A forwarded message can outrun the truth before anyone thinks to check it. Here is why, and where AI is already helping fact-checkers catch up, from Nairobi to Lagos.
Early in the pandemic, a message raced through WhatsApp groups in Kenya claiming that a baby had revealed the cure for Covid-19: drink sugarless black tea at dawn. It was false. But it was forwarded in good faith, carried by fear and by respect for one’s elders, and it travelled far faster than any correction could. The same pattern repeated across the continent and beyond. In Nigeria, false messages that an anti-malaria drug could prevent the virus sent people rushing to pharmacies, and some overdosed and were hospitalised. No politician, no grand conspiracy. Just ordinary people sharing something that felt true.
That is the problem in miniature. It is not that we lack information; it is that a convincing falsehood is simply easier to pass on than a careful fact. And the gap between the two is widening.
Why the false travels faster
This is not a quirk of any single platform. In a landmark 2018 study in Science, researchers at MIT tracked true and false stories across Twitter and found that falsehoods spread faster, reached more people, and ran deeper into networks than the truth, and not by a small margin. The main driver was not bots. It was us. Novelty and outrage make a false claim more shareable than a sober correction. A lie fits in a single line; the truth needs a paragraph, a source, and time that few of us spend.
Why human fact-checkers cannot keep up
Professional fact-checkers do painstaking and often thankless work, and there are now hundreds of them, from Africa Check in Johannesburg to Dubawa in Lagos. But verification is slow by design, and the feed is not. During the pandemic, African fact-checking groups in the World Health Organization’s Africa Infodemic Response Alliance debunked more than 1,300 false reports, and still the rumours kept coming. In a country like Nigeria, with more than 200 million people, the sheer volume can swamp any newsroom.
The same tools, pointed the other way
Here is the uncomfortable twist. The technology that now manufactures fakes, cloned voices, invented quotes, synthetic images, is the same technology that can be turned toward catching them. Point it at detection instead of deception, and it begins to close the speed gap.
What already exists, and how it is actually used
Some of the groundwork is years old. Google’s Fact Check Tools gather published fact-checks from around the world into a single searchable index, using a shared tag called ClaimReview that also lets software match a fresh claim to one already debunked. ClaimBuster, built at the University of Texas at Arlington, scores sentences for how ‘check-worthy’ they are and has been used to flag claims live during US presidential debates. NewsGuard’s analysts rate the reliability of news sites, including a running catalogue of Covid-19 and vaccine ‘superspreaders’, and, more recently, of sites that publish AI-generated misinformation.
But the most telling examples are not in Silicon Valley. Since 2019 the UK charity Full Fact has built AI tools together with Africa Check and Argentina’s Chequeado, funded by a Google.org grant, that scan news sites and transcribe live radio and television to surface claims worth checking. In the run-up to Nigeria’s 2023 election, Africa Check, Dubawa and FactCheckHub used those tools to monitor broadcasts and flagged dozens of checkable claims within days; in 2024 Africa Check used the same AI transcription to keep pace with South African parties’ manifesto launches. Dubawa went a step further and built its own chatbot, now embedded in WhatsApp, that answers questions in Nigerian English and Pidgin, the languages people actually argue in. In East Africa, PesaCheck, part of Code for Africa, runs similar tools across more than twenty countries. None of these replaces the human checker; they hand her a head start.
If AI can generate a lie in seconds, it should help us question one just as fast.
The hard part, seen from outside the West
None of this invites easy optimism, and the view from beyond the West makes the risks sharper. AI models still hallucinate, stating false things with total confidence, which is alarming in a tool whose entire job is accuracy. They inherit the biases of their training data, and most of that data is in English. As Dubawa’s engineers put it, global models may not give you good information about Africa, which is why they trained their own on local fact-checks and Pidgin. The tools also lean heavily on Big Tech money and infrastructure, a dependence that worries researchers who want the technology without the strings. And where digital literacy is lower, a false claim can do more harm, not less. For now the honest word is ‘semi-automated’: the machine narrows the haystack, a person still has to find the needle.
Your best fact-checker is still you
While the tools mature, the most reliable check remains the person holding the phone, and the evidence for that is strong. In a 2023 study across 16 countries, researchers found that simply nudging people to think about accuracy before sharing made them noticeably better at telling true headlines from false ones. Africa Check’s own rule of thumb, learned from a year inside Kenyan WhatsApp groups, is blunter: if a message has been forwarded, it is probably unverified. So check who is saying it. See whether anyone credible reports the same thing. Notice when a post seems built to make you angry, because anger is what travels.
Closing the gap
The next decade will produce more synthetic content than all of human history before it, and plenty of it will be good. The point was never to stop that. The point is to make sure our ability to check what we see grows as fast as our ability to fake it, and that this ability reaches Lagos and Nairobi as surely as London and New York. The speed gap is where the damage lives, and closing it starts with two small words.
Pause, then verify.

