How Your Old AI Slop Is Dragging Down Rankings

That craptent surge? It never worked anyway…

This is one more delicious/tragic irony: The rankings those craptent-dependent B2B marketers were chasing were something they never reached.

According to the study above, 86% of articles ranking in Google Search are authored by humans. Just 14% are AI-written. And 82% of the articles that ChatGPT and Perplexity cite are human, with only 18% springing from LLMs. Moreover, how do you think AI-generated articles rank when they do appear in Google? If you guessed “lower,” give yourself a gold star.

One wrinkle here is that the study didn’t evaluate AI-assisted content in which there’s been a heavy human hand in its editing and revision. Which may (probably, IMHO) be more effective. Unsupervised or minimally overseen content, though, will suffer like the rest.

Where a craptent content corpus can catch up with you

So here’s the problem. Any issues with AI content quality aren’t confined to just the page and website where a marketer has published this kind of material.

Google has its knives out for “scaled content abuse” that’s intended to jack up rankings. Their own spam policies explicitly state that a page or an entire site can be de-ranked or just omitted if it’s spamming content.  In their eyes, pushing out content at scale just to manipulate your search rankings, whether it’s via AI or by humans, is liable to get flagged by the Search Police.

So a content abuser might not get slammed just for a specific page. Google confirms it passes judgement based on “site-wide signals,” so abuse in one corner of a site can lead to penalties that impact the entire site. That traffic a B2B marketer was chasing  with AI slop can actually go into reverse gear because of the AI slop.

How’s that for an unintended consequence?

The AI overviews got tougher to crack

There’s a single date where you can mark the transition from this being an SEO problem to being an AI visibility problem. That was May 15, 2026, when Good updated its Search Central documentation.

That’s when it clarified its spam policies with regard to generative AI responses in search, including AI Overviews and AI Mode. Every policy now governs what appears in AI answers, and those rules address scaled AI content abuse. Core ranking systems now focus on high-quality content, and different systems block content spam. AI search is subject to both.

What’s this mean? That those 2024-25 posts a marketer happily churned out using AI may now be a dead weight that’s dragging your visibility down, down, down. 

But outside of Google..?

There are two caveats to this, so let’s check them out.

Only Google is doing this: First off, no LLM vendor has published something that resembles Google’s site-wide demotion mechanic, so Perplexity, Claude, ChatGPT, etc., aren’t executing bans themselves. But there’s inference from what they can retrieve. When their bots go a-crawlin’ for something citeworthy, AI-drafted pages mainly lose. It’s like a bake-off where you cake didn’t take the ribbon, but that’s not a judgement on your entire bake shop. So your site might not take a hit from LLM searches. Nobody’s sure yet.

Once everyone knows the tricks, they stop working: When a team of researchers tested some of the most prevalent tactics for getting cited by AI across six subject areas, they discovered that most of them had negligible effect. What did work? That old, boring tactic about being the most relevant and useful source on the topic. But it only works until everybody else starts doing it, at which point the benefits start to disappear. So a marketer ought to be looking for new and meaningful things to say to maintain their advantage.

What to do to ensure AI visibility