How your readers can tell when you’re using AI (and how to add the human touch)
In poker, a “tell” is a small, unconscious signal that gives a player away. AI has tells, too.
There’s no escaping AI these days. It’s everywhere we look: in advertising, in business, in our personal lives. And chances are, your company or clients are using it when they write.
As AI gets more ubiquitous, so do its habits. Perhaps you’ve seen complaints about the em dash — or many lists of three — as signs of AI-written work. Let’s look at some of the other “tells” that make readers pause and think, “Wait a minute. A human didn’t write this.”
The more tells, the more obvious
No single quirk proves AI authorship. After all, humans use em dashes and tidy lists, too. We’ve all used these devices — that’s where AI learned them! The artificial feeling is the accumulation of tells: repeated patterns, predictable structure, and the sing-song feel of many AI-constructed paragraphs in a row. Taken together, your reader will increasingly get the sense that this text was generated, not written.
What does it matter?
You may be asking, “But if AI makes writing so much easier and allows us to get content to our readers faster, who cares if we use it or not?” The answer is credibility and trust. If your audience gets the feeling a machine wrote something, they’re going to think you didn’t put time or effort into communicating your message, and they’ll assume you’re the kind of writer or company that takes shortcuts. That’s not the kind of impression you want to make.
Trust is built on the sense that there’s a thoughtful human on the other side of the page. Once trust slips, it’s difficult to win back. So yes, it matters.
Two kinds of tells
There are two ways you can tell whether AI was used in a piece of writing. One is at the sentence level: gimmicky sentence types that AI loves to roll out and humans don’t use nearly as often or as repeatedly. The other is a larger-level tell: the sense that instead of a human writing about a topic, you have a machine generating language on the topic. Here’s a look at both those types.
The sentence-level tells
1) Whether X or Y, it’s Z.
- Examples:
- “Whether you enjoy skiing or snowboarding, winter is a great season to get outdoors.”
- “Whether you’re new to AI or a veteran prompter, here are some ways to finesse your prompts.”
- How to fix it:
- It’s okay to use this once or twice, but avoid repeating it throughout your document or at the beginning of multiple sections.
2) The question? The answer. (Or “The X: The Y.”)
- Examples:
- “The problem? Scale.”
- “Is it worth it? Absolutely.”
- “The issue: How to secure your data.”
- How to fix it:
- Instead of gimmicky punctuation, just use “is”: “The problem is scale.” Or skip the general sentence altogether and go straight to your specifics.
3) It’s more than X. It’s Y. (Or “It’s not just X — it’s Y.)
- Examples:
- “Star Trek is more than a TV show. It’s a lifestyle.”
- “What happens next is not just success, but growth on an exponential scale.”
- How to fix it:
- Skip the first half of the comparison altogether: “Star Trek is a lifestyle.” Or use the boring, but less gimmicky, “and”: “You’ll achieve success and growth on an exponential scale.”
4) Examples, examples, and examples. (Sometimes: Example. Example. And slightly longer example.)
- Examples:
- “You’ll see improved efficiency, manpower, and profits.”
- “No mess. No fuss. No wasted time.”
- How to fix it:
- Obviously, this is something that many human writers do on top of AI. So, try to check the number of these lists in your document and cut them down or vary them. If you’re not sure where to snip, think about what your examples are meant to show. Is each one necessary to give an idea of the scope of your concept?
5) The general: the specific.
- Examples:
- “This product delivers where it counts: visibility and ease of use.”
- “You’ll see a marked improvement — better stamina, improved mood, and, of course, a smaller waistline.”
- How to fix it:
- Just skip that general statement. The product delivers visibility and ease of use. AI is addicted to starting general and then getting specific (more on this later), so it says more than it has to say. You, as a human, can get straight to the point.
6) A general purpose at the end of a specific sentence
- Examples:
- “The product provides task tracking and automation to help your organization stay agile.”
- “We use a centralized management portal to communicate with health care providers and make sure their billing is compliant, to better deliver care to our members.”
- How to fix it:
- I’ve seen long sentences, full of details, that end with some sort of general platitude. The details speak for themselves. You don’t need to add the general purpose statement. Cut it!
7) Weird and repeated word choices
- Examples:
- “Yet” where “but” will do just fine
- “Since” where “because” or “so” would work better
- “As a result” or “This means that…” where simpler wording would be clearer
- How to fix it:
- Use the appropriate word. And avoid signposting language like “This means” or “Therefore” or “Consequently.” That advice goes beyond just removing AI tells and strengthens all your writing.
8) The em dash, yes, but specifically at the end of your paragraphs
- Examples:
- “Together we can achieve great things — by collaborating and applying our knowledge.”
- “All these factors contribute to the extreme weather we’ve been seeing — wildfires, floods, and intense storms.”
- How to fix it:
- Em dashes in general don’t scream AI, but if you find yourself using them as a crutch to give your paragraphs a snappy conclusion, try some other ways to finish your thoughts. Often, you don’t need a snappy conclusion at all. Deliver the information and let it stand.
The larger-picture tells
Paragraphs
Odds are you’ve seen this paragraph structure before:
- An opening using general information or a vague concept
- A second sentence that applies that information/concept to the specific topic
- Middle sentences using lists, usually lists of three (see #4 above)
- A final sentence that uses an em dash (#8 above) or a colon to wrap things up in a more general way
There’s nothing wrong with a paragraph that does these things. AI, which is, at its heart, a probability machine, produces paragraphs that look like this exactly because humans use this structure so often. But when you start to see multiple paragraphs that look this way, that’s when things start to feel generated, rather than written.
Generalities
Another tell, at the section or chapter level: You’ll probably see AI generate an opening paragraph that’s so general it might as well be about anything. In the next paragraph, it might drill down to your topic, but the first paragraph will say something like “The pace of innovation is accelerating” or “Climate concerns are higher than ever before.” Something vaguely in the universe of your point, but really not necessary to state because everyone already knows it.
Vagueness
Finally, and this is the toughest tell to spot: AI doesn’t make your points. It generates language about them. If the reader finds that the words seem to be circling the point you want to make but don’t exactly land on it, they’ll get the feeling that a human didn’t write it.
To fix this, always be specific. Say what you mean to say. You’ll avoid accusations of AI writing, and you’ll get your point across clearly. This may mean writing fewer words than you’re used to, but that’s a good thing. Bloated language and corporate speak are severely overrated.
So if AI is no good, what should we do?
Hire human writers. A good human writer will create engaging copy without falling into the repeated sentence and paragraph patterns that you see in AI-generated copy.
Drop Dragonfly a line! Our writers produce vivid, human-penned writing, and our editors are experts at spotting AI text and rewriting to make it more human.
