Read part Part I here

The Unresolved Tension:

With a fair measure of grace and discernment, I would probably permit a few remarks that are meant as flattery now and then, provided they are motivated by harmless intent and benign ignorance, and cause little damage to the recipient rather than genuine disrespect. Conversely, fluency does not garner more admiration than suspicion; I believe they are equally balanced.

I have debated with people who believe the opposite, highlighting global careers, academic achievement, and professional mobility, and firmly asserting that fluency opens more opportunities. However, I believe my  argument remains valid because it considers both realities fairly. For instance, admiration for language fluency carries many negotiated risks, and it’s often conditional, situational, and revocable. Especially when you consider that the same fluency that garners admiration in certain contexts and approval in institutions can, anecdotally, also unsettle close kinships and other relationships, as well as complicate one’s sense of identity within their cherished sense of belonging.

Black English speakers who command the language at a high level often report a subtle shift in perception within their community and beyond, as well as among their peers. Often perceived as ‘Different,’ ‘an exception,’ ‘Not like the others, ’ ‘not Black enough,’ and ‘Not like us’, because colonial languages, in fact, often redraw racial boundaries rather than dissolve them.

I have been told by someone of non-white ethnicity that I am ‘a white man in a black man’s body’. This policing of  ‘Blackness’ – as contentious and context-specific a concept that is – is not limited to speech. I have also watched a highly accomplished Black-British dancer of West African descent be told at a party, by another Black person of African descent, that they ‘dance white,’ even as hip-hop and African music filled the room.

The irony in that remark was both ridiculous and mind-blowing, since it came from someone who was not a professional dancer. Moreover, when you factor in the language in which it was uttered, it takes a painful turn. As soon as the comment landed, it drew a boundary around who was allowed to embody certain features of Black authenticity – whatever that means; certainly not the brilliant dancer who spoke with a British inflection. That moment was no longer about dancing skills and more about perceived cultural alignment, speaking directly to kinship and recognition: how fluency, movement, and expression—whether linguistic or physical—as illustrated in this video can trigger suspicion when they do not conform to transposed expectations. Once again, it demonstrates how unresolved inherited (and in some respect, inverted) colonial bias serves to invalidate experience with an audacity that questions identity and belonging.

Similarly, this is the same battlefield in which the African-English fluency finds itself immersed. A battlefield laced with internalised complexities and contradictions, caught between being celebrated in professional spaces and policed in social ones, at every turn, dictated to by the expectation of others, rather than our abilities. Sometimes mocked in popular culture and imitated by machines, yet our everyday speech retains the memory of what history tries to forget; a memory that still influences how African speakers sound and how they are heard today.

This tension suggests that the structural dynamic of colonial classrooms, that once disrupted the survival of many indigenous languages all over the world, never truly vanished. Instead, they have evolved, entering into media expectations, academic institutional norms, and the algorithmic codes and designs of new technologies like AI spaces, where fluency imports the oppressive weight of history, but not freedom. Thus, it creates an even greater burden on the intellectual capacity of the black body. So, if you are still reading, ask yourself: Does Black mastery of language still require explanation?

Is AI Grammar Racist?

Let’s return to the pending story introduced in Part I, in which a friend’s university submission was questioned for suspected AI-assisted writing. For an African intellectual, moments like these carry a concerning weight, as they place an additional burden on the Black body. It subtly delegitimises our capacity to produce knowledge without incurring suspicion, scrutiny, or proof of authenticity—forms of attention rarely imposed on other bodies. Essentially, African English speakers who deploy intricate grammar and intriguing vocabulary, already encounter reactions that white or Western speakers rarely do. Fluency, in those cases, in fact appears expected, nay, ‘natural.’

However, the tension grows as African English becomes integrated into modern technologies, especially AI. It is confirmed that African linguistic registers have been used to train AI systems, yet those same systems increasingly contribute to the delegitimisation of African speech in the real world. In 2024, I read an article published in The Guardian—TechScape: How Cheap, Outsourced Labour in Africa Is Shaping AI English—which highlighted how biased training data causes AI language patterns to mirror African professional English. Thus, creating a feedback loop in which African speech styles are recognised as ‘artificial’, while AI outputs are treated as neutral or authoritative. The irony here is as comical as it is problematically structural, to say the least. It is also difficult to ignore when viewed historically.

African intellectual voices most prominent in global discourse often come from graduates of colonial or postcolonial education systems. Their mastery of English—meticulously refined—once brought pride, praise, pay, access, success, and mobility. It also drew suspicion, envy, and resentment. Those same linguistic standards passed down through generations are now used to train artificial intelligence. Yet it is increasingly seen as separate from ‘natural’ spoken rhythms, especially when compared with modern Western speech, including vocabulary, speech patterns, accent, and cadence. This process is reinforced by the nature of AI training data itself. English-dominant AI datasets overwhelmingly emerge from elite domains: academic journals, government archives, corporate communications, and digitised colonial records. Essentially, spaces that privilege a specific register—formal, abstract, institutionally rigid and fluent—are long familiar to African professionals navigating global systems of power.

While writing this piece, I questioned how to preserve or defend African-English fluency. I realised, however, that the answer is not about defence. African-English does not need defence or justification; instead, it requires recognition without suspicion, especially in digital spaces. It also calls for better-designed technologies—systems that rely less on ingrained patterns and more on contextual understanding. These improvements are ultimately necessary to fix errors, decrease reliance on habits, and remove unwarranted scrutiny and suspicion that penalise the intellectual creativity of black identities in the digital space. Especially since, at present, AI can only recognise repetitive patterns; it is not (yet) able to identify bodies or reason logically. And here lies the problem.

Though AI is purely technical, at scale, however, the patterns it detects can become prone to biased interpretations or racialisation. The flagging of certain lexical choices as AI-assisted writing by scanners and the assignment of flawed meanings to them, pose a serious risk to those whose credibility has always been questioned. For example, writers of African descent whose ability to deploy colourful vocabulary are increasingly cautious in their use of terminologies, such as ‘delve,’ ‘explore,’ ‘tapestry,’ and ‘leverage,’ in their work because of how often they appear in the outputs of AI systems like Chat GPT, compared to the wider internet. If escalated, any literal work that contains any form of polished, abstract, or industry-relevant vocabulary can easily be labelled as ‘AI-ese’. Thus, delegitimising the intellectual and professional effort, skill, and competence behind the outcome.

However, for those writers, their style did not develop in a vacuum. In fact, one of the digital spaces where sophisticated and vibrant vocabulary has long been common is the African web, especially in Nigerian business and professional environments, which are influenced by colonial and postcolonial education systems that rewarded lexical diversity, grammatical accuracy, and a broadened vocabulary.

What follows is a stark inversion. English shaped under colonial pressure—often mocked as excessive, performative, or inauthentic when spoken by African speakers—now circulates as ‘neutral’ intelligence once detached from the speaker and voiced by machines. African-English has contributed to the development of artificial intelligence, yet African speakers are still questioned about that same fluency. The outcome is one of sombre indignity, as AI gains linguistic authority. If ‘AI-ese’ sounds like African-English, then African-English is redesigned as artificial — not because of its lack of humanity, but because its humanity has been abstracted, standardised, and deemed acceptable only when no Black person speaks it.

Read Part III here

The original version of this series first appeared as a long read on Substack. Follow Anthony for more insights.