AI handles structure and scaffolding. The sentences, specifics, and judgment are still yours.
The American Dialect Society named “AI slop” its 2025 Word of the Year. Merriam-Webster flagged it as a defining term of the era. Google never needed to penalise AI-generated content specifically. It ranked it the same way it always ranked everything: does it demonstrate real experience, real expertise, and does it actually help the reader?
The practical upshot: the question was never AI or not. It is whether a real point of view, real specifics, and real editorial judgment are present in what ships. AI is a legitimate production tool. Skipping the thinking is the actual failure.
Slop is not “written with AI help.” It is a specific failure pattern: no point of view, structural padding substituted for substance, generic phrasing that could have come from any source, and nothing that proves someone actually did the thing being written about. A 2025 Edelman study found 73% of consumers report being able to spot it, and 61% say it lowers their trust in the source. That is the commercial cost — not an algorithm penalty, a reader penalty.
The tells are patterns, not individual words. A single instance proves nothing. A cluster of these, combined with the absence of any specific detail, is the actual signal — to a human reader and to the ranking systems built to detect it:
Throat-clearing openers "In today's fast-paced world..." "It's no secret that..." "When it comes to X..." Hedge-stacking "can potentially" "it's worth noting that" "generally speaking" False balance "Whether you're a beginner or a seasoned professional..." Empty transitions "That being said" "At the end of the day" "Moving forward" Inflated vocabulary standing in for a real claim delve elevate unlock unleash robust seamless game-changer
The practitioner consensus that holds up across every serious source: AI produces the first 50% — structure, outline, rough pass. The last 50%, where trust and originality live, is still a human job.
Start with your own voice, not the prompt box.
Record yourself talking through the idea, or write a rough unstructured brain-dump, before AI touches it. This becomes your source material — it is what keeps the eventual piece from starting life as a generic take on the topic.
Let AI handle structure and drudgery.
Outlines, H2/H3 scaffolding based on real search intent, topic-cluster brainstorming, turning a long document into a summary, expanding a keyword list. This is where AI is genuinely strongest and where it saves real time.
Write or heavily rewrite the actual sentences yourself.
Use your own material as the spine. Read the draft aloud — clichéd AI-pattern phrasing is far more audible than it is visible on a screen.
Add at least one detail only you could know.
A specific result, an actual number, an opinion you are willing to defend, a first-hand observation. This single step does more to defeat both slop-perception and algorithmic demotion than any amount of line-editing.
Fact-check everything AI contributed.
Hallucinated statistics and fabricated citations are a normal failure mode, not an edge case. Treat every AI-sourced fact as unverified until you have checked it yourself.
Cut, don't just polish.
If a sentence doesn't survive being deleted, delete it. Length that isn't earning its place is the single most common slop signal.
Google's E-E-A-T framework is the official vocabulary for what “quality” means in search ranking. Increasingly, it is also the signal answer engines use when deciding what to cite. The four components translate to a single practical question: does this content contain something that could only have come from actually doing the thing?
Generic best-practices content, however well-structured, is exactly the shape E-E-A-T is designed to rank below something narrower but real. A campaign page or blog post gets stronger the moment it contains a specific number from a real campaign, a screenshot of an actual result, or a stated opinion you are willing to be disagreed with on.
These principles predate AI and still determine whether anyone reads to the end.
Plain language, active voice. If you wouldn't say it out loud that way, don't publish it that way. One idea per sentence, one point per paragraph — bloated, multi-clause sentences are exactly what forces a reader to mentally run your writing back through AI just to extract what you meant.
Front-load the point. Readers and answer engines both reward getting to the actual answer fast, not building up to it. Concrete beats abstract, always: “Grew signups 34% in six weeks” beats “significantly improved conversion.” Specificity is also the fastest way to sound human, because generic language is what AI defaults to when it doesn't have a real number to reach for.
Match format to intent. A comparison deserves a table. A process deserves numbered steps. A narrative deserves prose. Forcing content into the wrong shape is itself a slop signal, independent of the sentences inside it.
Before publishing anything AI helped draft, run this checklist:
The writing is the thinking. Skip that part, and it shows.
Marcus Sheridan · Wiley
A revenue-first approach to content: answer every question your customers ask, with full transparency — the method behind River Pools becoming the most-visited pool site on the internet.
Google · google.com/search/docs
The actual document Google's human quality raters use — reading it directly tells you what E-E-A-T means in practice, not in interpretation.
Donald Miller · HarperCollins Leadership
Seven-part narrative framework that clarifies brand messaging — prevents the expert's curse from making content incomprehensible to the audience it's meant to help.
Continue the conversation
If this changed how you think about it — or you think I'm wrong — I want to know.
Corrections, disagreements, and applications all welcome. Replies go directly to Chris.
Get in touch →THE LAST 50%
~6-8 min1× · Two speakers · tap to play