Aos Fatos launches tool to combat disinformation in health ads
New digital solution analyzes advertising on major Brazilian news portals
The fact-checking website Aos Fatos has launched Check-up, a tool that collects and analyzes advertisements on major Brazilian news websites. The goal is to assist in analyzing health disinformation circulating in content recommendation boxes provided by platforms such as Taboola, Outbrain, and MGID.
With an open repository on GitHub, the solution was funded by Codesinfo – Innovation Fund for Combating Disinformation, an initiative by Projor sponsored by the Google News Initiative.
“Check-up was designed to understand how health disinformation is spread through ads displayed on major news websites,” explains Bruno Fávero, innovation director at Aos Fatos and project manager.
Aos Fatos itself used the tool, a Python library, to collect and analyze more than 240,000 ads from nine Brazilian portals.
How it works
The tool consists of three main modules: a crawler to collect links from specific sites, a scraper to capture and archive detected ads, and a thematic classifier based on an advanced language model (LLM).
“Check-up is a scraping and data extraction tool that can be used for investigations or research on native ads circulating on the internet. The code can be easily adapted to work with sites outside the initial scope of the project,” explains Fávero.
During development, the team had to overcome technical challenges. “One of the difficulties we faced was bypassing ad platform restrictions designed to prevent ad collection,” says Fávero. To address this, the team added VPN support to Check-up.
Since the tool requires programming knowledge to use, Aos Fatos has also made the raw data collected available for download, allowing researchers and journalists to access it even without programming skills.
The project also has plans for expansion and improvement. “We want to increase the number of websites covered by the tool and develop a classifier that predicts the risk of health disinformation in an ad,” Fávero reveals.
He also encourages the participation of the developer community: “Other developers can adapt Check-up’s code to scrape ads from more websites, create different classifiers for the ads, and develop an interface.”
