The first time I opened the notebook, my grandmother’s voice still lived in the margins. Not her actual voice—she has been gone seven years now, her Yoruba long softened into the Ibadan earth—but the rhythm of it. The way she’d start a story not at the beginning but somewhere in the middle, where the heat of the thing already breathed. When I reached Lagos, the water had already entered the market. I wrote that sentence down and felt, for a moment, that I’d caught something. Then I turned the page and found only blankness. What came next? What did the water smell like—diesel, rotting plantain, the salt-sweat of bodies pressed into a danfo bus? Who was with her? I’d heard the story a dozen times as a child, half-listening while shelling groundnuts, and now the details had fled.
This is the specific grief of the second-generation archivist: you inherit the weight of a story without the full body of it. You remember what the telling felt like—how your grandmother’s hands moved when she described crossing the Niger, the startled pitch of her laugh when she remembered a narrow escape—but the sequence, the names, the dates, they slide through your fingers like river water. You’re left with fragments. A recipe for ofada stew scrawled in your mother’s handwriting on the back of a church bulletin. A parable about a tortoise and a calabash you can only half-recall. The knowledge that your grandmother once hid her savings inside a headwrap, but not the reason why.
For years, I told myself I’d write it all down when I’d gathered enough. When the fragments cohered. When I could sit with my aunties and record their voices, transcribe every pause, footnote every proverb. But my aunties are aging too, and the pandemic taught me that waiting for the right moment is a form of forgetting dressed up as patience. So last autumn I began drafting the first chapters of a memoir about my grandmother’s migration from Ondo State to Lagos, and from Lagos to London, and back again. I sat at my desk with the notebook, the half-remembered parables, the recipes, and the heavy silence where her voice used to be. I could not see the shape of the thing.
I had scenes. Vivid, embodied scenes. The morning she left Ondo, walking barefoot to the motor park because her sandals had split and she would not ask her stepmother for new ones. The afternoon in Brixton when she saw snow for the first time and thought it was ash falling from a hidden fire. But between the scenes, nothing. No connective tissue. No architecture. I knew the house of the story but could not find the rooms.
This is where I need to tell you about the machine.
I typed the fragments into a document—four thousand words of half-memory, oral history, and recipe instructions that read like poetry—and fed them into an AI book generator. Not to write for me. I want to be clear about that, because I have spent months wrestling with the ethics of this, and I am still not certain I’ve landed on the right side of the question. I did not ask the machine to produce prose. I did not ask it to imitate my grandmother’s voice—a machine cannot imitate the particular way she pronounced pelu as if the word itself were a small gift. I asked it to do one thing: generate a book outline. To take my fragments and arrange them into a possible structure. To give me a skeleton I could dress in flesh, voice, and the specific Yoruba cadence that lives in my body even when the words escape me.
The Authors Guild, in their AI Best Practices for Authors, draws a line between using AI as a tool for brainstorming and structure versus using it to generate the final creative work. I cling to that distinction. The outline the machine produced was not my grandmother’s story. It was a scaffold—ten chapters, each with a thematic arc and suggested scenes. Some of the suggestions were wrong, by which I mean they were generic. The machine did not know that my grandmother’s first job in London was cleaning offices in Canary Wharf while the towers were still going up, that she would sometimes pause to watch the cranes and think about the carpenters in her father’s compound who had never seen a building taller than a palm tree. It could not know that. But the outline gave me something I’d been missing: a sense of proportion. It showed me that the story might begin not with her departure but with her return, that the grief of leaving and the grief of coming back might be braided together rather than separated by chronology. It gave me a container, and I filled the container with what I knew.
Still. Every time I opened the document, I felt a small betrayal sitting in my chest like an undigested meal. Whose story was this, if a machine had helped me tell it? My grandmother refused to own a microwave because she did not trust what she could not see cooking. She measured ingredients by the handful, by the feel of the dough against her palm. She would not have understood what I was doing. She might have laughed—a sharp, surprised laugh—and asked if the machine had also learned to pound yam. I would have had no answer.
Purdue OWL’s introduction to creative writing discusses the particular challenge of writing from memory, especially in creative nonfiction: the way memory is not a fixed archive but a living, shifting thing that changes each time we reach for it. I have returned to that idea often in the months since I began this project. My grandmother’s story was never static. She told it differently depending on who was listening. To her children, the migration was a lesson in sacrifice. To her church sisters, it was a testimony of God’s provision. To me, the youngest granddaughter who could not speak Yoruba fluently, it was softer, simplified, translated into a language I could hold. Which version is the true one? All of them. None of them. The machine does not understand this. The machine wants a single narrative arc, a coherent timeline, a protagonist with clear motivations. But my grandmother contained multitudes, and so does her story.
The outline sits on my desktop now, annotated in my own handwriting. I have crossed out chapter titles and written new ones in the margins. I have moved scenes around, trusting my body’s memory over the machine’s logic. The chapter the machine titled “Arrival in London” I have renamed “The Year She Stopped Dreaming in Yoruba,” because that is the detail my mother gave me over the phone last month, unprompted, when I asked her what she remembered most about those first years. She told me she knew she had changed when she started dreaming in English and could not find her way back. That sentence was not in the fragments I fed the machine. It arrived later, a gift from my mother’s own archive, and it rearranged everything. The machine’s outline was useful precisely because it was imperfect—it showed me where the story I had inherited diverged from the story a neutral observer might construct. The gaps were the point.
I am still learning how to talk about this process without feeling like I need to justify it, or apologize for it, or explain why I did not simply wait until I could afford to hire a developmental editor who understood the specific context of Yoruba migration narratives. The truth is that I wanted to move forward while the impulse was still hot in my hands. I wanted a companion that would not tire of my fragments, that would not need me to explain why a recipe for moin moin belonged in the same document as a description of a Brixton bus stop in 1972. The machine did not understand the connections, but it held space for them. It gave me permission to see the story as a story, rather than a sacred object I was too afraid to touch.
Is that preservation or betrayal? I do not know. I suspect the answer lives somewhere in the tension between the two, which is, I am learning, where most of my life lives. I am the granddaughter of a woman who crossed a continent and an ocean with nothing but a headwrap full of savings and a name that meant “the crown is sweet.” I am a writer who uses a laptop and an internet connection and, now, an AI-generated outline. These things feel contradictory, but perhaps contradiction is the inheritance of the diaspora child. Perhaps the most honest thing I can do is refuse to resolve it.
Last week, I read the first completed chapter aloud to my mother over a video call. She was silent for a long moment afterward, her face pixelated and frozen on the screen. Then she said, You have remembered her voice. She did not ask how I had structured the chapter. She did not ask about the machine. She only recognized the voice—her mother’s, and also mine. That recognition is not something any outline, human or artificial, can manufacture. It has to be built, sentence by sentence, from the inside of a life that knows the story because it carries the story in its body.
I am still feeding fragments to the machine, and I am still arguing with everything it gives back. The outline is not the memoir. The memoir is the thing I am making, slowly, with my grandmother’s recipes propped against my monitor and her name written on a scrap of paper I keep in my pocket. The machine is a tool, like the mortar and pestle she used to grind pepper—it does the preparatory work, but the flavor comes from the hand that holds it.
