The Twelve Tells

What a tell is, and what it is not

A tell is a habit that is visible on the page. It describes how writing reads. It says nothing about who or what typed it, and any tool that claims otherwise is selling you a coin flip in a lab coat.

Plenty of people wrote this way long before the models did. Consultants, agency copywriters, anyone who learned to write from other people's LinkedIn posts. A piece can trigger all twelve tells and still be human to the last word, and a piece can trigger none and still be empty.

So the cost is not being caught. The cost is being skimmed. A reader who feels the pattern does not report you. They stop reading, they stop trusting the byline, and they scroll. That happens whether or not anyone can prove anything.

One more thing before the twelve. We never score writing out of 100. A review names which tells fire, how many times, and where, plus one line of plain judgement. That is the whole verdict, and it is more useful than a number because you can act on it.

The rhythm tells

Five of the twelve are about pace and shape. They are the cheapest to fix and the fastest to spot, which is why they are worth clearing first.

Metronome. Sentence after sentence at near enough the same length. Fifteen words, fourteen, sixteen. Nothing in a model's training rewards variance, so everything drifts toward the mean, and prose at the mean ticks like a clock. The fix is physical: read the draft aloud. If your breathing stays regular, the writing is wrong. Put a three word sentence next to a forty word one and listen to what happens.

The Rule of Three. Three adjectives, three examples, three bullets, every time. Three is the trained rhythm of persuasive prose because it reads as complete. The cost is that a reader who notices the pattern stops believing the list was filled to the truth rather than to the count. Count your lists. If every one is a three, break one. Use two. Use five. Real thinking is lumpy.

The False Floor. A one or two word sentence carrying gravity it has not earned. 'Simple.' 'Wrong.' 'Here is the thing.' Those are specimens, not our words. It mimics a confident speaker landing a point, except a confident speaker lands it after the point rather than before. Allow yourself one per piece, after a genuinely hard passage that needs the air. If one turns up in your first three lines, cut it.

The Escalator. Transitions that announce themselves instead of doing work. 'Moreover.' 'Furthermore.' 'That said.' 'But here is the kicker.' Again, specimens. They are connective tissue with no argument inside them. Delete every one and reread. Roughly nine in ten will not be missed, and the survivor is the one doing real logical work.

The Emoji Scaffold. Structure imposed on content that never needed structuring. Ticks and rockets down the left margin look organised, so they feel like effort. What they actually do is convert an argument into a list, and a list cannot argue. If the points connect, write them as prose and let the connections show. If they do not connect, you have three posts rather than one.

The claim tells

Four tells are about what you assert and what backs it. These are the ones that quietly drain authority, because each of them produces the shape of a claim while removing the risk of making one.

The Negation Pivot. Two sentences that deny one thing to assert another. 'It is not about the tools. It is about the mindset.' Specimen. The shape produces the feeling of an insight without requiring one, since any two nouns slot in and sound profound. Say the second half on its own. 'This is a leadership problem' is a claim. The pivot version is a claim wearing a costume. If you truly need the contrast, earn it by naming who believes the wrong thing.

The Hedge Sandwich. A claim padded front and back until it cannot be wrong, and therefore cannot be interesting. 'While it is true that remote work has challenges, it is worth noting that many companies have found success, though of course results vary.' Specimen again, and a painful one. Cut the front hedge, cut the back hedge, keep the middle. If the naked claim frightens you slightly, that is the first sign you have written something worth reading.

Borrowed Authority. Appeals to evidence that never gets named. 'Studies show.' 'Research suggests.' 'Experts agree.' It buys credibility at zero cost, and any reader who goes looking finds nothing, which costs you more than the sentence ever bought. Name the source, the year and the number, or drop the claim. One checkable figure is worth ten rounds of 'studies show'.

Vocabulary Inflation. Words almost nobody says out loud, clustered in one piece. The cluster to watch, quoted as a specimen and not as our own writing: delve, leverage, seamless, robust, landscape, realm, tapestry, testament, elevate, foster, myriad, pivotal. None of them is banned by the language. The problem is that they arrive together, and together they read as borrowed. Say the sentence to another person. If you would not use the word to their face, it does not go in.

The story tells

Two tells live in narrative, and they are the hardest to fix with find and replace, because the problem is in what you chose to include rather than in any particular word.

The Symmetrical Anecdote. A story where every detail serves the point, nobody behaves badly, and nothing irrelevant happens. 'Last year a client came to me frustrated. We tried a new approach. Within three months, revenue doubled. The lesson was clear.' Specimen. It reads that way because it was reverse engineered from the moral, and real memories are not built in that direction. Put in one detail that does not serve the point. The weather. What they were wearing. Something you got wrong and only noticed months later. Friction is the proof of a real memory.

The Tidy Ending. A close that resolves everything neatly, then asks a question to manufacture comments. Symmetry is the default aesthetic, so every piece gets a bow. The cost is that a bow tells the reader there was never any real risk in the argument. End on the sharpest thing you said, not the safest. The best endings feel slightly unfinished, because thinking is. If you want comments, say something a reasonable person could disagree with.

The deep tell: No Cost

Tell twelve is worth more than the other eleven combined, and it is the one that survives every rewrite.

No Cost is what it sounds like. The writer risks nothing. No admission of having been wrong. No detail that makes them look worse. No named opponent, no number specific enough to be checked, no opinion that could cost them a client.

A model has nothing at stake by construction. It cannot be embarrassed, cannot lose a customer, and cannot be argued with at a conference next March. That absence stays legible even after every other tell has been scrubbed out, which is why a clean, competent, entirely safe post still reads like nobody was home.

The harshest read we give is reserved for a piece that trips this tell alone and none of the others. Technically clean, professionally competent, completely empty.

The fix is the only one on this list that costs you something. Name the thing you got wrong. Name the practice in your field you think is nonsense. Give the actual number instead of the rounded one. This is the tell no rewrite removes and no model imitates, so it is the one worth building a habit around.

Why a format will not save you

The obvious response to all of this is to go looking for a better template. Our corpus argues against it.

Across 29,599 posts from 53 creators, 2,867 land as Exceptional. Split those by archetype and the Exceptional rate stays flat, somewhere around 9 to 10 per cent whichever shape the creator writes in. The story-led creators, the framework teachers, the contrarians: broadly the same odds.

Read that carefully, because it cuts both ways. The shape you pick is not what decides whether a post lands, which means no format is going to rescue writing that risks nothing. It also means your preferred shape is not holding you back. What is left, once format stops explaining the outcome, is the twelve habits and what you were willing to put at stake.

The training loop

Naming a habit is what makes it visible, and visible is what makes it fixable. The loop has three parts and takes about ten minutes a week.

First, the two rounds either side of this lesson. Same game, fresh draw, and the second one is compared against your own first score rather than against anyone else's. There is no cohort number in this product because we do not have one, and a made up comparison would be the most machine-flavoured thing on the page.

Second, the tells pass in the editor. Paste a draft and it names which tells fire, how many times, and where they sit in your text. It does not score you out of 100 and it never guesses at who wrote the piece. It hands you a list of habits and their locations, and you decide which ones to keep.

Third, the weekly rerun. Put last week's published post through the same pass. The tells you keep hitting are your tells, not the general ones, and two or three will account for most of what fires. Add the words you keep reaching for to the forbidden list in your Voice Manifest so the next draft flags them while you are still writing.

One last thing about the game. Every human piece in it is writing a real person actually published, which is why the pool is being stocked slowly rather than filled quickly. A round built on invented human writing would test you against our impression of how people write. That is not a test worth passing.