Fact-checkers are experts at thorough research and careful writing, but many confessed Thursday that they struggle with thinking on their feet and venturing out of their comfort zones.

Eleven fact-checkers presented their organizations’ biggest successes and failures to kick off the third annual Global Fact-Checking Summit in Buenos Aires. Most of the success stories underlined what fact-checkers are good at — checking facts and making an impact on political discourse. But the failures were especially compelling, candid stories about their challenges and mistakes.

Speed can be as important as thoroughness during a crisis, when statements are fresh in everybody’s mind, but not all fact-checkers have the resources or the ability to live fact-check events. In the case of Dogruluk Payi, a government prohibition on immediate publication restricted their ability to live fact-check disasters and explosions in Turkey.

“When something happens in Turkey, we have publication bans before the ambulance or the police [comes],” said Batuhan Ersun, the Turkish site’s coordinator. “The first thing they do is to put the publication down.”

Other fact-checkers said they struggled to distill their messages into brief nuggets of information that could stick in people’s minds — particularly on television.

Lupa, a Brazilian fact-checking agency, frequently uses data analysis and graphics in its web fact-checks, but Lupa’s director, Cristina Tardaguila, discovered it’s difficult to do that on TV when she made her first appearance in February.

Tardaguila went on Globonews — which established a partnership with Lupa this year — to explain a fact-check on the severity of a microcephaly outbreak, along with Marcelo Castro, Brazil’s Minister of Health. Tardaguila explained how the national government was dramatically understating how many cases had been reported and challenged Castro’s defense of his department’s reporting, but her lack of experience on TV made it hard for her to debate a government official skilled at damage control.

Tardaguila said it is challenging to make healthcare records interesting on the screen without the ability to show data tables and infographs to the audience.

“The minute the show ended, I realized how I could have done it better,” she said. “To be better requires someone from TV by my side translating my article into television.”

Most fact-checkers are trained in journalism and news writing, but TV appearances test their ability to concisely summarize their work under the bright lights of a live telecast. TV offers great opportunity for fact-checkers to broaden their audiences, but they need to master a new set of skills.

Other failures discussed at the Buenos Aires conference included several efforts by fact-checkers to expand with new products. PolitiFact tried to launch a podcast last year with the help of the E.W. Scripps radio and TV chain, but editors discovered it was hard to turn a compelling fact-check into a suspenseful conversation when it simply consisted of two reporters who are both journalists for the organization. Without an outside interviewer asking the questions, there was no intrigue.

“The setup didn’t lend itself to one crucial element—drama,” said Katie Sanders, the deputy editor of PolitiFact. “With us playing both the host and the guest, it didn’t lend itself to those kind of surprises.”

Chequeado, a fact-checker in Argentina, tried to build an app in 2014, but there were no major events that galvanized enough interest in the project and not enough money to develop the app and carry it through to completion. Chequeado released the app to the public, but abandoned the project after it did not attract many users. Laura Zommer, Chequeado’s executive director, noted that there was not enough public interest in the app to allow it to take off without the necessary supply of resources.

“We thought we were better than we are… If you decide to do something technologically, you have to be sure there are interests outside your newsroom in that project,” Zommer said. “Never develop without enough money – at least for one year or so – to experiment.”

In addition to the confessions about failures, the fact-checkers also shared success stories. Full Fact broadened its audience by using Wikipedia in a creative way. The British fact-checker trained volunteers to edit and correct more than 20 Wikipedia articles about whether the UK should leave the European Union.

PolitiFact raised money from crowdfunding to annotate the State of the Union address live and send journalists to Iowa and New Hampshire before the caucuses and primary election. Lupa live-tweeted to essentially annotate television coverage of how members of Congress voted during the impeachment process in April.

Speed was of the essence for Dan MacGuill, a fact-checker for the Irish site TheJournal.ie who quickly debunked a social media post about low parliamentary attendance for a mental health debate as the post was beginning to go viral.

“It was anti-viral…It took a claim that had kind of taken over the Internet in Ireland for a short time and it kind of hijacked it.” MacGuill said. “Because I lived in the U.S., I could publish it by 3:30 the next morning, and that was particularly important for viral claims that have such short lifespans.”

Correction: an earlier version of this article contained a typo. The director of Lupa, the Brazilian fact-checking site, is Cristina Tardaguila not Cristiana. We regret the error.