In the final weeks of the 2024 presidential campaign, a video of Dr Matthew Opoku Prempeh, then the New Patriotic Party’s running mate, appeared on social media. He stood before a crowd of small‑scale miners and pledged that seized excavators would be returned to them, allowing the workers to resume their livelihoods.
The clip spread rapidly across WhatsApp groups, instantly polarising supporters and environmental activists. Within hours the NPP issued a statement, and journalists began reviewing the footage from multiple angles. A sign‑language interpreter was present, and two independent interpreters offered differing translations, deepening the confusion.
Ghanaian journalists, lacking forensic tools, escalated the clip to WITNESS’s Deepfakes Rapid Response Force, a global panel of media‑forensics specialists. Four expert teams examined the video and reached a unanimous verdict: there was no evidence of AI manipulation. The clip was real.
That conclusion, however, highlighted a broader issue. The real weapon was not a deepfake; it was the very possibility that anything could be fabricated. WITNESS noted that in 2024, a third of cases escalated to its rapid‑response team involved politicians attempting to discredit authentic material by claiming it was AI‑generated.
In Ghana, political information travels through WhatsApp groups rather than through traditional newsrooms. Over 90 % of internet users on the continent use the platform, and voice notes become the native format—intimate, forwarded, unindexed, and unfalsifiable by anyone outside the group. A fabricated clip can reach an uncle, a pastor, or a group admin whose word carries more weight than a government statement. By the time debunkers like GhanaFact or Dubawa publish a correction, the message has already been reshaped and forwarded through dozens of groups.
Trust in Ghana is largely interpersonal, not institutional. Where institutions have a history of mistrust, people rely on relatives and community elders. AI‑generated misinformation exploits this by mimicking local accents, proverbs, and cultural references, making a deepfake feel authentic to the target audience.
Attempts to counteract fake news through labels and watermarks have failed. During the 2024 election cycle, fake “news cards” with forged branding proliferated. Newsrooms responded by stamping fakes with a “FAKE NEWS” watermark, but partisan actors quickly hijacked the stamp, re‑issuing unflattering content with the same label. The instrument of verification became an instrument of confusion, as the public could no longer trust the source of the label.
South Africa’s Electoral Commission has recently introduced a Disinformation Code ahead of the November local elections, mandating that parties label AI‑generated material and correct false claims within 36 hours. Ghana’s experience suggests that without a shared belief in who can declare something false, such measures may be ineffective.
Moving forward, the challenge for Ghana is to rebuild trust in both institutional and interpersonal channels, ensuring that misinformation—whether deepfaked or not—can be identified and corrected before it spreads unchecked.
