Thin Content: There Is No Word Count, and There Is No Thin Content Penalty
· Royking Niba
Search Google’s spam policies for the phrase thin content and you will not find it. There is no policy by that name, no manual action called thin content, and no minimum word count anywhere in Google’s published documentation. The term is industry vocabulary that got adopted so widely people started treating it as a Google concept, and then started fixing sites against a threshold Google never set.
That does not mean weak pages are safe. It means the actual mechanisms have names, and the names tell you what to do. Three things are usually going on, and they have different remedies.
The three real mechanisms
| Mechanism | What Google actually says | How it lands | Remedy |
|---|---|---|---|
| Scaled content abuse | Many pages generated for the primary purpose of manipulating search rankings and not helping users, however produced | Manual action or algorithmic demotion | Remove the pages and the generator behind them |
| Thin affiliation | Publishing content with product affiliate links where the descriptions and reviews are copied directly from the merchant without original content or added value | Manual action | Add real testing, pricing insight or comparison, or drop the pages |
| Doorway abuse | Sites or pages created to rank for specific, similar queries, leading users to intermediate pages that are not as useful as the final destination | Manual action or demotion | Consolidate into the page the user actually wanted |
| Ordinary quality assessment | Not a violation at all. Core updates are broad recalibrations, not penalties | Gradual ranking loss across a core update rollout | Make the pages genuinely better, or reduce their number |
The fourth row is where most so-called thin content problems actually sit, and it is the one people misdiagnose most expensively, because they go looking for a violation to fix and there isn’t one. Nothing is broken. The pages are simply not good enough to beat what else is available, and a core update re-evaluated that.
Why the word count test fails
Read those three policy definitions again and notice what is absent from all of them: any reference to length. Scaled content abuse is about purpose and volume. Thin affiliation is about originality and added value. Doorway abuse is about interposing a page between the user and what they wanted. A 300-word page can satisfy every one of those tests. A 3,000-word page can fail all three.
I have watched a client spend eleven weeks padding 1,800 product pages from 90 words to 400 words, and the padding was written to a template. What they produced was 1,800 pages that were longer and still interchangeable, which is a worse position than where they started, because they had now also spent the budget. Length was never the variable.
The test I use instead
Three questions per page, in order. The first failure ends the assessment.
- Can a reader finish their task on this page, or do they have to search again? A page that sends people back to the results is thin regardless of its length.
- Is there anything here that a general-purpose language model could not have produced without this business? Original data, testing, client outcomes, pricing, stock, local knowledge, a photograph of the actual thing. If there is nothing, the page is commodity content and it will be treated as such.
- Does this page exist because someone needed it, or because a template generated it? The honest answer is usually obvious and usually uncomfortable.
That second question is the whole game, and it is not my invention. Google’s own guidance on generative AI features tells publishers to create valuable, non-commodity content and to bring a unique point of view, and states that doing so influences presence in AI search more than anything else. The same property decides ordinary rankings. A page nobody could have written but you is defensible. A page anyone could have written is competing on authority alone, and most sites lose that competition.
Triage: what to do with a few thousand weak pages
The instinct is to delete everything, and it is usually wrong, because a chunk of those pages earn links or traffic that nobody has measured. This is the sort I run. Every page goes in exactly one bucket.
| Bucket | Criteria | Action | Typical share |
|---|---|---|---|
| Improve | Real intent behind it, some traffic or links, and you have something genuine to add | Rewrite with the non-commodity ingredient. Keep the URL. | The smallest bucket, and the most valuable |
| Consolidate | Several pages chasing one intent, none of them winning | Merge into the strongest URL, 301 the rest | Large on sites with a tagging habit |
| Remove | No traffic, no links, no intent, no plan | 410 for genuinely gone. Do not redirect it into a healthy page. | Usually the biggest bucket |
| Noindex and keep | Useful to users, never going to rank, and you want it crawled and followed | Noindex, follow. Remember this saves no crawl budget. | Small: filters, internal utility pages |
| Leave alone | Earning something, even if it looks thin to a crawler tool | Nothing. Measure before you cut. | Always larger than people expect |
The sequencing rule that saves the most grief: measure contribution before you delete anything. Export traffic, referring domains and assisted conversions per URL first. I have seen a mass pruning take out the twelve pages that held a site’s entire link profile because they were short, and short was the only filter anyone applied.
The second rule concerns redirects, and it is the same one I apply to any demoted section: serve 410 for content that is gone rather than redirecting it somewhere healthy. Redirecting a thousand removed pages into a category page does not transfer their value. It tells Google that a category page is the best answer to a thousand different queries, which is how a contained problem becomes a sitewide one. That mechanism is the same one I describe when unwinding a borrowed-authority section in parasite SEO and site reputation abuse.
How long recovery takes
If a manual action is attached, the path is legible: fix, file, wait for the review. If the loss was algorithmic, there is no notice and no ticket, and the reassessment happens when Google next re-evaluates the site. That is slower and it is worth being honest with clients about, because the gap between doing the work and seeing the result is where most pruning projects get abandoned halfway, leaving the site in a worse state than either the before or the after.
Telling those two cases apart before you start is the single most useful hour you can spend, and it is the subject of manual action penalty or algorithmic demotion. The wider diagnostic order, including how to establish whether you are looking at a penalty at all, is in my pillar on Google penalty recovery. And if you are auditing rather than firefighting, the content pass that counts thin and duplicate pages sits inside the SEO audit checklist.
Royking Niba is an SEO and GEO consultant specialising in penalty and spam-update recovery, with more than 8 million organic visits recovered for clients. Every policy definition quoted here was checked against Google’s published spam policies on 22 September 2026.
Leave a Reply