ai slop is machine-made content, produced in bulk, pointed at people who never asked for any of it. for two years that was an insult you threw in a comment section. in 2025 it grew paperwork. a dictionary defined it. google wrote it into spam policy. spotify pulled seventy-five million tracks and published the number.
the shape of it never changes. one prompt becomes forty posts. forty posts become a channel. the channel exists to catch a thumb on its way past. nobody reads the output, and that includes whoever hit publish.
the dictionary version
macquarie dictionary made ai slop its word of the year for 2025 in november 2025, defining it as low-quality content created by generative ai, often containing errors, and not requested by the user. three clauses, all of them load-bearing.
low quality is the part everyone can already see. errors are the diagnostic part, because a model with no source for a claim will produce the claim anyway, and it sits there on the page looking exactly like the true sentences on either side of it. the third clause is the one people skim, and it decides everything. a draft you asked a model for and then edited is a tool doing a job. the same text fired into a feed you happened to be scrolling is slop. identical words, different situation.
the announcement was also the fourth time in the dictionary's history that the committee's choice and the public vote landed on the same word. people agree about this one.
google calls it scaled content abuse
search has been fighting this since long before the models got good, and it uses much drier language. google defines scaled content abuse as when many pages are generated for the primary purpose of manipulating search rankings and not helping users, and names using generative ai tools to generate many pages without adding value for users as an example, on a policy page last updated 2026-08-28.
read the trigger closely. it is scale plus intent, both at once. one model-assisted explainer on your site passes. four thousand of them trip the rule. the same page also counts scraping feeds, search results or other content to generate many pages, including automated transformations such as synonymifying and translating, which shuts the thesaurus loophole before anybody gets excited about it.
that clause is older than the current wave. spinning articles is an ancient trick and the policy already covered it. generative models just made a tired scam cheap enough for everyone to run at once.
spotify and the seventy-five million
music shows the volume problem most clearly, because a track is small and a catalogue has no ceiling. when spotify tightened its rules on 25 september 2025 it said it had removed over 75 million spammy tracks in the past 12 months. that is one service, over one year.
the response came in three parts, and each one has an obvious equivalent for anybody posting into a feed.
first, a spam filter that identifies uploaders using mass uploads, duplicates and seo hacks and stops recommending them, rolled out gradually so legitimate creators were not caught in it. second, an impersonation rule, which allows vocal impersonation only where the impersonated artist has authorised it. third, a disclosure standard built by ddex that lets artists mark where ai contributed to vocals, instrumentation or post-production, with tens of thousands of ai credits submitted daily.
the third part is the sneaky one. tens of thousands of ai credits a day means the honest use of these tools is enormous and completely ordinary. so all three policies aim at the same target, which is volume. and the punishment is demotion. you get no warning email, no strike, no notification. your reach flattens and stays flat while you keep posting into a room that stopped listening.
the tells
slop identifies itself inside the first paragraph, and grammar has almost nothing to do with it. the giveaways are structural.
- every section runs the same length, because nothing in it had more to say than anything else
- claims show up with no name, number or date attached
- the closing line restates the opening
- the specifics could be swapped for any other specifics and not one sentence would break
- nothing in it is wrong in an interesting way
the last one is the tell that never misses. a person writing about something they know will get a detail slightly wrong, pick a fight with the obvious answer, or spend two hundred words on a thing only they care about. slop does none of that, because there is nobody inside it to argue with.
what it means if you post
models are allowed. use them. what all three policies chase is a human being having made a call somewhere in the chain, and each of them measures that differently. google weighs pages built for rankings against pages built for readers, spotify weighs mass uploads against catalogues, and a feed weighs whether anybody stopped scrolling for long enough to count.
which makes specificity the entire defence. one post about something you saw with your own eyes is what all three policies are trying to protect, and forty assembled from what a model guessed about it are what they are built to catch. the accounts that come through this intact are the ones holding something that could only have come from the person running them.
if you want the mechanical version of that argument, why posts spread covers what a feed is measuring when it decides whether to show your post to more people.
here is the boring practical bit that never makes it onto a policy page. the extra work is tiny. one real detail you did not get from a model. one number you looked up yourself. one opinion you would still defend in a group chat at two in the morning. do that per post and the post stops matching the description in any of the three policies.