Keeping Pace with Misinformation: Why We Built an AI Team

When Full Fact was founded in 2010, fact checking was largely a manual process. Fact checkers listened to speeches, combed through documents, scanned newspapers and wrote up their findings. That work remains vital today. But the sheer volume, speed and variety of misinformation has changed.

False claims now spread instantly on social media, in live broadcasts, in podcasts that run for hours and in short-form video across multiple social media platforms. Even the most committed fact checking team can’t listen to everything, read everything and respond quickly enough.

That is the gap Full Fact AI was created to fill.

From experiments to a dedicated AI team

As early as 2016, Full Fact began testing machine learning to help spot checkable claims in political debates. By 2019, winning the Google.org AI Impact Challenge gave us the resources to build a dedicated AI team and scale this work globally.

The aim was never to replace fact checkers. Instead, the goal was to use AI to do the heavy lifting at the front of the process - transcribing, detecting, matching and monitoring claims - so that human fact checkers can focus on what they do best: verifying evidence and sharing their findings with the public.

What the AI team builds

Our AI team develops tools that:

• Transcribe and extract claims from audio, video, and text - surfacing statements that fact checkers might want to investigate.
• Prioritise and rank content so teams spend less time scrolling and more time checking what matters most.
• Spot repeat claims. When a falsehood resurfaces months later, or in another country, fact checkers get an early alert.
• Support multilingual, global monitoring so our partners across Africa, Europe and beyond can adapt tools to their own contexts.

These systems already process hundreds of thousands of sentences per day and have supported fact checkers in over 50 countries and in elections from Nigeria and the US to the UK and South Africa.

Why in-house AI matters

We could have outsourced this work. But we built an internal AI team because:

• Fact checking is different: Off-the-shelf AI isn’t built for the nuances of misinformation. It takes domain expertise and close collaboration with fact checkers to get it right.
• Trust and transparency is vital: We need tools that are robust, explainable and designed to serve the public interest.
• Global collaboration matters: Our partners need adaptable tools, not one-size-fits-all systems. Building in-house means we can listen, learn and update quickly.

Preparing for what comes next

Full Fact started an AI team because fact-checking must keep pace with the speed, scale and formats of modern misinformation.

AI tools help us find, filter and follow claims at scale. Human fact checkers provide the expertise, judgement and accountability. Together, they give us a fighting chance to make public debate more accurate and fair.

We’ve just launched a new LinkedIn page for Full Fact AI where we’ll share updates, insights and stories from our work at the intersection of AI and fact checking. If you’d like to follow along, join the conversation and see what we’re building next, follow us on LinkedIn.