$ man content-wiki/ai-slop-guide

AI slop is the collection of writing patterns that make AI-generated content instantly recognizable as AI-generated. Not because the ideas are bad — because the delivery follows predictable patterns that no human naturally writes in. Em-dashes everywhere. Authority signaling phrases. Three parallel dramatic sentences. Bookend summaries that restate the introduction word-for-word. These patterns exist because language models learned from a training corpus full of a specific kind of polished, performative writing. The model defaults to that style unless you actively constrain it.

PATTERN

1. Em-dashes: delete all of them. Use periods, commas, or restructure. Natural alternatives are ellipses and arrows. 2. Authority signaling phrases: the uncomfortable truth, let me be clear, here is what nobody tells you, the hard truth is, here is the reality, what most people miss. These are performance, not substance. Exception: here is how and here is the play are directional in builder voice, not performative. 3. Narrator setup lines: here is the thing about, here is where it gets interesting. Delete the setup. Start with the actual point. 4. Dramatic rhetorical framing: but here is the part where, and that is when it clicked, want to know the crazy part. State what happened. Let the reader feel it. 5. Three parallel dramatic sentences: you cannot see it, you cannot copy-paste it away, you have to know it exists. One direct statement lands harder. Cut to one. 6. The bookend summary: opening with a thesis, closing with the exact same thesis rephrased 800 words later. AI wraps content in neat bows. Real thinking goes somewhere new by the end. 7. Self-branded concepts: this is what I call. Just explain it.

PATTERN

8. Artificial drama sentences: the shift sounds simple, it is not. Show why it is hard with a specific example instead of telling the reader it is hard. 9. Colon-listed everything: the result: better data, the impact: faster sales. Reads like a PowerPoint slide. Write natural sentences. 10. The humble brag disclaimer: I do not have all the answers, but. Share your take or do not. The disclaimer makes it worse.

PATTERN

These need judgment, not automatic deletion: 11. Engagement bait endings: so here is my question for you. Generally avoid. But on meme posts, asking what is your version and drop it in the comments fits the lighter tone. 12. Bullets for arguments: arguments belong in prose. But workflow steps, tool lists, and technical implementations use bullets and emoji markers naturally. The rule: bullets for execution, prose for reasoning. 13. False dichotomies: it is not X, it is Y. Generally avoid. But contrasting old way vs new way — manual SDR grind vs automated Clay workflow — is showing evolution, not a false dichotomy. 14. Bold headers as transitions: headers for navigation only. LinkedIn posts are whitespace-driven, not header-driven.

PRO TIP

These look like AI tells but are actually authentic voice patterns. Do not flag them: Ellipses for trailing thoughts: true story, it feels prehistoric. Arrows for workflow steps and progression. Emoji section markers (checkmarks, wrenches, links, brains, puzzle pieces) for structuring workflow walkthroughs. Here is how and here is the play as directional openers into workflow breakdowns. Pop culture references mixed into technical content. No gatekeeping as a value statement with resource delivery in comments. The distinction: AI patterns are formulaic and repetitive across all content. Your natural patterns are contextual — they appear in specific content types where they serve a function. Ellipses show up in reflective posts, not in technical breakdowns. Emoji markers show up in plays, not in essays. Context determines whether a pattern is authentic or slop.

FORMULA

Before publishing any AI-assisted content, run this scan: (1) Search for em-dashes — delete all. (2) Search for here is the thing, here is where, let me be clear, the uncomfortable truth — delete all. (3) Check the opening and closing — if they say the same thing, rewrite the closing. (4) Count parallel sentence structures — if you find three in a row with the same rhythm, cut to one. (5) Check for colon-listed statements — rewrite as natural sentences. (6) Read it out loud — if it sounds like a keynote speech, it is slop. If it sounds like you explaining something to a friend, it is voice. This takes 3 minutes per post. Those 3 minutes are the difference between content that sounds like everyone else's AI output and content that sounds like you.