The same page often exists at four or five different addresses. A trailing slash, a tracking parameter, an old URL nobody redirected, a printer version somebody linked once and forgot. To a visitor, it is one page. To Google, ChatGPT, and Perplexity, it can read as four or five separate ones.
That split matters more than it used to. An answer engine choosing what to cite has to pick one version to trust. If the signal is scattered across duplicates, it may pick the thin one, the outdated one, or skip citing the page at all.
How One Page Becomes Four
On a typical Arizona service business site, duplication is rarely intentional.
It accumulates.
A service page exists at the plain URL and again with a UTM parameter from an old ad campaign. A category page returns the same list sorted three ways, each with its own address. The site moved from http to https, or added www, and the old versions were never redirected. They were just left standing.
None of this looks like a problem in a browser. Every version renders fine. The problem is invisible until something reads the site one URL at a time, with no idea that four of them are the same page.
What a Canonical Tag Does
A canonical tag is a line in a page's code that says: this address is the real one, treat the others as copies. It does not remove or hide the duplicate pages. It tells a crawler which one to count.
Most site builders write a self-referencing canonical on every page automatically, pointing a page at itself. That part usually works out of the box. Where it breaks is everywhere the same content exists somewhere else: a printer-friendly page, a syndicated blog post, a filtered version of a product list. Those need a canonical pointing back to the original, and most sites never add one.
This sits in the same technical layer as schema markup and structured data. It is invisible when it works and expensive when it does not, which is why AEO and GEO work spends so much time on plumbing nobody sees.
What a Split Signal Costs
When five URLs carry the same content, none of them accumulates the full weight of that content on its own. Links, mentions, and internal references spread across versions instead of landing on one.
For a visitor, this rarely shows up directly. They find the page, read it, move on. For Google, for ChatGPT, and for Perplexity, the effect is quieter and more expensive. A page that should be the strongest one on the site ends up competing with its own copies for the same spot.
I have seen a business's own page get outranked in an AI Overview by a syndicated copy on a directory site, because the directory copy had one clean URL and the original had several.
How I Check For Duplicates
I start with a full crawl, not a spot check. A handful of pages looking fine tells you nothing about the other four hundred.
Then I sort every URL returning a live status by content similarity. Near-identical pages with different addresses surface fast this way, and so do the parameter versions most site tools bury deep in their reports.
The last step is the boring one.
I open the source of the pages that matter most and confirm the canonical tag actually points where it should. A tag that exists but points at the wrong page is worse than no tag, because it looks handled and is not.
The Mistakes That Make It Worse
Canonical chains are the most common one. Page A canonicalizes to page B, and page B canonicalizes to page C. A crawler is not required to follow the whole chain, so the signal can just stop.
Canonical tags that disagree with the redirect setup are another. If a redirect sends visitors to the new URL but the canonical tag still points at the old one, a crawler gets two different answers about which page is real.
The third is a canonical tag copied across a template without checking it. A page that canonicalizes to a different page's URL because a developer reused a component is a quiet bug, and it can sit there for years, uncorrected.
None of this is exotic. It is the same kind of structural check behind entity consistency across surfaces, because a business that agrees with itself on which page is real is easier for a machine to trust. A site migration is usually where this breaks, since old and new URLs sit side by side for a while by design.
Canonical tags are not the most interesting SEO problem. They are one of the least interesting and one of the most commonly wrong. Fixing them is unglamorous, cheap, and worth doing before anything else on the page gets attention. That check is part of every SEO / AEO / GEO Foundation engagement.