You’ve read some today. The email that takes four paragraphs to ask one question. The post that sounds thrilled about nothing. The report where every sentence reads fine and the whole says nothing. Researchers at Stanford’s Social Media Lab and BetterUp call it workslop: AI writing that looks finished but can’t move the task forward.
The economics are brutal. Producing words now costs almost nothing. Reading them costs what it always did. Every minute a writer saves by not thinking gets billed to a reader who has to. That transfer is the slop tax, and by 2026 it has a major business impact.
In September 2025, BetterUp and Stanford surveyed 1,150 US desk workers. Four in ten had received workslop in the past month, and each piece took nearly two hours to fix. The bill lands on reputation too. 42% trusted the sender less afterward. About half thought less of their ability. Nearly a third wanted to stop working with them. People fix the draft, and they remember who sent it.
The platforms noticed. In July, LinkedIn added a “seems like AI slop” report button and says it blocks hundreds of thousands of automated comments a day. Within weeks, more than a million people had used the button. Slop is now something readers report, not just something they skim past. Ryan Roslansky, who ran LinkedIn for six years, called the pattern a doom loop this week: documents nobody wrote, sent to inboxes nobody reads, “a very expensive way to avoid writing and reading.”
Workday’s January 2026 research found the same leak at scale. 85% of employees save one to seven hours a week with AI. Nearly 40% of that saved time goes straight back into correcting, rewriting, and verifying machine output. 77% of daily users review AI work as carefully as human work, if not more. Only 14% consistently come out ahead.
The dangerous part: slop feels like speed. In METR’s 2025 controlled study, experienced developers expected AI to make them 24% faster. Measured, they were 19% slower. One narrow setting, sure. But the work felt quicker while it was getting slower, and that gap is where slop lives. Stack Overflow’s survey of 31,476 developers rhymes with it: two thirds named “almost right” answers their top frustration.
Slop isn’t a vibe. It’s measurable. Dmitry Kobak’s team scanned more than 15 million biomedical abstracts and found LLM fingerprints on at least 13.5% of 2024’s papers, with “delve” suddenly everywhere after decades of minding its own business. Rhythm leaves prints too: people scatter their sentence lengths, models keep an even pulse. What you can’t do is reliably catch the author. The RAID benchmark showed detectors collapse under light paraphrasing. Fine. The useful question was never “did a machine touch this?” It’s “does this waste my reader’s time?”
That question has a fix, and it’s a workflow, not a vibe check. Treat model output as a draft, never a deliverable. Verify that every name, number, quote, and link survived, because fluent rewrites drop facts quietly. Read it aloud once; your ear catches the even pulse your eye forgives. Cut any sentence that could sit unchanged in a competitor’s document. Then measure it before you send it, and believe the score over the glow of finishing early.
None of this is anti-AI. Gallup finds 65% of employees at AI-adopting companies report positive impact. The tools are fine. Shipping their first draft is the problem. Ten minutes of verification against two hours of someone else’s cleanup isn’t a close call.
That’s why Zero Slop exists: free, open source, scores your draft locally, rewrites it without touching a fact, and learns your taste from the edits you keep. The machines made words cheap. Attention still isn’t. Spend it like it’s the reader’s money, because it is.