You have probably heard the rule: your content needs to be a certain percentage unique or Google will penalize you for duplicate content. Thirty percent new text, some people say. Others say forty. The number changes depending on who you ask, and that alone should tell you something about how much of the advice on duplicate content seo is guesswork dressed up as fact.
Here is the honest verdict. Two pages can share wording, structure, or a topic without an automatic SEO penalty. But if their main content is the same or very similar, Google can group them together and usually show just one of them in search results. There is no published percentage that guarantees two pages will both rank. The clustering behavior is confirmed, straight from Google's own documentation. The specific number everyone quotes is not confirmed anywhere, because Google has never published one.
That gap between what is confirmed and what gets repeated as fact is worth sitting with for a minute, because it is where a lot of wasted rewriting comes from.
Where the "must be X% unique" claim goes wrong
The claim you have probably run into combines two separate ideas, and neither one holds up the way people assume.
The first idea is that there is a specific number, a word count or a percentage of changed text, that Google checks against. If your page clears that bar, you are safe. If it does not, you get flagged. This sounds precise, which is part of why it spreads so easily. But Google's search documentation does not describe a threshold like this anywhere. It talks about primary content, meaning the substance of a page, being the same or appreciably similar to another page's primary content. It never attaches a number to that comparison.
The second idea is that falling short of that bar triggers a penalty, the same way keyword stuffing or hidden text might get a page demoted. Google addressed this directly in a Search Central post from 2008, and the answer was clear: there is no penalty for ordinary duplicate content in the sense most people mean it. Google does not punish your whole site because two of your pages resemble each other.
So the popular version of the rule fails on both counts. There is no numerical cutoff to hit, and there is no automatic punishment for missing one. What actually happens is more specific and, honestly, more useful once you understand it: content uniqueness seo comes down to whether each page answers a genuinely different question, not whether it clears some percentage.
What Google actually does with similar pages
Google's current documentation describes a process called canonicalization. When Google finds multiple pages with the same or very similar primary content, it groups them into a cluster and picks one page to represent that cluster in search results. That chosen page is called the canonical. It gets crawled more often. The other pages in the cluster, the duplicates, get crawled less frequently and typically will not show up as separate results, even if a person searches for something that would technically match them too.
This is not a penalty. It is closer to Google doing some housekeeping on your behalf, because nobody wants five nearly identical results clogging up a search page. Google's older 2006 explanation described duplicate content as blocks of text that match completely or are appreciably similar across URLs, and it noted plainly that people do not want to see the same result repeated over and over. That is still the underlying logic today.
You get to have a say in which page is treated as canonical. Adding a rel="canonical" tag tells Google which URL you prefer. But that tag is a signal, not a command. Google weighs it alongside other things, like redirects, whether a page is on HTTPS, and whether the URL is included in your sitemap, and it can still choose a different page than the one you marked. If Google picks a page you consider a duplicate instead of the one you meant to rank, that is worth investigating, because it usually means your intended page and the "duplicate" are not different enough where it counts.
Google's Search Console has a specific test for this. Open the URL Inspection tool and compare two fields: Google-selected canonical and user-declared canonical. If they match, Google agrees with your preference. If they do not, Google chose something else, and the guidance is direct about what to do next: make the content of the page you want indexed substantially different from the one Google picked instead.
Duplicate content, near-duplicates, and topic overlap are not the same problem
A lot of confusion here comes from lumping together things that are actually distinct, so it helps to separate them clearly. Most duplicate content seo advice treats these as one problem, and that is exactly where it goes wrong.
Duplicate content is when the same primary content sits at more than one URL. This happens more often through technical accidents than editorial choices: a page reachable through both HTTP and HTTPS, a product page that generates a different URL for every sort order, a session ID tacked onto the end of a link. None of this is a content strategy failure. It is a technical cleanup job, and fixing your canonical signals usually resolves it without touching the writing at all.
Near-duplicate pages are the trickiest category, because they look like two separate articles on the surface. Different titles, maybe a different intro paragraph, a slightly reordered outline. But if the actual substance, the core answer the page gives, is basically the same, Google can still cluster them the way it would an exact duplicate. Changing the headline does not change what the page fundamentally says. This is the situation where most well-meaning teams get caught off guard with near-duplicate pages, because they assume that visibly different wording means the pages are different enough, when what matters is whether the main answer is different.
Topic overlap without duplication is when two pages cover the same subject but answer genuinely different questions. A page explaining what keyword cannibalization is and a separate page walking through how to fix keyword cannibalization on an ecommerce site can share plenty of vocabulary without being anywhere close to duplicates, because a reader looking for a definition and a reader looking for a fix want different things from the page they land on. Shared terminology is not evidence of a duplicate cluster. What matters is whether the primary content, the actual substance of the answer, diverges.
Genuinely thin content is a fourth category entirely, and it is worth naming so you do not conflate it with the other three. A page can be thin, meaning it does not have enough substance to do its job, whether or not anything else on the internet resembles it. A page can also be substantial and still get clustered as a duplicate of something else. Thinness and duplication are different diagnoses that happen to sometimes show up on the same page, and they call for different fixes.
When overlapping pages actually compete with each other
Not every case of overlap means Google has clustered your pages. It helps to separate three situations that get talked about as if they were one.
The first is the clustering scenario already covered: Google sees the same or very similar primary content and picks one canonical. Here an alternate URL typically will not get its own listing in results, and you can confirm this directly through the URL Inspection tool.
The second is genuinely different: both pages remain distinct in Google's eyes, meaning neither is flagged as a duplicate of the other, but they are chasing the same search intent so closely that they end up competing for the same spot in results anyway. This is an editorial and performance problem, not a technical duplicate-content issue, and it will not show up the same way in Search Console. A shared keyword by itself does not prove this is happening. You have to look at what each page is actually trying to answer and how each one performs for the relevant searches.
The third is the one that gets mistaken for a problem when it is not: two pages share a subject but serve different purposes, and the distinction is real and visible in the content itself. This is fine. Sharing a subject is not the same as competing for the same result.
A workable rule for deciding what to do with your own backlog: if a topic genuinely has one reader need with substantially the same answer either way, that points toward maintaining one strong destination rather than publishing a second, thinner variation of the same thing. If two reader needs are genuinely different, separate pages are justified, but only if the main content on each one actually makes that difference real. Two headlines pointing at the same underlying answer do not count as two different needs.
Cases where a blanket rule breaks down
A few situations complicate the picture enough that they deserve their own callout, because applying the general rule too literally here causes real damage.
Regional and translated pages are the clearest example. Google's own guidance lists near-identical same-language regional pages, say a US and a UK version of the same content, as a case worth canonicalizing, usually alongside an hreflang setup that tells Google which version serves which audience. But pages with primary content in genuinely different languages are not duplicates of each other just because they cover the same topic. Translating only your header and footer while leaving the body copy in the original language does not create a real second-language page, and Google will likely still treat that as the same content wearing a different wrapper.
Paginated content is the opposite case: a series of URLs that Google explicitly says should not be collapsed into one canonical. If you run a long list, a directory, or a product category spread across several pages, Google's pagination guidance is specific that each page in that sequence should keep its own canonical URL rather than pointing back to page one. Treating pagination like ordinary duplication and consolidating it away can quietly remove content from the index that should have stayed there.
And then there is the boundary you should never cross while trying to solve this the fast way. Google's spam policies describe doorway pages, meaning pages built to be substantially similar to each other purely to capture more search traffic for city names or minor keyword variants, funneling all that traffic to one real destination. They also describe scaled content abuse, generating a high volume of pages mainly to manipulate rankings rather than to help anyone. Neither of those is the same thing as having two legitimately similar articles on your site by accident. But it is worth knowing where that line sits, because "just publish more variations" is exactly the instinct that walks a site toward it.
A practical check before you rewrite anything
Before spending hours rewriting a page to hit some imagined uniqueness threshold, run through this instead.
Start by naming what each page is actually for. Write down, in one sentence, the specific question each URL answers. If you cannot state a real difference between two pages this way, that is a strong signal you are looking at one topic that has been split into two weaker pages, not two topics that both deserve to exist.
Next, check what Google actually did, rather than guessing. Open the URL Inspection tool in Search Console for the page in question and look at the Page indexing section. If it says "Alternate page with proper canonical tag," that page is pointing to an indexed canonical and this is normal, not an error to fix. If it says "Duplicate without user-selected canonical," Google chose a canonical on its own, which is also expected behavior in most cases, not a red flag. If it says "Duplicate, Google chose different canonical than user," your preference lost, and the fix is making that page's content substantially different, not adding more canonical tags. "Crawled, currently not indexed" is a separate signal entirely and does not by itself mean duplication.
If the pages really are variants of one underlying page, pick the version you want to represent it and align your canonical signals so Google's choice matches yours. If they are meant to be separate but Google has clustered them anyway, look for a technical explanation first, then make the difference in the actual content clear enough that it cannot be mistaken for a rewording.
One more thing worth knowing before you check your work too soon: Google's own troubleshooting guidance says a page can stay in a duplicate cluster for up to two weeks even after you have made real changes, and clearer, more substantial edits tend to resolve faster. That is a possible delay, not a promise, so do not read a lack of change after two days as proof that your fix did not work.
The verdict, and what changes it
The confirmed part of this is Google's clustering behavior itself, and the fact that there is no automatic penalty for ordinary duplication in the way most people fear. Both of those come straight from Google's own documentation, current and past. What remains genuinely unknown is any specific percentage or word count that separates "unique enough" from "too similar," because Google has never published one, and no vendor uniqueness score can substitute for that missing number. If Google ever does publish a concrete threshold, or contradicts this in future documentation, that would change the verdict here. Until then, the useful standard is not a percentage. It is whether each page has a distinct purpose and primary content that actually reflects it.
If you are managing a backlog of dozens or hundreds of already published pages, that standard is a lot easier to apply with a real map of what you have already published and what each piece is actually about, rather than relying on memory or a spreadsheet that goes stale the day after you update it. That kind of visibility is what makes a consolidation decision defensible instead of a guess.
