What is Cofacts?

An open fact-checking system in Taiwan: when citizens forward a suspicious message they received on LINE or elsewhere to a chatbot, they get back an answer researched and written by volunteers. Anyone can take part. It grew out of g0v, the civic hacker community, and its data can be used by anyone.

Has your family group chat ever received a message like "drinking this cures the disease" or "such-and-such is dangerous"? You may wonder whether it is true, but checking it yourself takes effort. Cofacts is a system in Taiwan where you simply forward such a message to a chatbot and get back an answer researched by fellow citizens.

In short

Cofacts is a fact-checking system born in Taiwan: forward a dubious story to a chatbot (a program that converses automatically), and you get back an answer researched and written by volunteer citizens. Its Chinese name is "真的假的", meaning "Really? Or fake?" [1][2][3].

Anyone can write answers. The data collected can be used freely by anyone [1][2][3].

For example

Rumors inside conversations no one else can see. Messaging apps like LINE, which is widely used in Taiwan, keep conversations hidden from outsiders. That makes rumors hard to investigate and counter. Johnson Liang, a core member of Cofacts, writes that the very messages users send to the bot become a valuable record of what is happening inside these closed social networks. In March 2018, Cofacts processed this record so that individuals could not be identified and made it public [2].

Before the same-sex marriage referendum (2018). Liang writes that Cofacts' analysis features made it possible to show in a graph how widely messages containing "homosexuality" were circulating. Cofacts was also interviewed for international media reports on how Taiwan fought anti-LGBT fake news before the vote on same-sex marriage [2].

A "hub" for other bots. Cofacts' database is published in a form anyone can use without restriction (CC0). So a popular bot made by someone else, "美玉姨" (Auntie Meiyu), also uses Cofacts' data [3]. Audrey Tang described "the Cofacts playbook of being almost a hub, a repertoire", with "any number of downstreams" [4].

According to Tang, even elderly parents who can use a smartphone but are not comfortable with a computer can become participants who ask for checks or forward the results to their families, simply by adding such bots [4]. Tang says the same system also exists in Thailand: "Cofacts is not just Taiwan, it's also Thailand" [5].

How it works

Forward it, and an answer comes back

According to Plurality, in Cofacts "participating citizens rapidly respond to both trending social media content and to messages from private channels forwarded to a public comment box for requested response" [1].

In Liang's 2019 review, even when users passed 70,000, the number of editors writing answers did not grow. So in July 2018 Cofacts added a feature that asks senders to write what they find suspicious when they send a message [2].

There were three aims: to filter out joke submissions, to let editors choose what to investigate, and to have the senders themselves question the information — a first step in media literacy (the ability to read information critically) [2]. In August, a button was also added for sharing a helpful answer with friends [2].

Who runs it

Cofacts is a spin-off project of g0v (pronounced "gov-zero"), a community of civic technologists [1]. Billion Lee (李比鄰) is introduced as a co-founder [6].

In 2018, Cofacts held 50 in-person meetings on Wednesday evenings. Every two months it also held small gatherings of editors, who spent 100 minutes writing answers together, Liang records [2].

According to Liang's 2019 article, Cofacts drew attention by chance in early 2017 while it was still unfinished, and users flooded in faster than editing volunteers. The English name Cofacts was decided in April 2017 [3]. It received a civic tech grant from g0v, but Liang writes that those who provided the money do not interfere in how it is run [3][2].

Not a neutral third party

Liang acknowledges that since anyone can take part, the accuracy of each individual answer cannot be guaranteed. Instead, anyone can add another answer at any time [3]. Answers include a category for "contains personal opinion", and all answers to a message are shown side by side. If you do not like who is writing the answers, you can also set up your own bot with the published data and code [3].

Liang writes that by not restricting who editors are, Cofacts is an attempt to counter false information through freedom of speech. Whether it succeeds or fails, Liang says, it will be a reference for countries facing the same problems [2].

Bringing in AI

In 2023, Tang said that Cofacts had already trained a language model (an AI that handles text) and was having the AI add context and prompts for thinking at the point when a piece of false information is first seen [7]. In 2024, Tang explained that it provides "basically zero-day responses to zero-day viral disinformation, before any fact checker can look at any viral message", linking newly emerged rumors to related warnings from past cases [8].

In a 2025 interview in German, Tang described Cofacts as something like a Wikipedia of fact-checking. Tang said the government does not run it, experts also support it, and it is used in school media literacy classes [9].

What is debated

Suspicion that it is a government tool. In 2019, a column portrayed Cofacts as something like an agency policing speech, and hinted that it was acting on the orders of Tang or the government of the time. Liang rebutted the points one by one: the Auntie Meiyu bot was made by someone else, the project has never stopped, and Liang had met Tang only as a fellow g0v participant [3]. Tang also said in 2024 that Cofacts is not on the government side but remains "in a civil society and funded by nonprofits and donations and so on" [8].
Not adopted by the government. A Daily Beast article (2023) criticizes the fact that Cofacts was not adopted as a main government tool against false information. Lawmakers, it says, preferred strong measures such as fines on citizens who spread false information [10].
The burden on volunteers. Cofacts itself listed as a problem in 2018 that the number of editors did not grow even as users increased [2].

Where it fits in Plurality

Chapter 2-2 of Plurality, "The Life of a Digital Democracy", in its section "Information integrity", presents Cofacts as central to Taiwan's "superpower" of responding rapidly to misinformation [1]. Citing research, the book writes that "these systems can typically respond faster, equally accurately and more engagingly to rumors than can professional fact checkers, who are much more bandwidth constrained" [1].

The same section also covers other LINE bots and the approach of Taiwan's efforts against false information known as "humor over rumor" [1]. The overall picture of Taiwan is covered in the article on digital democracy in Taiwan.

In that citizens add supplementary notes, Cofacts is a system close to Community Notes on X. In a 2023 interview, Tang listed Polis, Cofacts and Community Notes together as technologies that show not what is loudest, but what bridges people whose opinions differ most [7]. This idea is covered in more detail in the article on bridging.

Sources

Plurality, 2-2 "The Life of a Digital Democracy" — https://www.plurality.net/v/chapters/2-2/eng/ ［CC0］
Johnson Liang, "2019 第一天，來看看兩歲的 Cofacts" (On the first day of 2019, a look at two-year-old Cofacts) (2019-01-01, in Chinese) — https://medium.com/cofacts/2019-%E7%AC%AC%E4%B8%80%E5%A4%A9-%E4%BE%86%E7%9C%8B%E7%9C%8B%E5%85%A9%E6%AD%B2%E7%9A%84-cofacts-d2330f01727b
Johnson Liang, "摘掉『公正第三方』的高帽子 — Cofacts 對陰謀論的回應" (Taking off the "neutral third party" hat: Cofacts' response to conspiracy theories) (2019-01-11, in Chinese) — https://medium.com/cofacts/%E6%91%98%E6%8E%89-%E5%85%AC%E6%AD%A3%E7%AC%AC%E4%B8%89%E6%96%B9-%E7%9A%84%E9%AB%98%E5%B8%BD%E5%AD%90-cofacts-%E5%B0%8D%E9%99%B0%E8%AC%80%E8%AB%96%E7%9A%84%E5%9B%9E%E6%87%89-bd948d128d6c
Audrey Tang, Interview with Microsoft Research, Part II (2024-01-01) — https://github.com/audreyt/transcript/blob/main/2024-01-01-Interview-with-Microsoft-Research%2C-Part-II.md ［CC0］
Audrey Tang, Conversation with AI Objectives Institute (2024-05-06) — https://github.com/audreyt/transcript/blob/main/2024-05-06-Conversation-with-AI-Objectives-Institute.md ［CC0］
"國防在線急 S2E7" (with Billion Lee, video, in Chinese) — https://www.youtube.com/watch?v=mCHU0RsVbL8
Voice of America interview with Audrey Tang (2023-09-23, in Chinese) — https://sayit.archive.tw/2023-09-23-%E7%BE%8E%E5%9C%8B%E4%B9%8B%E9%9F%B3%E5%B0%88%E8%A8%AA%E5%94%90%E9%B3%B3%E9%83%A8%E9%95%B7%E9%80%90%E5%AD%97%E7%A8%BF ［CC0］
Audrey Tang, Interview with Your Undivided Attention podcast (2024-01-24) — https://github.com/audreyt/transcript/blob/main/2024-01-24-Interview-with-Your-Undivided-Attention-podcast.md ［CC0］
D+C, "Taiwan hat die gesellschaftliche Polarisierung abgeschwächt …" (interview with Audrey Tang, 2025, in German) — https://www.dandc.eu/de/article/taiwan-hat-die-gesellschaftliche-polarisierung-abgeschwaecht-und-ist-heute-besser-gegen
Daily Beast, "Taiwan tried to digitize democracy with vTaiwan. It was a huge flop" (2023) — https://www.thedailybeast.com/taiwan-tried-to-digitize-democracy-with-vtaiwan-it-was-a-huge-flop/

We welcome error reports and questions ("What is…?"). Submissions become public issues.

Report an error in this article ・ Ask a question