实核查和提供准确来源方面的不可靠性。 AI
影响 凸显了当前AI在事实核查方面的不可靠性以及用户尽职调查的必要性。
排序理由 用户个人经历和对AI不准确性的详细描述。
在 Mastodon — mastodon.social 阅读 →
AI 生成摘要 · Google Gemini · 来自 9 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
Yes, I want to believe that I am diligent about checking AI answers. But am I really? Do I have the energy and the will to check everything? I doubt it. I probably slipped many times. When I didn’t care enough about the answer. When the answer sounded pleasing enough to me at the…
But taking an AI instead for answering questions? This conversation happened in October 2026. ChatGPT has been around for almost four years, and we have known about the hallucination problem for even longer. How do I fix the AI in Google’s search? Sure, it also warns us that “AI …
The relevant edit can absolutely be linked to. It was not added on August 22. It was not added by an anonymous contributor, but by a volunteer editor with tons of experience. Those things can slip through, and the community can fix it, as happened today. Or maybe it is true, and …
“Can you give me a link to the diff that introduced this on the 22 August?” Google: “Because Wikipedia's specific page architecture dynamically generates temporary revision IDs rather than permanent links for unverified edits, providing a functional direct hyperlink to that preci…
I asked. Several times. It kept responding that the answer generation failed or something like that. Eventually, the Google AI answered. Me: “Can you figure out who entered the statement about the altered scene to Wikipedia?” Google: “Because Wikipedia’s edit history is completel…
“While the text you initially asked about is currently written verbatim on Wikipedia's Spider-Man: Brand New Day page, the elaborate details I gave you in my first response to explain why it happened were entirely made up by me.” To make it clear: not only did it first hallucinat…
I ask: “Where's the link to the Wrap interview?” And now it gets really interesting. Instead of giving me a link, Google answers: “The statement that "the scene was altered in post-production due to its perceived sexual undertones" is actually a fabricated claim from user-edited …
Google immediately launched the AI mode, and stated that the text is from the Wikipedia article (correct). It further explained that “during test screenings, younger viewers expressed discomfort with the perceived sexual undertones of the moment. In response, the creative team ma…
This is not really about Spider-Man, but about the Truth, about AI, and Wikipedia. From time to time I stumble into a rabbit hole where I start fretting about a single fact, and try to get down to the bottom of it. Sometimes I write about these. This time, I stumbled upon two sen…