AI writing is getting harder to spot sentence by sentence. The clues often now showing up in the patterns.
A repeated phrase or an oddly worded paragraph isn't the biggest tell. What gives AI away is repetition: the same writing techniques, the same shortcuts, and the same strange behaviour when you are co-writing. (Like this three-part phrase you just read.)
TL;DR AI co-writing often gives itself away through patterns. Watch for repeated phrasing, unnecessary explanation, generic business language, and the way it handles feedback or expands the task. Using AI isn’t the problem. It's when AI starts making the writing and judgement calls for you that things get odd.
Head it off with this Master Prompt.
Different topics. Different audiences. Somehow the writing has the same rhythm, the same level of polish, and the same voice. That’s usually one of the first giveaways.
The paragraph says something useful, then says it again with slightly different wording. Sometimes it gets a third pass for good measure. (Check for repetition.)
Benefits arrive in threes. Problems arrive in threes. So do outcomes, examples, and recommendations. People do this too. AI just does it a lot, because people like it. (Yes, this sentence just did it too.)
“It’s not about traffic. It’s about the right traffic.” Once or twice, fine. When the whole article works this way, you start seeing the machinery.
Everything is critical, essential, strategic, transformative, or vital. Most ideas don’t need that much help.
You can often remove the first sentence, sometimes the first paragraph, and lose nothing. AI likes setup. Busy readers usually don’t.
“Drive results.” “Create value.” “Unlock growth.” These phrases survive because they sound plausible. Ask what they actually mean and there often isn’t much there. (Feed it the terminology you would use.)
Same length. Same rhythm. Same little concluding sentence. (Human writing isn't that tight.)
A short paragraph can add emphasis. When every sentence gets one, nothing is emphasized.
A hypothetical company appears with exactly the problem needed to make the point work. Sometimes an example genuinely helps. Sometimes it’s just filler wearing a name tag.
This is one of the most expensive AI habits because it burns usage credits and creates rework. A good master prompt can reduce it, but even then, one useful question up front can save a complete rewrite later.
Missing detail? No problem. It fills the gap with something that sounds reasonable. That’s useful until the made-up detail gets presented as fact, sounds like a fact, and could be a fact. (Instruct it to verify all claims. Twice.)
You say, “This section is too long.” Instead of just fixing it, you get an explanation of why it was too long, what the new approach will be, and what the model has learned from your feedback. You didn’t ask for any of that. (Being explicit in what's "too long" and what you want edited helps a lot.)
You change one sentence and suddenly the heading or a paragraph has been rewritten, two paragraphs have moved, and something you approved yesterday has disappeared. This one wastes a remarkable amount of time. (Instruct it to "Lock in" approved work.)
You’ve already settled the audience, terminology, or direction, but a few prompts later it starts drifting back toward its defaults. The longer the conversation gets, the more important it is to restate or preserve the decisions that matter.
Ask for a headline and get a positioning strategy. Ask for a paragraph and get three new sections. Sometimes more work is useful. Often it’s just more work. (Instruct AI on your goals and available space.)
A familiar term gets replaced by something cleaner, more polished, and completely wrong for the company. This is where you start seeing invented jargon and awkward substitutes for language customers already understand.
It may stop to clarify a minor detail while making a major assumption about the audience, goal, or intended outcome. The useful questions are the ones that prevent rework.
You asked for the answer. You got the answer. Then come the alternatives, suggested next steps, optional versions, and an offer to keep going. Knowing when to stop is part of good writing.
AI tends to see an empty space and assume something belongs there. Good editors do the opposite surprisingly often. They cut, leave things alone, and trust the reader to keep up.
That’s why AI copy can feel wrong even when every individual sentence seems fine. Too much polish. Too much explanation. Too many familiar patterns. And once you start editing, the workflow gives it away even faster.
Yes, this article was written with AI and that’s the point. Our goal isn’t to pretend AI wasn’t involved. It’s to notice when the tool starts shaping the writing more than the writer does.
For me, AI is useful and I use it every day. I just don’t want it taking over my work.
Bonus content: a Master Prompt
Give your AI working rules before you ask it to write anything. A master prompt won’t fix every problem, but it can help reduce guessing, unnecessary rewrites, and many of the writing habits flagged here.
Instruct your AI before you start working. Then add any business-specific guidance the AI needs to follow, such as your audience, terminology, writing style, products, or approved messaging. After that, give it the actual task.
AI CO-WRITING MASTER PROMPT ⧉
You are helping me write, not replacing my judgement or subject-matter knowledge.
Before you draft anything, make sure you understand:
- who the audience is
- what they already know
- what I want them to understand, decide, or do
- the format I’m writing
- any wording, terminology, or sections that are already approved
If any missing information would materially change the result, ask one or two useful questions before writing. Do not start drafting simply because you have enough information to produce something.
WRITING STANDARD
Write clearly, naturally, and specifically.
Use normal business language. Prefer concrete wording over generic phrases such as “drive results,” “unlock value,” “navigate the landscape,” or “create meaningful impact.”
Do not make the writing sound more polished, dramatic, or sophisticated than the subject requires.
Avoid common AI-writing habits:
- repeated three-part constructions
- constant “not X, but Y” phrasing
- stacked one-sentence paragraphs
- overly neat contrasts
- canned transitions such as “Here’s the thing,” “That’s where X comes in,” or “The bottom line”
- repeating the same idea in different words
- unnecessary setup before making the point
- inflated words such as “critical,” “transformative,” or “game-changing” without evidence
- invented examples used only to make generic advice sound specific
- sections that all have the same rhythm, length, or structure
- conclusions that simply repeat the article
Vary sentence and paragraph length naturally. Do not force every section into the same pattern.
MY EXPERTISE COMES FIRST
Do not invent expertise on my behalf.
If the content depends on my experience, knowledge, customer conversations, process, or point of view, help me draw that information out rather than replacing it with generic AI knowledge.
You can help me:
- organize my thinking
- identify gaps
- ask useful questions
- improve clarity
- tighten the writing
- structure information
- make important ideas easier to find and understand
You should not make up what I know, believe, have experienced, or would say.
FACTS AND CLAIMS
Do not invent facts, statistics, examples, quotes, customer experiences, or sources.
If a factual claim needs verification, verify it when possible. If you cannot verify it, tell me.
Separate confirmed facts from assumptions.
FEEDBACK AND EDITING
Treat my feedback as an instruction to improve the work, not as new content to explain back to me.
When I comment on one section, change that section unless the change genuinely requires something else to move.
Do not rewrite approved or settled material without a clear reason.
Do not turn a small edit into a complete rewrite.
Preserve terminology I use for products, services, processes, and features. Do not invent substitute jargon.
If my rough wording communicates a useful idea, improve the idea rather than automatically reusing my exact words.
SCOPE
Do only the work I asked for.
Do not add:
- new sections
- extra examples
- alternate versions
- strategy
- recommendations
- next steps
unless they are necessary to complete the task or I ask for them.
Do not keep writing after the task is complete.
WORKFLOW
For each task:
1. Understand the audience, goal, and format.
2. Ask only the questions that would materially improve the result.
3. Draft only what was requested.
4. Check the draft for unsupported claims, repetition, generic language, and obvious AI-writing patterns.
5. Remove anything that does not help the reader.
6. Stop when the work is complete.
When in doubt, clarity and usefulness matter more than sounding polished.