Two auto parts sites, the same articles on both. One lost 97 percent of its organic traffic. The other grew 17 percent over the same period. That is as close to a controlled experiment as this work ever gets, and it answers a question clients ask constantly: when Google finds duplicate content across two sites, what decides which one survives? The answer had nothing to do with which site published first.
The two sites
A US auto parts retailer came to me in April 2024 with traffic that had fallen off a cliff. Its sister site, serving the Canadian market, was running the same articles and doing fine. Both were reviewed on the same day with the same tool, which makes the comparison unusually clean.
| Measurement | US site | Canadian sister site |
|---|---|---|
| Organic traffic per month | 6, down 97 percent | 1,857, up 17 percent |
| Ranking keywords | 4 | 4,000 |
| Authority score | 9 | 23 |
| Backlinks | 1,600 | 17,400 |
| Referring domains | 541 | 1,500 |
| Traffic trend | Peak around 1,200, then collapse to near zero | Steady growth through the same period |
Six visits a month from four ranking keywords is not a demotion. It is a site that has effectively left the index as a commercial proposition.
What the audit found
- The articles on the two sites were substantially the same. Why: Not similar in topic. The same posts, appearing on both domains. Google had two candidates for the same content and no reason to rank both.
- Publishing was high volume with no editorial control. Why: A large archive built quickly, which is the pattern the scaled content abuse policy introduced in March 2024 was written to catch.
- No meaningful on-page optimisation on either site. Why: Titles, headings and internal structure had not been worked on. Neither site was competing on merit, which left the duplicate question to be settled on other signals.
- Technical issues across the affected templates. Why: Not the cause of the collapse, but they capped what the site could recover to even if the content problem were solved.
- No brand signals at all. Why: Nothing on the site or off it identified who ran it or why a reader should choose it over the identical pages elsewhere.
Why Google picked the other site
This is the part worth dwelling on, because the usual assumption is wrong. People expect Google to keep the original and discard the copy. That is not what the systems optimise for. When the same content exists in two places, Google decides which version to show, and it decides on the strength of everything around the content rather than the content itself.
Look at the two rows that matter. The Canadian site had 1,500 referring domains against 541, and an authority score of 23 against 9. It was, by a wide margin, the better-supported domain. Given a choice between two identical pages, the systems kept the one with more independent signals behind it and stopped surfacing the other.
The practical lesson for anyone running multiple sites on shared content: the weaker domain is not getting a free ride from the stronger one. It is the one that gets switched off. If you are considering a second site to cover another market, the content has to be genuinely different for that market, or you are choosing which of your own properties to sacrifice.
Where duplicate content ends and policy begins
Two separate things were happening here and they are worth keeping apart, because they carry different consequences.
Duplicate content across two domains is not a penalty. Google has been consistent about this for years: it deduplicates rather than punishes. The losing site simply stops being shown, which feels identical to a penalty from the inside but is a filtering decision rather than an enforcement action.
Bulk publishing with no editorial value added is a different matter, and since March 2024 it has been an explicit spam policy. Scaled content abuse covers producing many pages mainly to rank rather than to help, regardless of who or what wrote them. A site sitting on both sides of that line has a deduplication problem and a policy problem at once, and clearing only the first would not have brought the traffic back. The full policy history is in the Google Panda update and its modern descendants.
The roadmap
- Decide what this site is for, before touching anything. Why: A US site duplicating its Canadian counterpart has no reason to exist as a separate property. Either it serves the US market with US-specific parts, suppliers, shipping and pricing, or it should be consolidated into the domain that is working.
- Rewrite or remove the duplicated archive. Why: Every page that exists on both domains is a page Google has already decided about. Leaving them keeps the site in the losing half of that decision.
- Stop bulk publishing. Why: Cleaning the archive while the process that filled it continues produces the same outcome again within months. This was the item the client had to own.
- Build on-page fundamentals. Why: Titles, headings and internal linking. Basic work, but the site had none of it, so it was competing on nothing.
- Establish brand and authority signals. Why: Real ownership details, accountable contact information, and links earned rather than accumulated. This is what makes a domain worth keeping when a duplicate decision is made.
- Clear the technical findings. Why: They did not cause the drop, but they limit the ceiling on any recovery.
The honest assessment
A site at six visits a month with four ranking keywords is not in a recovery situation in the usual sense. There is almost nothing left to recover to. The realistic options were to rebuild it as a genuinely distinct US property, which is a launch rather than a recovery, or to consolidate into the domain already working and stop paying for two.
I said so plainly, because the alternative was a recovery engagement that would produce a site with slightly better on-page tags and still no reason to exist. When the business question and the SEO question are the same question, answering only the SEO one wastes the client’s money.
What this case teaches
Publishing the same content on two domains is a choice about which one to lose. Google will pick, and it will pick on domain strength rather than on which site published first or which one you care about more.
A 97 percent loss and a 40 percent loss are different diagnostic categories. Partial losses usually mean a quality reassessment you can work back from. Near-total losses point at deduplication or a policy violation, and the first question is whether the site has a reason to exist rather than what to improve.
The comparison site is the most valuable diagnostic you have. Two properties, same content, same owner, opposite outcomes. Everything that differed between them was a ranking signal, and reading that difference took an afternoon rather than a month of theorising. The same logic applies within a single site, which is why the page scope analyzer reports the pages that gained alongside the ones that fell.
Running two sites on the same content?
If one is fading while the other holds, that is a deduplication decision rather than a penalty, and it needs a business answer before an SEO one. Send both domains and within 48 hours I will tell you which one Google is keeping and whether the other is worth saving. Free, no obligation.
Get free recovery auditFigures are from third-party domain analysis of both sites captured on the same date during the April 2024 audit. Neither site is named and identifying details are withheld under the confidentiality terms set out in the editorial policy. Diagnostic findings and recommendations are reproduced from the audit report delivered at the time. No recovery outcome is claimed, because the roadmap required a business decision about whether the second site should continue to exist.