ChatGPT launched at the end of November 2022. Four years on, it is hard to find a corner of Africa’s creative economy it has not reached.
Scriptwriters brainstorm with it. Producers master tracks with AI tools. Advertising agencies draft campaigns in it. Students get flagged for using it. Streaming platforms scan for it, courts argue over it, and Nigerian filmmakers now run entire festivals around it.
This is an overview of what happened: the tools, the platforms, the sectors, the lawsuits, the work, and what comes next. It holds the enthusiasm and the scrutiny together, because both are earned.
The short version is that access widened and ownership did not. African creatives gained cheaper tools and faster workflows. The data centres, the models, the platforms and most of the rule-making stayed elsewhere.
The first year: shock, strikes and a first ad
The early signs arrived quickly. ChatGPT passed 100 million users within months. In 2023, Hollywood’s writers and actors struck for months, with AI protections among their central demands. SAG-AFTRA’s 118-day strike ended with protections against unauthorised digital replication.
In Africa, the first stories were about work and marketing.
Workers in Kenya who helped make ChatGPT less toxic were reportedly paid between $1.32 and $2 an hour. That detail set the tone for what followed. The people who built the safety layer of the world’s most famous AI product were paid a fraction of what its users imagined.
In August 2023, Safaricom, Kenya’s biggest advertiser, unveiled what it called Africa’s first AI-generated TV ad. Kenyan creatives took notice. Corporate contracts are a core income stream for designers, photographers and copywriters, and a campaign made without them was a signal, not a stunt.
The tool race: one chatbot became many
In late 2022, “AI” mostly meant ChatGPT. By 2026, it is a crowded market.
ChatGPT reached 900 million weekly users by February 2026, according to OpenAI. Sensor Tower puts it above 1.1 billion monthly users, the fastest any app has reached a billion.
But its dominance has cracked. Sensor Tower’s 2026 report shows ChatGPT’s share of AI assistant users falling below 50% for the first time, to 46.4% by the end of May. Google’s Gemini held 27.7%, with 662 million monthly users. Anthropic’s Claude held 10.3%, with 245 million. Grok, Perplexity, DeepSeek and Meta AI each hold under 5%.
DeepSeek’s January 2025 launch was a shock. Its monthly users roughly tripled from 33.7 million to 96.9 million in three months, though it struggled to hold ground in the US.
Money followed the users. People were on pace to spend over $4.2 billion on AI apps in the first half of 2026, up from $1.83 billion a year earlier.
Creative tools split off into their own markets. Image generators like Midjourney, music generators like Suno and Udio, and video models like Sora and Veo each built separate audiences.
How much of Africa is using it
Africa’s usage is rising, but unevenly.
Microsoft’s Global AI Diffusion report for the first quarter of 2026 puts worldwide generative AI use at 17.8% of people aged 15 to 64, up from 15.1% a year earlier. The African picture, across the 46 countries it ranks, looks like this:
South Africa leads the continent, and it also sits above the global average. Egypt is close behind. Nigeria, with the continent’s loudest AI conversation, is at 10.1%, level with Ghana and several West African neighbours.
Two lessons sit in that table. First, adoption follows connectivity, language and income more than it follows hype. Second, the story is not Nigeria-shaped. Some of the fastest-moving adopters are in North Africa and Southern Africa.
Creatives are ahead of everyone else
Among creative professionals, adoption is much higher than among the general population.
Adobe and Harris Poll found that 86% of more than 16,000 creators across eight countries use generative AI. In a separate global survey of 1,780 creative professionals, 53% of those in the Middle East and Africa used AI daily, against 44% in the US. That is one survey, not a continental census, so it is a signal and not a statistic to build a policy on.
Among 100 African marketers surveyed, 71% said AI is part of their team’s workflow. Only 26% had formal training.
That last number matters. Africa’s creatives are adopting fast, often self-taught, and mostly without the training, contracts or protections that come with an institutional rollout. The tools are in the room. The rules are not.
How platforms responded
Platforms built AI into their products while tightening the rules around what it could be used for.
YouTube added its own tools, including Veo, Dream Screen and Dream Track. At the same time it began labelling disclosed AI content in 2024. In May 2026, it started applying labels automatically when its systems detect significant photorealistic AI use. Labels are permanent on content made with its own tools, or carrying metadata that marks it as fully AI-generated.
Monetisation rules tightened too. In July 2026, YouTube clarified that generic, mass-produced and repetitive content will not earn, whether it is AI-made or not. It reportedly terminated 16 channels with 4.7 billion lifetime views in January.
For Africa, this matters more than it first appears. YouTube is where many African producers earn, especially in film, and its Partner Programme covers only around a dozen African countries. A tighter rulebook lands hardest on the creators with the least room to absorb a mistake.
TikTok integrated content credentials in January 2025 to detect and label AI content automatically. Meta labels AI imagery across its apps. The EU AI Act’s transparency labelling rules apply from August 2026, and platforms that operate globally tend to apply their strictest rules everywhere.
Streaming went the same way. Netflix has reportedly used AI dubbing, claiming a 15% improvement in completion rates when viewers chose it over subtitles. In May 2026, Spotify signed a licensing deal with Universal for fan-made AI covers and remixes.
Dubbing deserves attention in an African context. Africa has some of the world’s most multilingual audiences, and cheaper dubbing could widen the reach of African films within the continent. It could also remove work from the human translators who do it now.
Music: volume exploded, value did not
Generative music grew faster than anyone’s rules.
Deezer’s daily AI uploads rose from about 10,000 in January 2025, or 10% of uploads, to around 90,000 by June 2026, which is more than half of its new uploads. The share climbed step by step: 28% in September 2025, 39% in January 2026 and 44% in April 2026.
But uploads are not listening. AI tracks remain a small share of what people actually stream, and the volume has attracted fraud. Spotify has removed tens of millions of spam tracks, and Bandcamp banned AI music outright. Deezer now says it will remove AI tracks that have not been streamed in six months or that are tied to fraudulent streams.
The business then turned to licensing. Universal settled with Udio in October 2025, and Warner settled with both Udio and Suno that November. As of May 2026, Sony was still litigating.
The pattern is familiar. When a technology scrapes first and asks later, the industry eventually forces a price on it. That price is being set in New York and Los Angeles. African rights holders, whose catalogues are less visible to the model builders, have much less leverage in those rooms.
Africa’s music story runs in parallel
Sub-Saharan Africa’s recorded music revenue rose 22.6% in 2024 to pass $100 million. That is real growth, from a small base.
Producers across the continent use AI mastering and stem-separation tools as routine. And South Africa has produced its first visible AI hit. “Suka!”, an amapiano-style track by a former podcaster turned music technologist, reached more than 1.2 million streams and climbed to number 10 on Apple Music South Africa after its December 2025 release. Listeners split. Some called it fresh. Others found the vocals mechanical and worried about what its success means for working artists.
That reaction is the whole debate in miniature. The tool works. Whether the work counts is still being argued.
Film, TV and animation: the most visible experiments
Nollywood’s experiments are the most visible in Africa.
Makemation (2025) was billed as Africa’s first AI-themed feature in cinemas. The Naija AI Film Festival drew 490 submissions at its debut. The Abuja International Film Festival’s 2026 edition, which ran from October 20 to 25, took the theme “Frames without Borders: AI, Ethics, and the Future of Filmmaking,” with masterclasses on ethical AI use alongside screenings.
Then there is Mystery of Osu Land, a Yoruba-language science-fiction feature of more than three hours, with about 80% of its visuals AI-generated. Its makers applied for a Guinness World Record in October 2025. Guinness declined in January 2026, because the category had not been formally established. The filmmakers called it an administrative decision and said the film’s reality did not change.
One Nigerian filmmaker claims AI can cut costs from $1 million or $2 million to about $200,000. Treat that as an anecdote, not a benchmark. But it points at the economics that matter.
Earlier this year, we looked at the three cash registers of African film: cinema, streaming and YouTube. A cinema-grade release can cost ₦250 million or more. A strong YouTube title runs closer to ₦9 million to ₦15 million. If AI lowers production costs further, it pushes more producers toward the register that is open to everyone, and it deepens the competition inside it. It makes the open door wider, and more crowded.
The flashpoint was a synthetic actor
Globally, the argument crystallised around synthetic performers. When “AI actress” Tilly Norwood surfaced in 2025, SAG-AFTRA, which represents over 160,000 performers, condemned her as a character built from performers’ work without consent.
Lionsgate, meanwhile, signed a deal with Runway to train video models on its library. Both facts sit side by side. A studio can license its own catalogue to a model builder. An individual actor, or an African studio without a legal team, usually cannot.
Writing, plagiarism and the courts
The word “plagiarism” took on two meanings.
The first is about the models
More than 125 AI copyright suits were active or recently resolved by August 2026. The biggest result so far is Anthropic’s $1.5 billion settlement with authors, roughly $3,000 per book, which received final approval in July 2026.
The lesson courts have signalled is narrow. Acquiring training data from pirated copies is expensive. Whether training itself is lawful is a separate question, and it is still unresolved. The New York Times’ suit against OpenAI and Microsoft was still pending at the time of the research behind this piece, with no ruling on the merits.
African authors, journalists and musicians are not parties to those cases. Their work is in the training data like everyone else’s. They have no settlement and, in most countries, no clear legal route to one.
The second is about writers being accused
AI detectors misfire, and they misfire in a pattern. In a Stanford study, seven detectors flagged 61% of essays by non-native English writers as AI-generated, and 97% of them were flagged by at least one detector.
For African writers and students working in English as a second or third language, that risk is real. South African universities have been debating detector use for exactly this reason. The cost of a false accusation is highest for those with the least recourse.
Slop
Merriam-Webster named “slop” its 2025 word of the year, defining it as low-quality digital content made in quantity by AI. Macquarie Dictionary chose “AI slop.”
Commentators described a “slop economy,” in which volumes of AI content are milked for ad money. Platforms’ responses, including labels, demonetisation and removal from recommendations, are aimed at that economy.
The risk for African creators runs both ways. They can be swept up in crackdowns aimed at mass-produced content, and they can be drowned out by it. A creator with real craft has to prove it is not slop. A creator using AI well has to prove it is not slop either.
Work: freelancers, the gig economy and the people behind the models
For freelancers, the picture is a split, not a collapse.
Upwork’s 2026 Future Workforce Index found that freelancers who use AI earn 34% more per hour than those who do not. But per-contract earnings in generative AI and creative production work fell 13%, which suggests lower-complexity AI work pays less as it scales. Contract starts in that category grew 90% year on year. Note that the survey sampled 2,400 US-based workers, so it describes the platform’s market, not Africa’s.
South Africa’s freelance report calls it a skills split, not a job apocalypse. About 59% of projects are AI-augmented, but creative supply (19.1% of freelancers) far outruns demand (5.6% of projects). Africa accounts for about 9.5% of AI-related gigs on Fiverr, with Nigeria leading.
Counts of African freelancers vary wildly, from the hundreds of thousands to over five million in Nigeria alone, depending on who is counting. That range shows how little reliable data exists. Policy built on guesses will be as unreliable as the guesses.
The hidden workers
The people who train and clean the models are still largely invisible.
In April 2026, Sama issued a layoff notice affecting 1,108 workers in Kenya after Meta ended a large contract with its Nairobi office. Meta has been in court with Sama since 2022 over claims from former content moderators about low pay and inadequate mental health support. The moderators are seeking $1.6 billion in compensation, and the case is ongoing. Sama says it pays four times the local minimum wage and provides extensive mental health support. It has shifted to other work, mainly AI data labelling for Meta.
The detail matters. Africa supplies some of the human labour that makes AI products safe and usable. It captures very little of the value those products generate.
Infrastructure, languages and law
Three structural gaps sit under every creative story.
Infrastructure
Africa holds only 0.6% of global data centre capacity. That is expected to roughly triple to 1.2 gigawatts by 2030, with $8.76 billion of new investment forecast by 2031. Outside South Africa, only about a third of built capacity is utilised.
Capacity that exists but is not used points to a demand and power problem, not only a building problem.
Languages
Africa has over 2,000 languages. The Masakhane community, with more than 2,000 researchers, works on them. Lelapa’s InkubaLM covers five (Swahili, Yoruba, isiXhosa, Hausa and isiZulu). Benchmarks still show a persistent gap versus English and French.
For creatives, this is not an abstract research problem. A tool that writes fluent English and clumsy Hausa quietly shapes what gets made, and in what language.
Law
Law lags behind. Nigeria’s copyright regime has no text-and-data-mining exception. Kenya’s draft Copyright Bill 2026 addresses AI works and training data. South Africa’s fair-use proposal could open training on local work.
Kenya tabled an AI Bill in February 2026, and South Africa submitted a draft national AI policy in March. Nigeria’s AI regulation bill was targeted for passage by March, and I could not confirm that it passed. In one survey, 75% of African media organisations said AI regulation is unclear or absent.
Corporate support exists, but it is small. Google’s AI programme with the Elba Hope Foundation puts $1 million behind about 100,000 creators. That is roughly $10 per creator.
The ledger
What improved
Access to tools that once needed capital or a studio
Lower production costs
Faster localisation and dubbing
A path in for creators without capital
New, better-paid AI-skilled freelance work
What got worse
A flood of low-quality content competing for attention
Unresolved copyright, with African rights holders mostly outside the negotiations
Pressure on writers, translators and illustrators
False accusations against non-native English writers
Thin local infrastructure and regulation
What to expect next
These are projections, not facts.
More automatic labelling and detection, shaped by the EU’s August 2026 rules.
Licensing becoming the norm in music, as in the Udio and Suno deals.
Courts deciding whether training is fair use, with the New York Times case as a bellwether.
Nigeria and Kenya moving from drafts to enforceable law.
Real revenue at risk. A CISAC study projects that 24% of music revenue and 21% of audiovisual revenue could be at risk by 2028, with dubbing translators the hardest hit.
The question underneath
Four years in, the useful question is no longer whether African creatives will use AI. They already do, at rates that match or beat much of the world.
The question is what they will own.
The tools are cheap. The training data includes African work. The labour behind the models includes African workers. The audience is African. Yet the compute, the models, the licensing deals and most of the legal precedents sit elsewhere.
An industry can thrive as a user. It cannot set terms as one. The next four years will be decided less by how well African creatives use AI, and more by whether they get a seat where its rules and prices are set.
Written by Layo
Lead Editorial Writer, Creative Brief Africa
Outside of her editorial work, she writes Curious Health, a newsletter focused on everyday health questions, explored with clarity and care.