Google said in June 2019 that you "usually won't see more than two listings from the same site in our top results," which means the third post you wrote on a subject was never a third chance at ranking — it was a spare. On a 30-post blog, keyword cannibalization rarely shows up as a penalty. It shows up as internal links that no longer point anywhere specific, an AI answer engine picking whichever of your three URLs it likes, and roughly three and a half hours per redundant post that could have gone into updating the original. The fix that actually pays is a topic record you check before writing, not a consolidation audit after the damage.
Nobody publishes a duplicate on purpose. The pattern is always the same: you write a good post on a subject in month three. In month eleven you have a slightly better idea about it and write that up, having half-forgotten the first. In month twenty-five a keyword tool suggests a phrase that is, functionally, the same topic in different words.
Now three URLs on a 30-post archive answer more or less the same question, and none of them ranks. Every SEO article tells you Google is "confused," so you go looking for a consolidation audit — a process that costs more than the posts did. Before doing that, it's worth separating the part of the standard story that holds up from the part that doesn't. A lot of it doesn't.
The classic framing — three pages on one topic leave search engines unable to tell which to rank — is contested by people with more data than either of us.
Ahrefs studied multiple rankings on their own site: 9,700 instances where more than one ahrefs.com URL ranked for the same keyword. They hand-reviewed a sample of 80. Exactly one needed action. Their conclusion: "classic SEO theory is wrong—you don't need to go and fix each keyword with multiple rankings." The same study found multiple rankings skew toward harder keywords — median Keyword Difficulty of 64 versus 37 for single rankings, at near-identical search volume. That looks more like strength than self-sabotage.
Google's John Mueller made the same point in September 2025: "If you have 3 different pages appearing in the same search result, that doesn't seem problematic to me just because it's 'more than 1'. … pages aren't duplicates just because they happen to appear in the same search results page." He also questioned whether hunting for cannibalization is a good use of time — a fair objection when your content team is one person.
Ahrefs' glossary agrees that when "the two 'competing' pages are fulfilling a different search intent, it probably isn't" a problem, citing their own #1 and #2 for "YouTube search terms" with a blog post and a tool page. Their longer guide to keyword cannibalization quotes Patrick Stox calling the search-engines-get-confused theory "preposterous," since Google understands what is on each individual page.
So: multiple rankings are usually fine. Near-duplicate posts are the narrower, real problem. Microsoft put the distinction cleanly in a Bing Webmaster Blog post from December 19, 2025: "Duplicate content doesn't trigger search penalties on its own, but it does reduce visibility by diluting authority, confusing intent, and slowing how updates reach both search engines and AI-powered discovery systems."
Not a penalty. A dilution. Here is what dilutes, specifically, on a blog your size.
Google's June 6, 2019 site diversity change is the clearest structural statement on this: "This site diversity change means that you usually won't see more than two listings from the same site in our top results. However, we may still show more than two in cases where our systems determine it's especially relevant to do so."
Two caveats: the announcement is from 2019, and Google itself hedges with "usually." Treat it as stated behaviour, not a hard current rule. But even hedged, it kills the intuition behind the third post — you wrote it hoping for a slot on a page where you were already at your ceiling.
The slots are also worth less than older advice assumes. Search Engine Land's keyword cannibalization guide puts position #1 at roughly 27.6% CTR and position #10 at 2.4%. And Ahrefs' December 2025 study of 300,000 keywords found the top-ranking page takes a 58% lower CTR when an AI Overview sits above it, tapering to −19.4% at #10; the same study a year earlier measured 34.5%. Three posts divide an already-shrinking pool.
This is the cost that hits a small blog hardest, and almost nobody names it.
Google's documentation on crawlable links and anchor text says "Every page you care about should have a link from at least one other page on your site," and gives a test for anchor text: "Try reading only the anchor text (out of context) and check if it's specific enough to make sense by itself."
Now try that test with three overlapping posts. What anchor text is specific enough to identify one of them? There isn't one — that's the definition of overlap. So you link to whichever version you happened to remember, your internal links scatter across three URLs, and none accumulates the signal Mueller calls "super critical for SEO. It's one of the biggest things you can do on a website to guide Google and visitors to the pages that you think are important."
To be precise: repeating an anchor is not the problem. Mueller said in April 2025 that "Having 4 identical links on a page to another page seems fine & common to me." The problem is ambiguity about which URL the anchor should resolve to.
Note also what is not being split. The "you're splitting your link equity" line is big-site framing. Ahrefs' study of roughly 14 billion pages found 96.55% of pages get no traffic from Google at all, and that of about 20 million pages with zero referring domains, only 2,997 ever cleared 1,000 monthly visits — about 1 in 6,671. If nothing external links to your three posts, there's no external equity to split. The split that's actually happening is internal.
This is the genuinely new cost, and most cannibalization advice predates it. From that December 2025 Bing post, written by Principal Product Managers Fabrice Canel and Krishna Madhavan: "LLMs group near-duplicate URLs into a single cluster and then choose one page to represent the set." Microsoft warns that "the model may select a version that is outdated or not the one you intended to highlight."
Read that against your three posts. On an AI answer surface they are not three candidates — they are one cluster, from which a model picks a representative. If your best thinking lives in the 2026 rewrite but the 2024 original has the shape a retriever likes, the model may quote the old one, and you don't get a vote. The upside, as Search Engine Journal summarised it: "When you reduce overlapping pages and allow one authoritative version to carry your signals, search engines can more confidently understand your intent."
Orbit Media's annual blogging survey — 808 content marketers surveyed in August 2025 — puts the average blog post at 1,333 words and 3 hours 25 minutes of production time, with roughly half of respondents publishing 2–4 posts a month. (The dataset skews B2B and US, so treat it as a benchmark, not a law.)
That's the honest price tag on a redundant post: about three and a half hours, near seven for two of them. Spend it updating the original and you're betting with the odds. Ahrefs' SEO statistics roundup reports that 72.9% of pages ranking in Google's top 10 are more than three years old, and that the average top-ranking page also ranks top-10 for nearly 1,000 other keywords. Your first post, if it's any good, is already harvesting the long tail your third was written to capture.
You don't need an audit process. You need four checks, cheapest first. The first three come from Search Engine Land's practitioner guide to detecting and fixing cannibalization.
site:yourdomain.com "target keyword" in Google. Lists the indexed pages on your site containing the phrase. On a 30-post blog this alone usually finds it.&filter=0 to a Google search URL. This disables filtering and domain clustering, so you can see which of your pages actually compete in the SERP rather than which one Google chose to show you.dimensions: ["query","page"] with a rowLimit of up to 25,000 (default 1,000). Group the rows by query, count distinct pages, flag every query with two or more of your URLs. It's a ten-minute script you can rerun quarterly.If you already pay for it, Semrush's Cannibalization report inside Position Tracking flags keywords where multiple pages of yours rank in Google's top 100. The warning sign between checks: rankings drifting down despite publishing more.
Finding overlap is not the same as needing to act. Remember: 79 of Ahrefs' 80 sampled cases needed nothing. Run each collision through four options.
Leave it alone when the pages genuinely serve different intent. If one post is a tutorial and the other a comparison, and both earn clicks, you're diversifying. Ahrefs' data included keywords where a second ranking added anywhere from 1.94% to over 9,000% more traffic.
301 redirect when one page clearly outperforms the other and the weaker one holds nothing worth keeping.
Merge, then 301 when both perform similarly but each holds a different angle — the common case for a blogger's three posts. Ahrefs did exactly this with two of their own competing guides on broken link building; the consolidated page's traffic surpassed the combined traffic of the two it replaced. They also cite Moz having three pages competing for "keyword cannibalization," the best of which reached only position #6.
Differentiate when both topics deserve to exist but the targeting has collapsed. Search Engine Land's example: two pages aimed at "small business accounting software," reworked so one targets "best accounting software for small business startups" and the other "cloud-based accounting software for small businesses." Verify the long-tail variants have real volume first.
The published evidence for consolidation is good. Chris Long of Nectiv documented duplicate-content consolidation cases in Search Engine Land: two identical franchise-listing pages where a 301 produced "organic traffic improved by over 200 percent to the page"; and a "Food Franchises" / "Fast Food Franchises" pair — the closest published analogue to a blogger's near-duplicates — where a canonical on the narrower page lifted organic traffic 47 percent. Sara Taher's case on SEO Riddler reports that after merging and redirecting the weaker URLs, "the performance of one URL is now even better than all of the previously cannibalizing URLs combined" — she publishes a graph, not before/after numbers, so take the claim rather than a figure.
Worth memorising: the fixes Ahrefs explicitly does not recommend — deleting pages, noindex, canonicalization (except for genuine exact duplicates), and de-optimising a page. Their sequence is merge into one guide, publish at the surviving URL, 301 the old page, then update internal links to the new URL instead of leaning on the redirect.
This is the part that scares people off, and it's mostly mechanical.
Use a permanent redirect. Google's documentation on redirects states that "The 301 and 308 status codes mean that a page has permanently moved to a new location," and that "the indexing pipeline uses the redirect as a signal that the redirect target should be canonical." Google advises using permanent redirects only when you're sure you won't revert them.
On Netlify: a line in _redirects is just /old-path /new-path 301, and 301 is the default status if you omit it; the equivalent in netlify.toml is a [[redirects]] table with from, to and status = 301. Per Netlify's redirects documentation, paths are case-sensitive, _redirects rules process before netlify.toml rules, and the first matching rule wins.
On Vercel: redirects live in vercel.json and run at the edge. Vercel's docs recommend 307/308 over 302/301 "to avoid the ambiguity of non GET methods," and state both "do not affect SEO." Vercel does not lowercase paths: /About and /about are different URLs, which is a quiet way to end up with a second copy of a post.
Then fix the internal links. Ahrefs calls this step out specifically: redirecting while leaving internal links pointed at the dead URL wastes the consolidation you just did. On a static site it's a find-and-replace across your markdown, and it belongs in the same commit as the redirect rule.
That coupling is the argument for treating posts as code. If your blog publishes through a GitHub pull request — the publishing path Magic Share uses, where merging the PR is the final publish step — the merged post, the redirect rule and every updated internal link land in one reviewable diff.
Everything above is remediation. It costs you the original three and a half hours, plus the merge, plus the redirect, plus the link cleanup. Prevention costs a lookup.
Yoast's guidance on keyword cannibalization is explicitly pre-publication: keep a keyword and topic map, and "assign a unique target keyword to each page." Search Engine Land's guide lands in the same place. Combine that with Google's anchor-text test and you get a record with seven columns, one row per published or planned post:
| Column | Why it's there |
|---|---|
| slug | The canonical URL; also what you redirect to if you ever merge |
| primary keyword | The uniqueness constraint — one keyword, one row |
| search intent | informational / commercial / transactional; different intent is the licence to keep both |
| the one question this post answers | A single sentence. This is the column that actually catches collisions |
| anchor text you'd use to link to it | Google's test: specific enough to make sense by itself |
| publish date | Context for what's stale |
| last-updated date | The prompt to refresh instead of rewrite |
The pre-write check is two lookups, and it takes under a minute:
When it collides, the output is an update, not a post. That is the whole discipline — and given 72.9% of top-10 pages are over three years old, strengthening a page that already has age and internal links usually beats starting a new one at zero.
The failure mode isn't the format, it's that nobody opens the file at the moment of deciding what to write. The record only works if something checks it before the draft exists.
This problem gets worse as writing gets cheaper.
Ahrefs' analysis of 900,000 newly created English pages in April 2025 found 74.2% contained AI-generated content — 2.5% pure AI, 25.8% pure human, 71.7% a mix. Graphite, sampling 43,000 CommonCrawl URLs, reported that AI-written articles overtook human-written ones in November 2024, reaching 51.7% of sampled articles by May 2025, and notes those articles "largely do not appear in Google and ChatGPT."
The defensible reading isn't "AI content doesn't work." It's that undifferentiated bulk doesn't work — an argument for a topic record, not against automation. Google's helpful-content self-assessment asks, under "Avoid creating search engine-first content," whether you are "producing lots of content on many different topics in hopes that some of it might perform well in search results?" And Google's spam policies list "Substantially similar pages that are closer to search results than a clearly defined, browseable hierarchy" as an example of doorway abuse.
To be clear: three honestly-written overlapping posts on a 30-post blog are not doorway abuse or scaled content abuse. Nobody is at risk of enforcement for forgetting what they wrote in 2024. But that language tells you how Google describes a stack of near-variant pages — not a description you want your archive to grow into. We covered it in our close reading of Google's scaled content abuse rules: method is never the offence; volume-without-value is.
This is why a topic ledger sits at the centre of how Magic Share works. The agent keeps a record of everything already written or planned for a site and checks it before drafting, so a run that would produce your fourth post on one subject produces something else instead. Schema checks then validate a target keyword, slug, meta title and description before a draft is saved — turning "assign a unique target keyword to each page" from advice everyone repeats into a constraint enforced at write time.
Run it yourself in a spreadsheet if you prefer. The point isn't the tool; it's that the check happens before the three and a half hours, not after.
No. Microsoft stated it directly in December 2025: "Duplicate content doesn't trigger search penalties on its own, but it does reduce visibility by diluting authority, confusing intent, and slowing how updates reach both search engines and AI-powered discovery systems." Google's John Mueller has similarly said multiple pages in one result set aren't inherently a problem. The cost is dilution and lost time.
Filter Search Console's Performance report to the keyword, then switch to the Pages tab — if two URLs earn impressions for that query, they're candidates. Confirm with site:yourdomain.com "keyword", and append &filter=0 to a Google search URL to disable domain clustering and see which of your pages really compete.
Ahrefs explicitly doesn't recommend deleting, noindex, or de-optimising. The standard sequence is to merge the useful parts into one page, publish at the surviving URL, 301 the old URL to it, then update internal links to point at the new URL rather than relying on the redirect.
It matters differently. Microsoft's Bing team says "LLMs group near-duplicate URLs into a single cluster and then choose one page to represent the set," and warns the model "may select a version that is outdated or not the one you intended to highlight." So near-duplicates on an AI surface aren't extra chances — they're one entry with a representative chosen for you.
Size isn't the variable; overlap is. But on a small archive the mechanics change: with few or no external backlinks — Ahrefs found only about 1 in 6,671 pages with zero referring domains clears 1,000 monthly visits — there's little external equity to split. What splits is your internal linking and your attention.
Multiple rankings are usually fine. Near-duplicate posts are the narrower problem: a ceiling you were already at, internal links with nothing specific to aim at, an AI answer engine choosing your representative page, and about three and a half hours per redundant post. Audit what you have — most of what you find will need nothing. Then write the record down so it doesn't happen again.
If you'd rather not be the one remembering, Magic Share drafts against a topic ledger of everything your site has published or planned, verifies every cited link before saving, and holds each post as a draft until you approve it. See how the drafting and verification pipeline fits together, browse the rest of Notes from the garden, or start free on the pricing page — your first three posts are free.
Magic Share researches, writes and fact-checks posts like this for any site — point it at your URL and review your first draft today.
Plant your first post