SearchPilot test: canonical alone matched a redirect for duplicates

A common piece of technical SEO advice holds that a 301 redirect is the strongest way to consolidate duplicate URLs, stronger than a canonical tag, which Google can choose to ignore. SearchPilot, a controlled SEO A/B-testing company, ran an actual test of that belief on a site that already had canonical tags in place, and the result was not the clean win the advice would predict. SearchPilot’s own published case study, not our analysis, is the source for what follows.

What SearchPilot tested#

The client’s trailing-slash URLs already carried a canonical tag pointing at the non-trailing-slash equivalent before the test began. According to SearchPilot, those pages were “getting a small amount of traffic indicating that some of the pages were still indexed and ranking,” despite the canonical already being in place. That starting condition is what made the test worth running: if canonical tags fully consolidated signal on their own, no residual traffic should have existed on the duplicate URLs at all.

SearchPilot’s variant added a 301 redirect from the trailing-slash URLs to their non-trailing-slash canonical target, on top of the existing tag, and compared the outcome against a forecast built from the site’s control group. The mechanism under test was narrow and isolated: one additional signal, a redirect, layered onto a page set that was already canonicalized, which is a cleaner test of the redirect’s marginal effect than comparing a redirected page against a page with no duplicate-handling at all.

What the result actually showed#

Two things happened, and SearchPilot reports them separately rather than blending them into one number. First, the redirected trailing-slash URLs lost organic sessions, which SearchPilot describes as “an expected decrease,” since a redirected URL no longer serves its own page at all and any of its remaining rankings or click-through had nowhere left to land. Second, the canonical target pages, the ones meant to benefit from the newly consolidated signal, saw only a change SearchPilot calls “marginally positive,” and its own wording states plainly that “the result remained inconclusive at a statistically significant percentage.”

That is a materially different outcome than the 22% traffic uplift SearchPilot reports in its separate test of canonicalizing to product variation pages. In that case, a clear, though still confidence-interval-free, positive number is reported. Here, the honest reading of SearchPilot’s own words is that the redirect did not produce a result the company itself is willing to call statistically real, on the side of the test that was supposed to show the benefit.

Why the client shipped it anyway#

SearchPilot’s case study notes that the client deployed the redirect site-wide despite the inconclusive result, based on the change matching general best practice and the absence of any observed negative trend overall. That is worth separating clearly from the test’s own finding: shipping a change because it seems safe and conventional is a defensible operational decision, but it is not the same claim as the test having proven a traffic gain. SearchPilot’s case study does not conflate the two, and neither should a summary of it.

Why we care#

The advice that a redirect always beats a canonical tag for consolidating duplicate signal is repeated often enough to sound settled. This one controlled test, run by a company whose business is running controlled tests, found the opposite of a clean confirmation: a canonical tag that was already in place appears to have already done most of the consolidation work, and adding a redirect on top of it cost real traffic on one side without producing a provable gain on the other. SearchPilot does not disclose enough about this test, page count, duration, or a confidence interval, to turn that into a universal rule. What it does support is a narrower, more defensible claim: before assuming a redirect is the stronger fix for a duplicate that already carries a working canonical, it is worth testing that assumption on the actual site in question rather than applying it as received wisdom.