A writer can spend years learning how to construct a sentence, only to lose his reputation because someone decides the sentence is too polished to be human.
That is the disturbing possibility raised by the controversy surrounding Nigerian writer Jerry Falade in the US and his debut crime novel, ‘Call Me, I’ll Hide the Body’. The manuscript reportedly received an offer of more than $2million during a 14-way auction before his agents withdrew it, saying they could no longer authenticate how it developed ‘from origin to completion’. Falade denies using artificial intelligence (AI) to write the book and says racial bias shaped the response.
The public may never know every detail. Authors must be truthful about their methods, and publishers are entitled to investigate possible misrepresentation. But this dispute raises a larger question: what happens when ‘parts’ that sound like AI become evidence that a human being did not write them?
That question matters in Lagos, Johannesburg, and Nairobi, and also in Atlanta, New York, and London. It matters to the African novelist seeking an international publisher, the employee whose email is dismissed as ‘obviously ChatGPT’, and the student accused of cheating because a detector assigns a percentage to an essay.
We are entering an era in which people may be required to prove that their own language belongs to them.
Most AI-detection systems do not discover hidden signatures showing that a chatbot wrote a paragraph. They estimate whether language resembles patterns associated with machine-generated text. But predictable writing is not necessarily artificial writing. Students follow essay formulas. Lawyers use standard phrases. Civil servants write in an official register.
In a 2024 test, ZeroGPT labelled the Declaration of Independence (written in 1776) as 97.93% AI-generated. Also, Stanford-led study found that seven widely used detectors repeatedly misclassified essays written by non-native English speakers as AI-generated, but treated essays written by native speakers more accurately. The researchers warned that such systems could penalize writers whose language is less varied or more predictable. Newer tools may perform differently, but even Turnitin (an online service used by universities and schools to check student essays for copied text) says AI reports can misidentify human and machine-written text and should not be the sole basis for adverse action.
Consider an international student in the United States. She has spent years being told that her English must be grammatically correct. She avoids slang, checks every verb and structures each paragraph carefully. Then a detector labels her work artificial because it is controlled and predictable. First, the institution tells her to master standard English through IELTS and TOEFL. Then it treats mastery as suspicious.
For many African writers like Falade, English is both a language and an inheritance shaped by colonial history. In Nigeria, Kenya, Ghana and other former British colonies, generations of students were taught forms of English influenced by British education. Teachers emphasized correctness, formal composition and structure. Those habits do not disappear when people cross the Atlantic. A Nigerian in the US may still write ‘whilst’ instead of ‘while’, ‘amongst’ instead of ‘among’, or ‘kindly’ where an American colleague would write ‘please’. A lawyer in Lagos may use a long, balanced sentence because that style signals a certain level of seriousness. English is diverse and shaped by local languages, cultures and histories, and the idea of what ‘natural’ writing sounds like may be culturally narrow. To an American editor, a phrase may sound unusually formal. To a Nigerian graduate, it may sound like the English taught at school.
Write with errors, and your intelligence may be questioned. Write without errors, and your humanity may be questioned.
There is another inequality beneath this debate. A Time investigation found that outsourced workers in Kenya helped label harmful material for a safety system associated with ChatGPT, with some taking home less than $2 an hour. Their work did not teach ChatGPT British grammar and should not be confused with how language models learn to write. But it reveals the unequal geography of AI. Much of the less visible labor, such as data labelling, content moderation and evaluation, is carried out elsewhere, including in Africa, often for far less money. African workers can be essential when the machine needs to become safer, but African (or black) writers may still approach global institutions as outsiders whose authorship is treated as doubtful. This is about power: who profits, whose labor remains hidden, and who receives the benefit of the doubt.
I agree that publishers must investigate deception. A book contract involves intellectual property, money and trust. If an author submits ‘substantially’ or ‘entirely’ machine-generated material but claims it is entirely self-written, publishers may have legitimate grounds for rejection.
But scrutiny needs rules. A fair inquiry should examine document histories, drafts, research files, correspondence and revisions. It should differentiate between spellcheck, grammar assistance, research, brainstorming, translation, sentence generation and entire chapters. Those are not the same activities.
Publishers, schools and employers should adopt one basic principle: no adverse decision should rest solely on an AI-detector score, a suspicious word or an intuition that prose sounds too polished. People deserve clear rules, transparent evidence, human review and a meaningful opportunity to respond before their reputation is damaged. Otherwise, the real problem is that institutions may begin treating people as imitations of AI.
Artificial intelligence learned from human writers. Its fluency is borrowed fluency. We should be careful when accusing a person of sounding like a machine. Very often, it is the machine that sounds like us.
No writer should have to misspell a word, break a sentence or weaken an argument merely to prove that a human being was present. In the age of artificial intelligence, the presumption of innocence must include something more fundamental: the presumption of humanity.
