Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

ChatGPT and Gemini show different citation patterns in REALI study

AI summary

REALI collected 1,800 answers from ChatGPT and Gemini across three waves of tests about 2026 Taipei mayoral-election topics. ChatGPT concentrated 26.2% of its news citations on Central News Agency, while Gemini's leading news source, United Daily News, accounted for 7.5%; Taipei City Government ranked first among non-news sources for both models. The study also found that identical questions could produce different cited issues over time, but REALI cautioned that its findings do not establish source credibility or causal influence.

In-depth insight

The results provide concrete evidence that AI assistants can favor very different news and non-news sources, making citation visibility a relevant consideration for publishers and information analysts.

ChatGPT、Gemini 資訊從哪來?研究揭兩款 AI 引用網站差異

生成式 AI 已成為民眾搜尋及整理資訊的常用工具,但同樣的問題交給不同 AI,背後參考的資料來源可能大不相同。AI 資訊分析平台 REALI 近期針對 ChatGPT 與 Gemini 進行測試,蒐集共 1,800 則回答,發現除了新聞媒體與政府網站,PTT、Dcard、維基百科及 YouTube 等平台,也會出現在 AI 回答的引用來源中,且兩款 AI 對不同網站的引用情形存在明顯差異。

REALI 透過旗下 GEO Positioning System(GPS)分析工具,以 2026 年台北市長選舉相關議題作為測試情境,於 10 月 30 日至 11 月 1 日分三波蒐集回答,每題、每模型、每波各測試 15 次,藉此觀察不同 AI 服務引用網站的情況,以及回答內容是否隨時間及提問方式出現變化。

#REALI#ChatGPT#Gemini#GEO Positioning System

Original · 3 min readRead the original
Group chat

No comments yet. Start the conversation.