Academic English revision

Your Coding Memo Reads Like AI. Your Supervisor Can Tell in 30 Seconds

Coding memos are your brain on paper. AI strips out the doubt, the dead ends, and the specific names — and that's exactly what makes them look fake.

Your supervisor opens your coding memo. Reads for thirty seconds. Closes it. Not because there's a typo. Because it reads like a generated summary, not like a researcher actually thinking through data. You wrote this memo as evidence of your analytical process — not as a miniature paper. But when AI "generates" it for you, it defaults to paper mode. Every sentence is correct. Every sentence is balanced. Every sentence is perfectly hedged. And your supervisor sees one thing: this student isn't thinking, they're performing thinking. The worst part? You can't even tell it's off. It's not grammatically wrong. It's not logically flawed. It's just... not yours. And somehow that's the easiest thing to detect.

A coding memo is your brain on paper. Your supervisor doesn't want to see your conclusion — that lives in your findings section. They want to see how you got there. Did you hesitate? Did you recode something three times? Was there a point where you got stuck for two days and came back with fresh eyes? Those imperfections are the most valuable part of the memo. AI is spectacular at erasing them. It turns your hesitation into "it is worth noting." It turns your stuckness into "this requires further exploration." It turns your genuine confusion into a paragraph of polished academic nothing. You've lost the evidence that you were actually doing the work.

Three specific patterns give away an AI-written coding memo. First: no "I." Real memos say "I initially coded this as X but I'm not sure that's right." AI writes "A recurring pattern emerged." Second: no dead ends. When you're actually coding, you walk down a path and realize it doesn't work. You write that down. AI never walks down a wrong path. It only "discovers" new patterns. Third: no specific names. Your memo says "P3 kept using that word over and over." AI says "multiple participants." The difference between a researcher's memo and an AI's memo is the difference between a person and a function.

Here's what the difference actually looks like. ❌ "The data revealed that participants experienced a significant degree of emotional distress related to their academic performance, which was consistently thematized across multiple interviews." → ✅ "P3 kept coming back to her grades being 'like my worth being graded.' I initially coded this as 'anxiety' but I don't think that captures it. Maybe identity? Revisit with P5." ❌ "A recurring pattern emerged wherein participants described a tension between maintaining academic standards and preserving personal well-being." → ✅ "Almost everyone mentioned the push-pull between 'doing well' and 'not losing myself.' P1 said it most bluntly. Is this one theme or two? Supervisor might push back if I merge them." ❌ "The findings indicate that participants' perceptions were shaped by both institutional factors and interpersonal dynamics, necessitating further exploration of the interplay." → ✅ "I'm starting to think 'institutional' and 'interpersonal' aren't separate at all. When P5 talked about department culture she was literally describing her supervisor. Same thing, different clothes. Flag for memo #7."

Here's a test: if you deleted every uncertain, hesitant, "I'm not sure" moment from your memo, what's left? If the answer is "a literature review paragraph," you've got a problem. Good coding memos read like you're explaining your thinking to a smart friend over coffee. You say "wait, I actually don't have this figured out yet." You say "I thought it was this at first, then I realized it was something else." Those fuzzy edges are your most credible evidence as a researcher. AI sands them all down. What your supervisor sees is a smooth, empty surface.

Practically: I let AI pull out the key excerpts from transcripts and flag possible theme directions. That part is genuinely fast. But the memo itself? I never let it "generate" that. I write a first draft — messy, fragmented, full of typos — then I run it through easydue to smooth the register and soften the harshest sentences while keeping my uncertainty intact. The time I save isn't for polishing. It's for actually figuring out whether that theme is real. Your coding memo is not a paper. It's a CT scan of your brain. A CT scan needs noise. Without noise, it's not a real scan.