Google Is Now Penalizing AI Content As Thin Content.

On 27th july 2026, Google just issued a manual action against a 20-year-old forum for "thin content with little or no added value." The flagged pages weren't spam. They were AI-generated answers, and that distinction matters for every brand running AI SEO at scale.

The Manual Action That Broke The Internet's Assumptions

In July 2026, a large tech forum received a partial manual action under code WNC-651700. Google's notice pointed at one URL pattern: the site's /threads/ directory, covering roughly 168,000 threads and half a million posts. Two other sections of the same site, an old syndicated headline feed and an archived knowledge base, were arguably thinner. Neither got flagged.

The penalized section had an AI chatbot posting as a "staff member" since 2023, with over 100,000 auto-generated replies.

SEO & AI Search Consultant Gagan Ghotra flagged this pattern publicly, and it lines up with what Google has said for years about thin content. The volume of the message didn't trigger the action. missing value did.

What Thin Content Actually Means Under Google's Guidelines

Google's spam policies define thin content narrowly, but the category has expanded in practice. The core examples:

• Thin affiliate content: pages that copy merchant descriptions word for word, with no original analysis
• Auto-generated content: text created programmatically without human review or added insight
• Scraped content: material lifted from other sites with minimal alteration
• Doorway pages: pages built to rank for a query but funneling users elsewhere
• Content from automated tools without review: this is the category most AI SEO programs risk falling into

Former Google engineer Matt Cutts described it precisely years ago; thin content lacks original insight, research, or analysis. The opposite is content that reflects genuine expertise.

AI answers, however grammatically clean, are received knowledge. They didn't come from lived experience with the product being discussed, so they don't satisfy what a forum visitor came for.

Can humans also write thin content without using AI?

Yes, humans can write thin content, too. Thin content isn't an AI problem. It's a value problem. Humans have been publishing thin content since before AI writing tools existed:

• Article marketing content: keyword-stuffed posts written fast, for volume, with no depth

• Syndicated republishing: the same press release or blog post pasted across a dozen client sites

• Templated location pages: a service page swapped city-by-city with zero local specificity

• Padding for word count: intros that stall the answer, filler sentences added to hit a target length

Google's Helpful Content System, live since 2022, doesn't check whether a human or a model wrote a page. It checks whether the page serves the person who searched. A 2,000-word blog post written entirely by a human, full of generic advice and no original data, fails the same test as a mass-produced AI page.

This helps us conclude that authorship was never the problem; value was.

Why AI Content Gets Flagged More Often

AI content fails this test more often for a structural reason. Models trained on existing web content tend to reproduce the median take on a topic. Ask five different tools to explain a concept, and you'll get five versions of the same framework, the same examples, and the same structure.

At scale, that similarity is precisely what Google's spam systems are built to detect: repeated patterns across many URLs, no unique data points, no named expertise behind the claims.

That's the real risk for AI SEO programs. Not that AI wrote the draft. The real risk is that the draft never left the "average of the internet" and was published anyway.

How do you fill the gaps in thin content and make it user-friendly and informational?

Treat every AI-assisted draft as a first pass, not a finished asset. Run it through four checks before publishing.

1. Add a data point AI can't generate: Pull a number from your campaigns, a client result, a proprietary survey, or a tool you built. If the page has zero data your competitors don't also have, it's thin by definition.

2. Attribute claims to a named expert: A quote from your Head of SEO, a specific methodology your team uses, and a named case reference. Bylines and validated input are trust signals that both Google's reviewers and ranking systems weigh.

3. Cover the follow-up questions, not just the headline query: Thin pages answer the title and stop. Full-coverage pages answer the obvious next three questions a reader has in the same visit.

4. Run a duplication check across your site: If five of your service pages share the same intro structure, same examples, and same call to action, you have internal thin content. Google's systems detect pattern repetition within a domain just as easily as across the web.

5. Track engagement, not just rankings. High impressions with low click-through and short time on page are the leading indicators of thin content before a manual action ever lands. Audit those URLs first.

Bottom Line

Google is penalizing pages that skip the step where a human adds something the internet didn't already have.

AI SEO works when the model handles structure, speed, and first drafts, and a person handles judgment, data, and expertise.

Skip that second half, and it doesn't. It doesn't matter whether a machine or a marketer wrote the page. Google will treat it the same way: thin.