Signs of AI Writing: 20 AI Slop Tells and How to Fix Them

The clearest signs of AI writing come in clusters: throat-clearing intros, adjectives stacked in threes, sentences that all run the same length, and claims with no number attached. One of these on its own proves nothing. Four or five in the same two paragraphs almost always mean nobody edited the draft.

This guide covers the 20 tells that are easiest to spot during an edit, with an example of each and how to rewrite it. It’s written for the person who approves content before it goes out under the company name: LinkedIn posts, sales emails, proposals and blog articles.

Why it matters if readers can tell

I see it every day on LinkedIn. Posts and comments from different creators that read the same, with the same hook, the same verbs and the same motivational sign-off. The feed is full of content nobody fully wrote and almost nobody finishes reading.

My take is simple: readers notice. They can’t always say why, but two lines in they know the text is generic and they scroll. For a company that sells judgment, like a consulting firm or a B2B vendor, that moment of doubt is expensive. If the writing has no owner, readers assume the offer doesn’t either.

Google, for its part, doesn’t penalize content for how it was made. Its official guidance on AI-generated content says it rewards helpful, original content however it’s produced, and that using automation to generate content mainly to manipulate rankings violates its spam policies. In practice, AI slop loses for the same reason any weak content loses: it gives the searcher nothing new.

The 20 signs of AI writing

I’ve grouped them by where they show up. Each one comes with a typical example and a fix.

Openings and endings

1. Throat-clearing intros. “In today’s fast-paced business world, AI is changing how we work.” The first sentence tells readers what they already know. Delete it and see if the piece starts better at sentence two. It usually does.

2. The recap ending. “In conclusion, AI offers tremendous opportunities for businesses.” It repeats what the reader just read. End on your strongest data point, a specific recommendation or the next step.

3. Motivational sign-offs. “The future is now. Are you ready to take the leap?” Swap it for something the reader can do tomorrow: “Pull the last five emails your team sent and count how many open with a generic greeting.”

4. Rapid-fire rhetorical questions. “Is your team wasting time? Are your processes slow? Do you feel like you’re falling behind?” Keep the one question your reader is actually asking, or none.

5. Formula headlines. “The Ultimate Guide to…”, “Everything You Need to Know About…”. Say in the headline what the reader gets and who it’s for.

Word choice

6. Inflated verbs. Leverage, unlock, elevate, empower, delve, revolutionize. Use the verb that describes what happens: “month-end inventory close went from three days to one” beats “optimizes inventory management.”

7. Brochure nouns. Ecosystem, synergy, paradigm, landscape, end-to-end solution, value-add. Name the thing. If it’s software, say what kind. If it’s a process, say which one.

8. Adjectives in threes. “A fast, efficient and scalable solution.” Language models group things in threes because it sounds complete. Keep the one adjective you can prove, then prove it.

9. Empty intensifiers. Incredible, fascinating, crucial, groundbreaking. Delete them and reread the sentence. If it got weaker, what it needs is a number.

10. Filler transitions. “It’s worth noting,” “it’s important to mention,” “moreover,” “that being said.” Cut them. If the point matters, readers will see that without a warning label.

Rhythm and structure

11. Sentences of identical length. Every sentence 15 to 20 words, every paragraph three lines. It reads like a metronome. Mix it up: a short sentence after a long one changes the rhythm immediately.

12. The fake contrast. “It’s not just a tool, it’s a transformation.” This may be the single most recognizable AI tic. It sets up an opposition nobody proposed so the second half sounds like a revelation. State the second half directly and back it with a fact.

13. Em dash overload. Asides set off with long dashes in nearly every paragraph. People use them too, just not this often. Turn the aside into its own sentence or use a comma.

14. Bullets where an argument belongs. One-line bullets, each with a bolded label and a colon. If the ideas depend on each other, write a paragraph. Lists are for steps or items that stand alone.

15. Bold everywhere. When everything is emphasized, nothing is. One bolded line per section is plenty.

Substance

16. Claims with no subject and no number. “AI can significantly improve efficiency.” How much, where, according to whom? Add the who, the how much and the timeframe. If you don’t have them, cut the sentence.

17. Hypothetical examples. “For example, a retail company could use AI to…”. Use a case you’ve actually seen, even without names, or a sourced figure.

18. Forced balance. Every point arrives with “however, it’s also important to consider…” and the piece never commits. Decide what you think and say it. Explain the caveats that matter once.

19. Zero firsthand experience. Nothing suggests the author did something, got something wrong or was in the room. Add a real scene: a meeting, a mistake, a conversation with a client. It’s the one thing a model can’t make up without lying.

20. Keyword stuffing. The exact same phrase every other paragraph, written for a search engine. Use it where it fits and vary the rest. Google has understood synonyms for years.

Do AI detectors work?

Not well enough to make decisions with. OpenAI launched its own AI text classifier in January 2023 and pulled it that July because of its low accuracy. In OpenAI’s own testing it correctly flagged only 26% of AI-written text and mislabeled 9% of human writing as AI.

A Stanford-led study also found that several widely used detectors consistently misclassify writing by non-native English speakers as AI-generated. If your team includes people writing in a second language, a detector score can flag their work for the wrong reason.

A detector gives you a percentage. An editor tells you which sentence to change and why. For anything that carries your company’s name, the editor is what you need.

How to edit an AI draft in 15 minutes

Here’s the order that works for a standard post or article:

Step	What you do	Minutes
First and last lines	Cut the generic opener and closer, then check whether the piece reads better without them	2
Word sweep	Ctrl+F the verbs, nouns and filler phrases from this list and replace them with what actually happens	4
Evidence	Every broad claim gets a number, a timeframe or a source; if there isn’t one, it goes	4
Experience	Add at least one case, mistake or opinion of your own	3
Read it out loud	Shorten or stretch the sentences that sound alike	2

Once a company starts writing with AI, the bottleneck moves from drafting to editing. Deciding who reviews, against what standard and how fast is part of rolling out AI across a team without the brand ending up sounding like everyone else’s.

FAQ

Does Google penalize AI-written content?

Not for the method. Google looks at whether content is helpful, original and trustworthy. What it does penalize is mass-producing pages to game rankings, whether AI or people wrote them.

How can you tell if ChatGPT wrote something?

There’s no reliable test. The best approach is to look at the signs together: if a piece hits several of the 20 on this list and contains no data or firsthand experience, it most likely came out of a model unedited.

What is AI slop?

It’s the term for low-quality content mass-produced with AI and published without real human review. Wikipedia maintains a field guide to signs of AI writing for its own editors.

Is it wrong to use AI to write?

No. AI speeds up the first draft and helps organize ideas. The problem starts when the draft gets published as-is, without the judgment or experience of the person whose name is on it.

Is your team already writing with AI and you want it to keep sounding like you? Let’s talk.