Quick Answer
AI slop is mass-produced machine-written content published with no human judgment behind it, and the rules against it are already written down. Google Search calls it scaled content abuse, YouTube calls it inauthentic content, Wikipedia deletes the obvious cases on sight, and Amazon makes you declare it. None of them ban AI, and all of them ban publishing at volume with nobody accountable for what went out.
Most of the anti-slop talk this year has been about taste. The part that reaches your traffic is duller and more useful.
Five of the platforms you publish on now carry the rule in their own documentation. Enforcement is named. None of them mention quality in the abstract, because they describe a production pattern instead.
If you write with a model open in the next tab, permission stopped being the question a while ago. This covers what each of the five documents says, the one thing all five ban, and the rule worth writing down before somebody writes it for you.
Five Platforms Have Written AI Slop Rules Down
Yes, the rules exist, and each one carries a consequence more concrete than disapproval. Each describes a production pattern rather than a tool. That is why none of them can be satisfied by switching models.
What follows is what each document says as it reads in September 2026. Every line was read on the policy page itself, not in the coverage of it.
1. Google Search Calls It Scaled Content Abuse
Google's spam policies define the practice in one sentence, on a page stamped last updated 28 August 2026. Scaled content abuse is, in Google's words, “creating large amounts of unoriginal content that provides little to no value to users, no matter how it's created.”
That last clause is the whole policy. The method is not the offense. A page a person typed to the same recipe breaks the same rule.
The examples name AI tools used to make many pages without adding value, scraping feeds into pages, and stitching other people's pages together. Google says it finds this through automated systems and human review, which can end in a manual action. The same page also names attempts to manipulate generative AI responses in Google Search as spam, which is worth reading line by line.
2. YouTube Withholds Ads From Inauthentic Content
YouTube's monetization policy no longer says “repetitious content.” Its own dated note records the change: on 15 July 2025 the rule was renamed to inauthentic content, and clarified to cover work that is repetitive or mass-produced.
Under that heading sits a line aimed squarely at AI slop. It covers content made with “generic or unoriginal templates giving the impression of mass production,” with none of the creator's own insight added. The review standard on the same page is blunter, and asks that content “not be mass-produced, generic, repetitive, or manipulative.”
The YouTube Partner Program still pays for video a model helped make. It pays when the model executes a vision the creator brought, and stops when the template does the creating.
3. Wikipedia Deletes Unambiguously LLM-Generated Pages
Wikipedia went further than withholding money. Criterion G15 of its speedy deletion policy lets one administrator delete a page outright, with no discussion first. It applies to pages showing signs that they “could only plausibly have been generated by large language models.”
Three signs qualify, and all three are evidence rather than style. Text aimed at the user, such as a knowledge-cutoff line or “as a large language model.” References that do not exist. Residue from the model itself, such as a citation artifact left in the markup.
Notice what is missing from that list. Nothing about cadence, nothing about the em dash, nothing a careful reader could mistake for a house preference. The criterion is narrow on purpose, because a broad one would sweep up human editors who write tidily.
4. Amazon KDP Makes You Declare AI-Generated Text
Amazon asks for disclosure rather than prohibition. It draws the line in the place most writers get wrong. Its content guidelines say content a tool created “is considered ‘AI-generated,’ even if you applied substantial edits afterwards.”
Kindle Direct Publishing calls the other case AI-assisted. You wrote it, and a model edited, refined or error-checked it. Only the first case has to be declared, at upload, every time.
So heavy editing does not launder a machine draft into your own work. It is the cleanest statement of the principle any of these five documents makes.
5. Medical Journals Refuse AI Authorship Outright
Academic publishing states it most flatly of the five. The International Committee of Medical Journal Editors says chatbots “should not be listed as authors because they cannot be responsible for the accuracy, integrity, and originality of the work.” Its recommendations also require authors to disclose any AI-assisted technologies they used.
The reasoning is worth borrowing even if you never submit to a journal. Authorship is an accountability claim, and a model cannot be held to one.
What All Five of These Rules Actually Ban
Not one of the five bans AI. What they ban is volume with nobody behind it, and that is a production decision a founder makes rather than a tool a founder buys.
The same three elements appear in each document. Output produced at scale. Content that is unoriginal, in the sense of adding nothing a reader could get elsewhere, and unattributed, in the sense that no person is answerable for the claims in it.
Strip any one of those three away and the AI slop label stops applying. That is the loophole, and it is a real one.
The scale problem is measurable. NewsGuard's AI Tracking Center counted 3,749 sites it classifies as AI content farms, across 16 languages, on 23 June 2026. Those sites publish machine-written news with minimal human oversight and no disclosure, and they are the population these policies were drafted against.
Most advice on content marketing for tech companies still treats volume as the lever. That is the behavior the policies now describe. The alternative is not writing less, but making sure each page carries a claim you can defend, which is why tracing every number back to its source is the practical version of this rule.
There is a second cost nobody writes into a policy. A team that stops wrestling with its own material stops learning from it, and the voice goes first.
How the Five AI Slop Rules Differ in Practice
The five agree on what is wrong and disagree on what to do about it. That matters, because the consequence decides how much of your attention each rule deserves. A deletion is immediate in a way a ranking effect never is.
| Policy | What triggers it | What happens |
|---|---|---|
| Google Search | Many pages with little value, however they were made | Lower ranking or removal, algorithmically or by manual action |
| YouTube Partner Program | Templated or mass-produced video with no original perspective | Ads withheld from the content |
| Wikipedia | Three named signs of machine origin, such as fake references | Page deleted by one administrator, no discussion |
| Kindle Direct Publishing | Content a tool created, later edits included | Disclosure required at upload, or the listing is at risk |
| Medical journals | A chatbot named as an author, or undisclosed AI use | Authorship refused, disclosure demanded |
Only Wikipedia acts on machine origin by itself. The other four act on the pattern around the work.
So the safest reading of all five is the strictest one. Assume nobody is checking whether a model wrote it, and everybody is checking whether it was worth publishing.
What This Changes for a Founder Publishing With AI
For a five-person company publishing twice a month, almost nothing changes at the enforcement layer. Two posts a month is not the production pattern scaled content abuse describes. Reading these rules as an imminent threat to a small publisher would be a forecast dressed up as a warning.
What changes is the standard you are read against. When much of what a buyer reads is machine-written, the specifics only you can produce become the thing that marks the page out. AI slop is the baseline your reader has already learned to skim.
That is the part an AI SEO service cannot skip past. It is where Better Marketing starts on a page: what this company knows that nothing else on the web knows. Trust is built in specifics, never in slogans.
A working rule fits in five lines. Write it down before you need it.
- A model may research, outline, summarize, edit and draft internally.
- Nothing reaches the public as a first draft without substantive human revision.
- No claim about a customer, a price, a security property or a product behavior comes from a model.
- A named person owns every public page and can defend every sentence on it.
- If that person cannot explain or improve it, it does not go out.
The test before publishing is one question. Would a knowledgeable buyer believe you wrote this because you know something useful, or because you know how to fill a page?
When Writing More Pages Is the Wrong Move
Sometimes publishing is not the constraint. Tightening an AI slop standard on work you should not be doing yet is motion rather than progress.
Content is a slow instrument. Content marketing for tech companies compounds over quarters, so if you need first customers inside eight weeks, founder-led outbound is the faster one. That is the job a B2B demand generation agency exists to run.
The same applies before product-market fit. If the thing you would write about changes every three weeks, the pages are wrong before they rank. The discipline that matters is talking to the twelve people who might buy.
Treat these policies as the standard your work has to meet when you do publish, not as a reason to publish and not as a reason to hand the whole problem to a B2B demand generation agency instead. Ads are rent. This is equity, and equity takes time to accumulate.
Write Your Own Rule Before a Platform Does
The anti-slop rules are not coming. They are written, dated and on the record. The only real question is whether your process would survive being read against them.
Ours is one line: a model can draft, but every number on a page has to trace to a source a reader can open. That is how Better Marketing writes.
Write your own version this week, name the person accountable for each public page, and judge any AI SEO service you hire by whether it starts in the same place. AI slop is what happens when nobody does that.