Killing the Tricolon: How AI’s Love of Three-Item Lists Gives It Away

Read almost any block of ChatGPT output and you’ll start seeing it everywhere. The model promises something “clear, concise, and compelling.” It describes a tool as “fast, reliable, and affordable.” Three items. Always three. Once you notice the pattern, you can’t unsee it, and neither can a careful reader.

This is the tricolon, the rhetorical habit of grouping things in threes, and AI is hopelessly addicted to it. The good news: it’s one of the easiest tells to spot and fix once you know it’s there.

Key takeaways

• AI models lean hard on the tricolon because three-item lists are statistically common in training data and feel “complete” to a model predicting the next likely word.
• A single tricolon is good writing. The tell is the relentless repetition: the same “X, Y, and Z” rhythm in paragraph after paragraph.
• You can catch it by literally searching your draft for “A, B, and C” constructions and counting them.
• The fix isn’t banning threes. It’s varying list length, breaking parallel grammar, and swapping vague trios for one concrete detail.
• Detectors don’t flag tricolons by name, but the flat rhythm they create is the low-variance pattern that nudges detection scores up.

Why AI writes in lists of three

The rule of three is real, and it works. Humans have used it for thousands of years because three is the smallest number that establishes a pattern. “Veni, vidi, vici.” “Life, liberty, and the pursuit of happiness.” “Blood, sweat, and tears.” A pair feels like a comparison. Four or more starts to feel like a catalog. Three lands.

Because the device is so common and so effective, it shows up constantly in the text these models were trained on. A large language model works by predicting the next likely token given everything before it. When it starts a list, the statistics push hard toward wrapping up at three items, because that’s where the training data most often closes the loop. Stopping at two can feel abrupt to the model’s probabilities. Going to five introduces less common structure. Three is the safe, smooth, high-probability choice.

So the model isn’t wrong to use the tricolon. The trouble is that it reaches for the safest option every single time. A human writer might use a three-item list, then a blunt one-sentence point, then a long winding clause, then a pair. The model defaults to the comfortable middle over and over, and that sameness is the giveaway.

If you want the deeper mechanics of why uniform, predictable text reads as machine-written, we covered the statistics behind AI detectors separately. The tricolon habit is one specific flavor of that broader predictability problem.

What the overuse actually looks like

Here’s a paragraph in the classic AI register. Read it and watch the rhythm:

> Our platform is fast, reliable, and secure. We help teams plan, build, and ship better products. Whether you’re a startup, an agency, or an enterprise, our flexible, scalable, and intuitive tools adapt to your needs.

Three sentences, four tricolons. Nothing in it is false, but the cadence is hypnotic in a bad way. Every sentence has the same beat: comma, comma, and. The brain stops processing meaning and starts hearing a metronome.

Now compare a version a person might actually write:

> Our platform is fast. Fast enough that builds people used to start before lunch now finish before they’ve topped up their coffee. It’s built for teams who ship constantly. Startup or enterprise, the tools bend to how you already work instead of forcing a new process.

Same claims, roughly. But the rhythm moves. One sentence is short. One leans on a small concrete image. One breaks the parallel structure entirely. There are no tricolons, and it reads like a person cared about it.

The “vague trio” problem

The worst tricolons aren’t just repetitive, they’re empty. “Efficient, scalable, and robust.” “Innovative, dynamic, and forward-thinking.” These lists exist because the model needed to fill the slot, not because each word earns its place. They’re the writing equivalent of clearing your throat.

A useful test: cover two of the three adjectives and ask whether the remaining one actually means anything specific. If “scalable” alone tells the reader nothing concrete, adding “efficient” and “robust” won’t save it. Stacking vague words doesn’t add up to precision. One real detail, a number, a name, a tiny scenario, beats three abstractions every time.

How to find tricolons in your own draft

You don’t need a tool for this. You need a search box and your ears.

Search for the structure. Use find-and-replace or a simple scan for the comma-comma-and pattern: “X, Y, and Z.” Highlight every hit. If a 400-word passage has five of them, that’s your problem flagged in yellow.

Read it aloud. This is the fastest method by far. The tricolon is a sound, and your ear catches a repeating cadence long before your eye notices the grammar. When two or three sentences in a row hit the same beat, you’ll hear the drone.

Count, then judge. One or two clean three-item lists in a few hundred words is normal, human writing. More than that, and you’re leaning on the device as a crutch. The threshold isn’t a hard rule. It’s about whether the pattern has become predictable.

How to fix it without writing worse

The goal is not to eliminate the number three. It’s to restore variety. A few moves cover most cases.

Cut one item. Often the third item is the weakest, added for rhythm rather than content. “Fast, reliable, and affordable” frequently works better as “fast and genuinely affordable,” because now each word has weight and the cadence shifts.

Add a fourth or fifth. Sometimes the honest answer is that there are four real things, and the model just trimmed to three out of habit. A four-item list breaks the expected beat precisely because it refuses to resolve where the reader’s ear expects.

Break the parallel grammar. Tricolons feel mechanical partly because all three items share the same shape: three adjectives, three -ing verbs, three noun phrases. Mix the shapes. “It’s fast, it won’t fall over under load, and honestly the price is the part people don’t believe.” Same three ideas, but the symmetry is gone and so is the robotic feel.

Replace the trio with one specific. This is the strongest fix. Delete “efficient, scalable, and robust” and write the one true detail underneath it, like a real load number you actually measured. Specificity is the most human thing you can put on a page, and it does the persuading that three adjectives only pretended to do.

A quick before-and-after

Before: *”Students choose us because we’re trustworthy, effective, and easy to use.”*

After: *”Students choose us for one boring reason: it works, and it takes about ninety seconds to learn.”*

The first sentence is three hollow adjectives in a row. The second names a concrete reason and a concrete time. It’s shorter, it’s specific, and it doesn’t sound like every other landing page on the internet. If you’re curious what that kind of writing tool actually costs, you can see what plans cost before you commit to anything.

Why this matters beyond just sounding human

The tricolon habit isn’t only an aesthetic issue. AI detectors work by measuring how predictable and how varied your text is. Constant three-item lists flatten your sentence rhythm and tighten your structure into a repeating loop, which lowers the variation detectors associate with real human writing. No detector has a “tricolon rule,” but killing the pattern raises your rhythmic variety, and that’s one of the signals that actually moves the needle. Keep in mind these tools report probabilities, not proof, so a score is a hint about your rhythm, not a final judgment.

It also just reads better. Real readers, the ones who decide whether they trust you, feel the monotony even if they can’t name it. Breaking the pattern helps you with both at once.

If you want more of this kind of sentence-level craft, there are more writing-craft guides on the blog. And once you’ve made your edits and want to check whether the rhythm still reads flat, you can run your draft through PaperBleach and fix the lines that need it.

Frequently asked questions

Why does AI write in lists of three so often? Two things stack up. The rule of three is genuinely common in English, so a model sees it constantly during training and learns to reproduce it. And because a language model predicts the next likely token, three items is a statistically safe place to wrap up a list. The result is overuse, far past the rate a careful human writer would choose.

Is using a list of three actually bad writing? No. The tricolon is one of the most effective devices in the language. The problem is monotony, not the device. Good writing mixes threes with pairs, single points, and longer lists. The fix is variety, not abstinence.

Do AI detectors specifically look for tricolons? No detector counts tricolons directly. What they measure is predictability and rhythm variation. Endless three-item lists flatten your rhythm, which lowers the variance detectors associate with human writing, so the habit feeds the broader pattern rather than tripping a specific rule. And detectors output probabilities, not proof, so treat any number as a signal.

How do I find tricolons in my own draft? Search for the “A, B, and C” structure and read the draft aloud. Your ear catches the repeating cadence faster than your eye. More than one or two per few hundred words usually means you’re leaning too hard on it.

What should I replace a tricolon with? If the three items are vague, replace the whole thing with one concrete detail. If they’re real, vary the length and break the parallel grammar so each item isn’t the same shape. Keep the meaning, kill the predictable beat.

The takeaway

The tricolon isn’t the enemy. Predictability is. AI reaches for three-item lists because they’re the safe, high-probability move, and it makes that move so often that the rhythm becomes a fingerprint. Mix in pairs, single sentences, and the occasional concrete number, and you’ll sound like a person again.