May 9, 2018

After graduating with his bachelor’s degree in journalism, Andrés Jiménez didn’t get a job at a news organization. So instead, he built a fact-checking bot.

“It was quite difficult to find a job and I just wanted to get started,” the Complutense University of Madrid alumnus told Poynter. “In the last couple of years, our profession has been scrutinized and I thought fact-checking would be the best option.”

Enter Facterbot, a Facebook Messenger chatbot designed to deliver recent, far-reaching fake news stories to users' inboxes. In addition to informing people about the most recent piece of misinformation, Jiménez — now a journalism innovation master’s student at Miguel Hernández University — said his goal is to help fact-checkers do their jobs better.

“False stories are shared more than the fact checks that debunk them,” he said. “I thought that sending the stories directly to them could combat that and make it a lot easier to discover if something is false.”

Jiménez is not alone in creating a Messenger bot aimed at delivering fact checks. In Brazil, two fact-checking organizations are building similar products — Aos Fatos’ Fátima and Agência Lupa’s Projeto Lupe!. Facebook is funding both projects, the latter of which is based on a model from Le Monde’s Les Decodéurs, in an effort to limit the spread of misinformation ahead of October’s general election.

Where Facterbot differs is in its approach.

While Fátima and Projeto Lupe! draw upon each organization’s respective fact checks to answer questions in real time, Facterbot delivers a general roundup of popular pieces of misinformation Monday, Wednesday and Friday. Users can choose from pre-selected responses to either learn more about each story or ask about different topics. It even offers translations in Spanish.


Think a reader’s digest, but for fake news.

“There has been some magnificent fact-checking. I wasn’t going to bring anything new in that sense,” he said. “But with this tool, I was going to provide other organizations with what they were missing and reach out to general public. I wanted to make the project not only appealing fact-checking-wise, but also technologically.”

Facterbot has its own mascot, Twitter profile, Medium page and weekly newsletter. On the back end, Jiménez uses Chatfuel, which lets people create chatbots powered by artificial intelligence without doing any coding themselves.

To find content, he trawls Reddit and 4chan, checks hashtags and accounts that have shared hoaxes in the past and monitors websites that fact-checkers have labeled as misinformers. Then Jiménez debunks the stories himself by using reverse image searches, finding primary sources online, checking websites’ Whois data and reaching out to those involved in the story. When bigger outlets debunk the same story, he’ll share that, too.

Many of the stories have to do with politics — Jiménez’s favorite subject — but he’s learned that a lot of fake news stories often touch on topics closer to home.

“They’re interested in politics, but funnily enough the ones that work best are those who talk about health, sports and arts,” he said. “That’s a lesson for me.”

Since launching Nov. 8, the project, which Jiménez has incorporated into his master’s program, has had some small-scale success.

Jiménez said the chatbot reaches a little more than 120 people, with about 40 percent of them reading it every day and 52 weekly active users. Within the next few months, he said he’d like to get to about 400 users and — in the long run — perhaps create packages where subscribers pay to receive a specific number of fact checks each month.

Of course, Facterbot isn’t an automated bot, so scaling its work would prove tricky should Jiménez attract more readers. He said that, if he was to grow the chatbot’s audience and monetize it, he’d hire a software developer to create a native server to save money. Then, he’d want to find another reporter to help with fact-checking.

“I have reached a goal that I had when is started, which is to feel useful — to feel like they gained value when they interacted with my bot,” he said. “Basically, (I’m) encouraging users to question every single thing they receive.”

Cris Tardáguila, director of Agência Lupa, tested Facterbot for about a week. She told Poynter in a message that she found it pretty useful for a high-level view of some fake news stories.

“The way the bot talks is pretty friendly and it is lovely to have the Spanish option,” she said. “I like the fact that it asks you if you need further information.”

At the same time, Tardáguila said the chatbot does have its limitations. There’s not much information on its about page, which she said hampers its credibility. It also didn’t recognize the names “Obama” or “Lula” and she only received two broadcasts in a week.

Still, she said it’s a welcome addition to the landscape of fact-checking chatbots.

“It is always interesting to have bots like this,” Tardáguila said. “It gives the sense that people are working to fight fake news.”

Support high-integrity, independent journalism that serves democracy. Make a gift to Poynter today. The Poynter Institute is a nonpartisan, nonprofit organization, and your gift helps us make good journalism better.
Daniel Funke is a staff writer covering online misinformation for PolitiFact. He previously reported for Poynter as a fact-checking reporter and a Google News Lab…
Daniel Funke

More News

Back to News