The internet found a word for what it’s been feeling

Sometime in the last year, "slop" became the word. Not "content." Not "AI-generated material." Slop, the same word you’d use for something scooped into a trough, applied without apology to the wave of AI-written articles, AI-drawn images, and AI-voiced videos now clogging search results and social feeds. Developer Simon Willison helped push the term into wider use in the spring of 2024. His reasoning was blunt: unwanted, unreviewed AI output thrust on people who didn’t ask for it deserves an ugly name, because that’s what it is (Willison). By late summer, The Atlantic was tracing how that same low-effort content had become a deliberate engagement-farming tactic, not just a byproduct of lazy shortcuts (Warzel). By fall, New York Magazine ran a widely shared piece cataloguing just how much of the internet’s new content had stopped being made by anyone who cared (Read).

I read all three of those pieces the year they came out, and I felt something I wasn’t expecting. Not anxiety about AI. Relief.

Here’s why. For two years now, every conversation I’ve had about AI and creative work has circled the same question: is this going to make the work worse. Slop is the answer to that question, made visible. It’s what happens when speed replaces judgment entirely, when nobody reads the draft before it ships, when the goal shifts from saying something true to saying something, fast, at scale. Slop isn’t a mystery. It’s not even really about the technology. It’s about what happens when a tool that can generate anything gets handed to someone who’s decided nothing needs a second look. The tool didn’t create that problem. It just made it a lot easier to have at industrial scale.

Which means the fix was never going to be banning the tool. The fix is the thing slop skips entirely: a person with real judgment, standing between the draft and the world, deciding what actually earns its place.

“If it’s mindlessly generated and thrust upon someone who didn’t ask for it, slop is the perfect term for it.”

Simon Willison, "Slop Is the New Name for Unwanted AI-Generated Content," May 8, 2024

What clients are actually afraid of, and it’s not the tool

I want to be specific about this because "clients are worried about AI" gets said a lot and it’s too vague to be useful. In the conversations I’ve had, nobody walks in afraid of the technology itself. They’re afraid of three specific things, and all three are trust problems, not tool problems.

They’re afraid of getting something generic, a brand voice that used to sound like them starting to sound like every other brand’s AI output, because the model defaults to the statistical average of everything it’s ever read, and the average was never their standard. They’re afraid of getting lied to, of paying for craft and quietly receiving an unedited first draft instead, with nobody telling them the difference. And they’re afraid of getting caught. More than one founder has told me, unprompted, that their real nightmare is a customer or a journalist spotting the tells in something their own brand put out into the world, the phrase that no human would actually write, the sentence rhythm that gives it away. That fear isn’t about AI ethics in the abstract. It’s about their name being on something they didn’t actually make.

Every one of those fears is really the same fear wearing three outfits: did a person actually stand behind this. That’s a trust question, and trust questions don’t get solved by avoiding the tool. They get solved by being radically clear about how the tool was used, every time, without exception.

AI sits at my table. It doesn’t run the room

I’ve written about this before, first when I laid out where I stand on AI in the abstract, and again when I walked through one real headline project and the decision I made about what to tell that client afterward. The rule underneath both pieces is the Picasso Standard: the issue was never the tool, it’s misrepresenting how the work got made. That’s the standard in miniature. This is the standard at scale.

What hasn’t changed, and what won’t change while I’m doing this work, is what happens after the tool does its part. Nothing leaves my hands that a person with real craft hasn’t read, judged, and either approved or rejected. Not a headline, not a caption, not a strategy deck, not a line of client-facing copy. If a draft doesn’t hold up, it doesn’t ship, whether it was written by an intern, a senior strategist, or a language model. The standard was never about who or what produced the first version. It’s about who stood behind the final one.

That’s a discipline, not a slogan, and disciplines only mean something when they cost you something. It would genuinely be faster, some weeks, to skip the review pass and send the first AI draft straight to the client with a light polish. Nobody would necessarily know. That’s exactly the moment the standard matters, because a rule that only holds when it’s convenient isn’t a rule. It’s a preference.

I also tell clients plainly when AI was part of a project’s process. Internally, I log it specifically: which tool, which stage, what it touched. Client-facing, I keep it simple: human-led, AI-assisted, said plainly, the same way I’d tell a client which writer was on their account. I’ve found that clients don’t flinch at the disclosure. They flinch at finding out later that I didn’t offer it.

McKinsey’s most recent global survey found that seventy-eight percent of organizations now use AI in at least one business function, up sharply from the year before (McKinsey & Company). That number tells you adoption is no longer the story. Adoption happened. The story now is what separates work that used AI well from work that used it as a shortcut around care, and that separation is the entire opportunity in front of anyone doing this work right now.

Why slop is good news for the people who refuse to make it

A flood of low-effort AI content sounds like a threat to anyone in creative work, and in a shallow reading it is. Search results get noisier. Inboxes fill with unsolicited "we can build your brand with AI in a week" pitches from outfits that never intended to have a person read the output. It’s harder for a client to trust anyone claiming to build brands right now, because so much of what’s flooding their feeds is proof that plenty of people won’t.

But that same flood is also the clearest argument I’ve had in years for what I actually sell. When everyone can generate content instantly, generating content stops being the differentiator. Judgment becomes the differentiator. The willingness to say no to a draft, even a fast and technically competent one, because it doesn’t sound like anyone, becomes the differentiator. That was always true. Slop just made it undeniable.

I don’t think the answer is a purity test about which tools an agency is allowed to touch. That’s not honest, because I use these tools daily and I’d rather say so than perform otherwise. The answer is a standard that doesn’t bend regardless of which tools are in the room. Research fast. Draft fast, if the tool helps. Then slow all the way down at the part that actually matters: the reading, the judging, the deciding what’s good enough to carry a client’s name.

“AI sits at my table. It doesn’t run the room.”

David Keyes

What this means for anyone building with me

If you’re evaluating a creative partner right now, here’s what I’d actually ask any agency, including me: what happens between the draft and the thing with your name on it. Who reads it. What do they reject. Will they tell you, plainly, when a tool was part of the process, or will you find out by accident later.

Those questions separate craft from slop, and they have nothing to do with which software sits open on someone’s desktop. I use AI openly to move faster on the parts of the work that benefit from speed. I don’t let it near the part of the job that actually earns your trust, which is a person deciding what’s good enough to be yours. That’s not a policy I adopted because slop became a headline. It’s the standard my work was built on. Slop just made it easier to explain why it matters.

If you want a partner who uses every tool available and tells you the truth about how the work got made, every time, let’s talk.