Thin content used to be a quiet problem: a short page with little to say might rank poorly, but it mostly sat there while your strong pages did the work. That bargain no longer holds. In the AI search era, your weakest pages shape how Google and AI systems judge the rest of your site — and they give answer engines nothing worth quoting.
The rules have tightened from two directions at once: Google’s spam policies and helpful-content guidance have been enforced harder since the March 2024 core update, and the September 2026 spam update arrived with a rewritten standard for AI-assisted content. Meanwhile, AI Overviews and AI Mode answer questions directly — a page only earns visibility if it contains something an answer can lift.
This guide covers what counts as thin content now, how to audit for it, and how to decide — page by page — whether to fix, merge, or remove it.
Key Takeaways
- “Thin content” is not a term Google defines in its spam policies — it is a page that fails Google’s own helpful-content test: no original information, no substantial value, nothing beyond the obvious.
- In AI search, a thin page has no extractable passage, so it cannot be cited — and citations are already a thin substitute for clicks.
- Audit at page level in Search Console before touching anything: find the pages that lost visibility, and for which searches.
- Fix pages that serve a real question, merge near-duplicates into one strong page, and remove only what has no audience.
- Google says removing lots of content just to look “fresh” does not help rankings — remove for quality, not for optics.
- Prevention beats cleanup: a pre-publish gate using Google’s self-assessment questions stops thin pages at the source.
What Counts as Thin Content in 2026?
Thin content is a page that gives a reader — or a search system — almost nothing it could not get elsewhere: a few generic sentences, a restatement of what every competing page says, or a page built around a keyword rather than a question anyone actually asked.
Google does not publish a “thin content” policy by that name, but its guidance describes the same thing: content that provides “little to no value”, pages mass-produced without care, and material that mainly summarises others without adding anything of its own. Test it against Google’s self-assessment list: does the page provide original information or analysis, a substantial description of the topic, or insight beyond the obvious? Would you bookmark it or expect to see it referenced in a book?
The usual suspects are programmatic pages stamped out for every city, product variant, or keyword combination with only the nouns swapped; short AI-drafted posts published without a human reading them; and old posts that were adequate five years ago and have not been updated since the facts moved on. Google’s spam policies also note that it generally presumes individual pages match the overall quality of the rest of the domain — a weak section is not quarantined from the strong one. Word count alone never decides it: Google is explicit that it has no preferred word count.
Why Does Thin Content Cost More in the AI Search Era?
Thin content costs more now because AI search changed what a page must do to earn visibility. The answer is assembled from passages, and a thin page contains no passage worth lifting — nothing self-contained, specific, and verifiable for an AI Overview or chatbot to quote.
That matters more because the click economics have already shifted. A randomized field experiment by Agarwal and Sen, covered in our piece on AI Overviews cutting clicks, found that where an AI Overview appeared, outbound organic clicks fell 39.8% — a causal estimate from a preliminary working paper on US desktop users, not a universal rate, but the direction is the point. Citations inside the Overview generated only about 8% of all clicks, according to the authors’ own summary. So the visit is harder to win, the citation is a poor substitute for it, and a page that cannot even earn the citation has lost twice.
There is a second, quieter cost: AI systems and Google’s ranking systems both judge your site from the whole corpus, not just your best pages. The September 2026 spam update underlined this — Google called it normal enforcement of existing spam policies, as our breakdown of the September 2026 spam update explains.
How Do You Audit Your Site for Thin Content?
Audit at page level, in Search Console, before you change a single word. Export clicks and impressions for the last three to six months, sort by the pages earning almost no impressions, and check the queries view for the worst — a page ranking only for its own brand name, or for nothing at all, is telling you something.
Google’s advice for auditing after a drop follows the same logic: look at which pages were most impacted, and for what searches, then assess those pages against its helpful-content questions. A page that lost traffic because its topic faded is a different problem from one that never earned traffic because it says nothing.
Layer in a crawl to catch what Search Console cannot show: word counts, duplicate groups, orphan pages with no internal links, and pages with no traffic that Search Console barely registers. Then score every candidate against Google’s checklist: does it answer a real question, add anything original, stay accurate and current, and leave the reader satisfied? Pages failing most of those go on the fix-or-remove list; pages failing all of them, with no traffic and no links, go straight to removal. Check Manual Actions in Search Console first: a manual action needs a different response — fix the violation, document it, and file a reconsideration request before starting a cleanup.
How Do You Decide Whether to Fix, Merge, or Remove a Page?
Decide with three questions, in this order: does anyone need this page, does it have anything worth keeping, and can it be made genuinely useful? Real search demand makes it a fix candidate; duplicating a stronger sibling makes it a merge candidate; no demand, no links, and no original material makes it a removal candidate.
The split is worth getting right because each option costs differently. Fixing takes the most effort, so spend it where the page can actually become the best answer to its question. Merging is cheaper and often better: two half-pages rarely beat one complete one. Removing is the cheapest option of all — which is why it gets overused, and why deleting pages that still earn traffic or links is how cleanups turn into traffic losses.
For a merge, pick the strongest URL as the survivor — the most traffic, the most links, or the best position — then fold the useful material from the others into it, rewriting so the result reads as one coherent page. Redirect the retired URLs to the survivor with permanent 301 redirects, one hop each, and update every internal link that pointed at the old pages. Our guide to redirect strategy for site migrations covers the same mechanics: one-to-one mapping, no chains, no bulk redirects to the homepage, and redirects kept for at least a year. Do not merge pages that answer genuinely different questions just because they share a keyword — the test is the reader’s intent.
Work through the list in batches, leaving gaps so you can see what moved in Search Console. Changing everything at once makes it impossible to tell which decisions helped.
What Does a Good Fix Look Like?
A good fix turns a page that says nothing into one that answers its question properly, with evidence a reader — and an answer engine — can use. Re-read it as a stranger would: what question should it answer, and does the first paragraph answer it? If not, rewrite the opening so the answer lands in the first two sentences.
Then add the substance the page was missing. Original information beats added length: first-hand experience, a worked example, a comparison you ran yourself, or a definition in your own words. Google’s self-assessment questions reward exactly this, and its E-E-A-T framing puts trust first, built from experience a reader can see. Show the work: name sources, date facts, and say who wrote the page. Our guide to writing introductions AI can quote walks through the same answer-first discipline.
Structure the fix so machines can read it too: question-form headings, a direct 40–60 word answer under each one, lists and tables where the content allows them, and clean semantic HTML. That extraction-friendly structure is what our guide to structuring content for AI crawlers describes. Finish by updating the title and meta description to match what the page now says.
When Is Removing a Page the Right Call?
Removing a page is the right call when it has no audience, no links worth preserving, no original material, and no realistic future — and you have checked all four, not just assumed them. Candidates: programmatic pages that never earned an impression, dated announcements for events long past, thin affiliate pages with no editorial value, and drafts published by accident.
How you remove it matters. A page you want gone should return a proper 404 or 410, or carry a noindex tag — Google’s spam guidance tells sites hosting scaled low-value content to exclude it from Search rather than leave it indexed. A 410 (gone) is processed faster than a 404 and signals intent. Do not redirect it to your homepage or an unrelated page: Google may treat bulk homepage redirects as soft 404s.
Two cautions before you delete in bulk. First, Google is explicit that adding or removing lots of content to seem “fresh” does not help rankings — remove pages because they are low value, not because a smaller site looks tidier. Second, check the page’s links and traffic one last time: a page with even a handful of real inbound links or steady visits is a fix or merge candidate. Deleting it destroys value you cannot get back.
What About AI-Generated Content on Your Site?
AI-generated content is not banned, and it never has been — but the standard for it is now written down more plainly than before. Google’s position is that how content is produced does not determine whether it is spam; whether it is helpful, original, and made for people does. Using automation mainly to manipulate rankings violates the spam policies.
What changed on October 1, 2026, during the September spam update, was the written standard for human review. Google rewrote its generative-AI guidance to say it is “critical to manually fact-check and review all AI-generated content for accuracy and trustworthiness before publishing” — and the obligation now explicitly covers title elements, meta descriptions, structured data, and image alt text, not just the body. Those four surfaces are exactly the ones automated pipelines generate in bulk without anyone reading them.
The same October revision added a warning that sits directly on top of thin-content audits: deceptive authorship — made-up author names, AI-generated headshots, false credentials — is explicitly a low-quality signal. If your cleanup turns up AI personas anywhere on the site, removing them is part of the same job. Real attribution, real review, and real expertise are the whole defence — and the whole point of the fix.
How Do You Stop Thin Content Coming Back?
You stop it with a gate, not with good intentions. Every page should pass the same short test before it publishes: does it answer a specific question, does it add something original, is every fact checked, and would you be comfortable with an AI system quoting its opening paragraph back to a searcher?
The gate also needs to catch the structural causes. Templates that generate a page per keyword, per city, or per product variant without unique material are thin-content factories, and no cleanup fixes a factory that keeps running. The same applies to AI-assisted drafting: the tool can research and structure, but a human must read, verify, and improve the output — titles, descriptions, schema, and alt text included — before it goes live.
Finally, make the audit a recurring habit. A quarterly pass over Search Console’s weakest pages, a check of new pages thirty days after publication, and a review whenever a spam or core update rolls out will catch thin content while it is still a handful of pages. Content quality is maintenance, not a launch state — and in the AI search era, the sites that maintain it are the ones whose pages keep getting quoted.
Frequently Asked Questions
Is thin content a Google penalty?
Not by that name. Google has no spam policy headed “thin content”, but the behaviours behind it are covered: scaled content abuse, scraping, doorway pages, and the helpful-content standard. A thin page usually loses visibility through ranking rather than a manual action — though the worst cases can draw one.
How many words make a page “not thin”?
There is no number. Google states plainly that it has no preferred word count, and a 300-word page that answers its question completely beats a 2,000-word page padded with filler. Judge by whether the page satisfies the reader’s question, not by its length.
Should I delete all my low-traffic pages?
No. Low traffic alone is not a diagnosis — a page can be new, seasonal, or aimed at a small but valuable audience. Delete only pages that fail on every front: no traffic, no links, no original value, and no plausible future. Fix or merge the rest.
Does removing thin pages help the rest of the site rank better?
Google gives no guarantee of that, and removing content just to look “fresh” does not help rankings. The honest case is narrower: it stops your site vouching for pages that fail its own quality standard, and it focuses crawl and attention on the pages worth ranking.
Can a fixed thin page recover its rankings?
Sometimes, and never on a schedule. Google gives no fixed timeline for recovery, and spam-related demotions are typically measured in months rather than days. Fix the page properly, make sure it is crawlable and linked, and judge it over quarters.
Thin content was always a liability; AI search just made the bill come due faster. Audit honestly, fix what serves a real question, merge what splits one answer across weak pages, and remove only what has nothing to offer. Then gate every new page, so the next audit finds a smaller pile — and the pages that remain are the kind an answer engine is glad to quote.
