[mc4wp_form id=33047]
The numbers were clear, compelling, and perfectly aligned with the narratives and indices built by Shehu Aminu, an investigative journalist based in Sokoto, North West Nigeria. His investigation into the prevalence of Gender Based Violence in Sokoto was taking shape, until he realised it required some credible statistics on reported domestic violence cases in marriages over the last 5 years.
With little access to public data from local authorities, he had turned to Kimi AI, an advanced artificial intelligence tool to fill the gaps, an option now available to journalists in inaccessible data environments.
But what started as a breakthrough quickly spiralled into doubt. The statistics, although precise and persuasive, could not be traced back to any verifiable source. The AI could not show its origin. When he asked for references, the response was “unfortunately, I cannot provide a specific URL to the data mentioned”.
“At that moment, I realised I was confronting a new threat to journalism, not disinformation from outside actors, but potentially fabricated data from the very tools designed to enhance our reporting,” Aminu said.
Data Inaccessibility in Nigeria
Across Nigeria newsrooms, Artificial Intelligence tools are being adopted to assist with research, data analysis, and even content generation. In a landscape defined by poor access to public records, unreliable government databases, and the chronic underfunding of data-driven journalism, AI offers an enticing shortcut.
Between 2023-2024, a survey of 245 Nigerian ministries, departments, and agencies (MDAs) revealed that only about 1.22 per cent fully complied with the Freedom of Information Act (FOIA), while over 150 MDAs consistently flouted information requests. From 2020-2022, the majority of MDAs scored below the benchmark for responsiveness, with 169, 173, and 153 MDAs, respectively, failing to respond or responding late to FOI requests.
Sourcing data in Nigeria, especially for investigative purposes, may be difficult in a country where governance affects people so directly on issues like fuel pricing, education, security, and internet access, data cannot be open, accessible, or trustworthy, said Muhammad Bello Buhari, a digital rights activist.
But in reality, credible government data is often locked behind bureaucratic secrecy, scattered across unresponsive portals, or outrightly unavailable. Even where data exists, it’s either outdated, politically massaged, or buried in practically unusable formats. He continued.
The growing inaccessibility of credible, real-time government and institutional data in Nigeria has created a major challenge for journalists, particularly those working on investigative stories.
For Buhari, the media has a role to play in making data accessible. “This is a call to push back harder. Journalists and editors must demand access, name agencies that obstruct data, and collaborate with civic tech groups to build open repositories. We need more data journalism, not less”.
A Growing Dependence
Despite the central role of public information in ensuring transparency and accountability, journalists often face bureaucratic bottlenecks, outdated portals, or outright data denial when seeking official statistics or reports.
This data vacuum undermines the depth and accuracy of their reporting, weakening the watchdog role of the press in a democratic society. As a result, many journalists are increasingly forced to seek alternative sources to fill these gaps.
One of the most significant alternatives has become artificial intelligence-generated data, which can simulate, aggregate, or infer patterns based on available inputs. However, this reliance introduces new concerns about accuracy, context, and ethical standards.
While AI tools can support data-driven storytelling, they are not substitutes for verified public records. This shift raises pressing questions about journalistic integrity, public trust, and the need for stronger data transparency laws to enable reporters to do their jobs effectively and responsibly.
“AI-generated data is becoming a valuable tool in journalism. It helps reporters to quickly analyse information, spot trends, and even draft content. But while it offers speed and scale, it is not foolproof.
AI can make mistakes, reflect biases in the data it is trained on, or produce results that sound convincing but aren’t entirely accurate”.
“That is why human judgment is still essential. Journalists need to fact-check, add context, and ensure the final story meets the standards of truth, fairness, and responsibility that audiences expect”, notes Abdallah el-Kurebe, Editor-in-Chief of ASHENEWS, a Nigeria-based news outlet with a focus on research reporting.
Hammed Abdulrasheed, a Nigerian journalist, said he had once relied on AI-generated data while working on a story. However, he was quick to add that he did so with a firm condition of always demanding to know the source of the data before using it.
He explained that tools like Perplexity AI, which aggregate open-source information, often provide external links to their data sources. “Perplexity AI, for example, is good for research. When you request data, it brings links to where it got the information, which is very important.” However, he noted that AI can occasionally provide inaccurate information and emphasised the importance of cross-checking.
“The best way to correct any potential inaccuracy is by requesting the original source, visiting the links, and personally verifying the documents. Only then can you decide whether it’s reliable enough to be used in a report”
The appeal is easy to understand. Tools like Meta, ChatGPT, DeepSeek, Kimi AI, and other generative AI platforms provide journalists with quick summaries, estimated data ranges, and language support in seconds. However, they often do so without transparency about their data sources, if such sources exist at all.
“Most generative AIs are not connected to live datasets,” explains Ibrahim Agua, a data analyst and AI expert. “They make predictions based on patterns, not on fact-checked, real-time information.”
He pointed out that Generative AI tools, such as ChatGPT, Copilot, and Gemini, can produce misleading outputs, often presenting fabricated data as authentic. “This issue is so prevalent that it is referred to as ‘hallucinations’ in AI-generated content”.
A Dangerous Shortcut To Online Research
Like in Aminu’s investigation, when this reporter prompted AI to provide data on the subject, it offered seemingly clear and detailed information. However, to verify his claim, this reporter further asked the AI to cite the source of its data, and once again, it failed to provide any.
For Abdallah, the issue began when he used ChatGPT to convert a PDF document into text. To his surprise, the system generated names that did not exist in the original file. When he flagged the error, the AI admitted to the mistake.
This incident left him concerned about the ease with which false information can slip through, especially in contexts where accuracy is non-negotiable.
“In the absence of primary data, AI fills the void with best guesses, not hard truths; it is a dangerous double-edged sword,” Buhari added.
He opined that while generative AI tools can assist with summarisation or trend spotting, they are only as good as the data that they were trained on. So, when journalists lean too heavily on AI to generate facts or simulate official data, they risk spreading inaccuracies, especially in sensitive areas like election reporting, public health, or national security.
He sees data as a public power, and governments must stop hoarding it like private property.
“We need a bold shift toward radical transparency where public data is proactively released in machine-readable formats, backed by open APIs, and subjected to routine audits and laws like the Freedom of Information Act must be more than decorative; they need enforcement”.
The Ethical Dilemma
At its core, journalism is about truth. Every statistic quoted, every trend visualised, must be traceable. Yet in an under-resourced environment where reliable data is a luxury, AI can easily tempt reporters with its speed and convenience.
Yahuza Bawaje, another journalist, confirmed that his former newsroom used AI to illustrate images for stories on acts of kindness and gender stereotypes. In some cases, they also relied on AI tools to transcribe English audio interviews into text. However, he emphasised that these outputs were rigorously cross-checked to avoid inaccuracies, especially since many of their interviewees were not native English speakers.
Bawaje also sources data from AI tools like Gemini and Copilot by prompting them with specific queries.
“These tools then generate links from various online sources, which I carefully review to determine whether they contain exactly what I need and whether the sources are credible”
This approach, he said, saves him considerable time compared to manual Google searches, and it has made his reporting more efficient. Still, Bawaje has never fully relied on AI-generated data. He consistently adheres to comprehensive ethical guidelines, either those established by his newsroom or by other reputable journalism organisations.
Buhari sees the problem not as one caused by AI or journalism itself, but by systemic failures rooted in the government’s inability to uphold the right to information. “AI is not the enemy here,” he said. “The real problem is the vacuum of official, verified, and accessible public data. Until that is fixed, journalists, researchers, and the public will continue to grasp at whatever tools they can, even flawed ones.”
Some journalists’ cautious approach reflected this concern. Despite their use of AI, they consistently emphasise the need to verify the sources behind any data generated. “Only then can you decide whether it’s reliable enough to be used in a report,” Bawaje adds, underscoring the continued importance of human oversight in digital reporting.
This dilemma is not unique to Nigeria. A 2023 study published in Media, Culture & Society discusses how journalists in Kenya and Ghana grapple with the use of AI in newsrooms, particularly regarding its reliability and ethical implications.
It stated that while AI offers efficiency, journalists in countries like Nigeria have expressed concerns over misinformation, lack of data transparency, and editorial pressure to adopt AI tools without sufficient training. In Kenya and Ghana, similar concerns have arisen. But the stakes are higher here, where fewer journalists have access to research databases, data analysts, or institutional support.
Holding The Machines Accountable
While many journalists are becoming dependent on AI, few have been trained on its limits.
There is a need for capacity-building on Artificial Intelligence literacy in the media. Journalists should treat AI-generated content the same way they treat a political press release: with caution and scepticism.
The solution, Agua, is not to abandon AI but to adopt safeguards, including cross-verifying AI outputs with known databases such as NDHS, NBS, and WHO, using AI only for leads, not conclusions and demanding transparency from AI developers about how models generate statistics, also labelling unverifiable figures as estimates, not facts.
Tech developers, meanwhile, have a responsibility to build a trust infrastructure, not just tools. That means investing in open-source platforms, verification algorithms, and partnerships that support local data ecosystems instead of importing opaque AI systems with foreign biases. Buhari added
“Tools like Dubawa, developed by CJID, offer journalists in low-resource settings a chance to bridge gaps with verified, localised facts,” says Hameed.
– This report was produced with support from the Centre for Journalism Innovation and Development (CJID) and Luminate
