ARTIFICIAL INTELLIGENCE (AI) INTEGRATION IN NIGERIAN NEWSROOMS REPORT

Executive Summary

Our survey of one-hundred and two journalists—editors and reporters in print, electronic and digital media—across the six geo-political regions of Nigeria explored how journalists in Nigeria are adapting to generative artificial intelligence tools, their safety concerns, policy challenges, and gender-related experiences. It is a follow-up to the stakeholder convening activity where we gathered actionable insights from newsroom leaders, journalists and Civil Society Organizations (CSOs) to validate preliminary project assumptions regarding newsrooms AI usage and safety concerns. The survey aims to establish a baseline for policy framework and training content as part of key activities under our AIM project. Based on the results of this survey, and insights from follow-up focus-group discussions, we find the following:

Key Findings

• The use of generative AI by journalists is significantly high. Of the surveyed population, 95% are using generative AI tools at least once a week. However, the frequency of usage varies. In a typical work week, 40% of the respondents use generative AI tools occasionally, 36% use it frequently, 13% people say they rarely use it, only 6% use it for almost every story while 5% responded they never used it weekly for their journalistic tasks.
• Our findings show that while journalists are adopting generative AI systems rapidly, there is a knowledge gap as a greater proportion of respondents (36%) have received no training on AI. Only 3% of the respondents are expert users of generative AI tools with confidence and ability to identify AI generated “hallucination” or fake information.
• In more detail, we find:
• • Usage of AI tools by journalists has increased in the last two years. The proportion of people using AI tools in their professional work within the last two years surged to 73% from 21% usage before the last 2 years.
• • ChatGPT is the single most popular AI tool with 83% usage at least once a week. Though ChatGPT dominates active use, weekly use cuts across other generative AI tools. 29% respond using Google Translate, Google Lense, 26%, Gemini 21%, Copilot 18%, Grammarily 16% while Quilbot records 12% and Peplexity 11% among the topmost used AI tools weekly.
• • Only 2% report using Nubia AI tool. Although the tool’s creators say it was built with the African context in mind, its actual use appears limited, likely because it is not widely known among the surveyed population.
• • The need for structured context-specific training on artificial intelligence is evidence. One-third of respondents (33%) reported being entirely self-taught—often relying on trial-and-error approaches—while only 10% indicated that their media organisations had provided any formal training.
• Findings on Journalists’ Safety and Risk Concerns
• Overall, our findings on journalists safety and risk concern show that while most respondents (67%) have not yet been directly targeted by AI-driven threats, they are highly aware of and concerned about AI-related risks—especially around source protection, job displacement, deepfake disinformation, and cultural/linguistic bias in AI tools.
• AI is seen by 59% of respondents as useful for factchecking and reducing errors, yet there is a strong undercurrent of mistrust about how well current AI systems understand Nigerian realities and how safely they handle sensitive information.
• In more detail, we find:
• • Over 60% (Very + Somewhat concerned) express clear anxiety about source exposure and data security when using AI tools.
• • Job displacement is the single most cited threat (37% of the respondents), reflecting anxiety about professional relevance, livelihoods, and the erosion of human editorial judgment.
• • A smaller but important group (14%) fears that AI may introduce or amplify errors, potentially increasing legal risk.
• • There is a significant awareness gap. A proportion of the respondents (21%) reported “Not concerned” (I didn't know this was a risk) when asked if they were concerned that uploading sensitive interview data to AI Tools (like ChatGPT) could expose their sources. Some respondents may be unconcerned not because the risk is low, but because it is not fully understood.
• • A proportion of the respondents (16%) say they avoid uploading sensitive data to AI tools, suggesting a self-imposed ethical or safety boundary, reflecting cautious professional practice.
• Findings on Newsroom Policies and Gaps
• The survey reveals significant structural and ethical gaps in how newsrooms are approaching AI adoption. Most respondents (51%) report an absence of formal policies. There are limited transparency practices, and unclear lines of responsibility. AI governance appears largely ad‑hoc, with individual journalists often left to make ethical decisions without institutional guidance.
• When asked who is primarily responsible for the ethical oversight of AI in your newsroom, 34% say individual journalists are responsible for ethical decisions around AI, with 27% saying no one is responsible. Communication from management—particularly regarding job security and tool provision—remains inconsistent, contributing to uncertainty within newsrooms. Overall, the data suggests that many organisations are using AI without the foundational frameworks needed to ensure accountability, transparency, and security.
• In more detail, we find:
• • A substantial policy vacuum as majority of respondents (51%) says their newsroom has no formal AI guidelines. Only 17% report having a clear, written policy, suggesting that structured governance is still emerging.
• • A combined 32% either don’t know (16%) or rely on informal verbal rules (16%). When we asked, does your newsroom have a written, formal policy regarding the use of AI? highlighting lack of internal communication on AI adoption in most newsrooms.
• • Audience transparency is not yet a priority in many newsrooms, despite growing public concern about AI‑generated misinformation. A higher proportion of respondents (51%) say their newsroom has no rule on disclosing AI‑generated content. 22% note disclosure is encouraged but not enforced, showing partial but weak commitment to transparency. Only 8% report mandatory disclosure. The lack of mandatory disclosure may undermine trust.
• • Ethical oversight is fragmented. 34% of respondents say individual journalists are responsible for ethical decisions around AI. 28% cite management or the Editor‑in‑Chief, 27% say no one is responsible. Only 11% point to the IT/Tech department.
• • Only a minority feel adequately informed about how AI may affect their roles, with 40% of the respondents reporting that AI’s impact on job security has not been discussed at all. 27% report transparent and reassuring communication, while 20% describe management’s communication as vague or unclear. 13% say information circulates only through informal rumours. A lack of structured communication may contribute to anxiety and speculation among staff.
• • Security practices are weak across most newsrooms. A larger proportion of the respondents (67%) say they are expected to use free AI tools, which often lack enterprise‑grade security. A significant number, 27% say they pay for their own professional tools, indicating a personal financial burden and inconsistent access to secure systems. Only 6% report that management provides paid enterprise subscriptions. The reliance on free tools increases exposure to data privacy risks and may violate organisational or regulatory standards.
• Findings on Gender-Sensitive Reporting & Practices
• The survey data shows that most respondents (72%) do not perceive AI outputs as gender‑biased, though 17% of the respondents say they have seen at least occasional stereotyping. Only 11% say it happens frequently. This suggests either genuinely low exposure to bias or low awareness of subtle patterns.
• While a strong majority (88%) believes AI training opportunities are equally accessible across genders, there is less consensus on whether AI‑generated disinformation disproportionately targets female journalists. Additionally, a significant portion of respondents (65%) say they avoid using AI for sensitive topics such as Gender‑Based Violence (GBV), which limits the pool of those who can assess AI’s performance in such contexts.
• Though respondents overwhelmingly view their newsroom training environments as gender-equitable, this likely reflects organisational pattern rather than AI-specific training, since most of the respondents had earlier indicated they had not received any AI training.
• Overall, the findings indicate low perceived risk, low active monitoring, and high confidence in internal newsroom equality, but also limited awareness of emerging harms and hesitation to rely on AI for sensitive reporting.
• In more detail, we find:
• • A greater proportion of respondents (61%) indicate strong interest in adopting gender‑sensitive reporting tools, but actual usage remains low. A significant number (25%) of respondents do not consider gender balance tracking necessary, indicating differing editorial priorities or limited awareness of its value. Only 14% actively use AI for gender‑balance monitoring.
• • While 88% of respondents believe training is equally accessible to men and women, 10% say men receive more opportunities, while 2% say women are prioritized.
• • A majority (45%) do not believe female journalists face greater AI‑driven disinformation threats. However, the high “no opinion” rate (27%) indicates uncertainty or lack of exposure to this issue. The combined 17% who agree or strongly agree represent a notable minority who perceive gendered digital threats.
• • Most respondents (65%) avoid using AI for GBV‑related content, likely due to trust concerns or editorial policies. Among those who do use AI, experiences are mixed: while nearly a quarter (23%) find the tone generally appropriate, over one in ten (12%) report frequent insensitivity, highlighting a risk area requiring attention.

Recommendations

This survey was conceived to inform our training curriculum and to guide the development of newsroom policy frameworks. While these objectives could also serve as recommendations, we have already begun implementing them in practice. As we actively pursue these objectives, we hope that media organisations, academic institutions, and other stakeholders in Nigeria and across the globe will find this work a valuable resource as they navigate the rapidly evolving intersection of journalism, media, and artificial intelligence.