Before a CMS migration, the two things most likely to quietly cost you traffic are orphaned assets (files nothing on the site links to anymore) and duplicate content (multiple URLs serving the same words). Neither shows up on a homepage. Neither triggers an alarm. Both survive the move by accident, dilute your crawl budget, split your ranking signals, and confuse the AI engines trying to figure out which version of a page is the real one. You find them by inventorying every URL and every file before cutover, not after the rankings dip.
Right, now let me explain why I opened with that.
Moving day, and the boxes nobody remembers packing
Picture your last house move. You thought you had, what, forty boxes? You end up with sixty-three. There's a box of cables for devices you no longer own. There's the second kettle. There's a mystery box labelled "KITCHEN?" in someone else's handwriting. You didn't decide to bring any of that. It just came along because moving day is chaos and nobody audits a box at 11pm.
A content migration is exactly this, except the boxes are URLs and nobody's carrying them up the stairs so nobody notices the weight. The mystery boxes are your orphaned assets. The second kettle is your duplicate content. And unlike a real move, the junk doesn't sit harmlessly in your new garage. It actively competes with your good stuff for Google's attention.
The attic: orphaned assets
An orphaned asset is a file that still exists at a URL but has no internal link pointing to it. The classic offenders are PDFs. Someone uploaded the 2019 pricing sheet, linked it from a landing page, then rebuilt the landing page two years later and forgot the PDF entirely. The file's still there. Google indexed it back in 2019. It still shows up in search for someone, still gets crawled, still gets served, and now quietly contradicts your current pricing to anyone who finds it.
Multiply that by every whitepaper, every old logo variant, every event flyer, every "final_FINAL_v3.docx" that got dropped into the media library. On a mature site that number is rarely dozens. It's usually hundreds, sometimes thousands.
Here's the trap at migration time. Most migration scripts move content types: pages, posts, products. Loose files in the old media library either get skipped (so every indexed PDF becomes a 404 the day you launch) or get bulk-copied to a new path (so every one of them silently changes URL and loses whatever links pointed at it). Either way, the attic follows you to the new house, just reorganised badly.
The fix isn't glamorous. Crawl the site, list every asset, and cross-reference against every internal link and your server logs. Anything with no inbound links and no traffic in twelve months is a kill candidate. Anything orphaned but still pulling search traffic needs a redirect, not a delete. That distinction is the whole game, and you can only make it if you inventoried the file in the first place.
The photocopier: duplicate content
Duplicate content is the second kettle. Same page, more than one address. It sneaks in through more doors than you'd think: http and https both resolving, www and non-www, trailing slashes, uppercase paths, tracking parameters like ?utm_source creating a fresh URL for every campaign, printer-friendly versions, tag and category archives that reprint the same posts, and pagination that echoes your first page.
Individually, harmless. In aggregate, you've handed a search engine four versions of your best article and asked it to pick a favourite. It picks. Sometimes it picks the printer-friendly one. Your link equity gets split across the copies, and the AI answer engines, which are even less forgiving about ambiguity, may cite the wrong URL or none at all.
Migration makes this worse because a new platform generates its own new duplicates while you're still carrying the old ones. Now you've got the second kettle and a third kettle you bought on the way to the new place.
What to actually do with each
The decision is different for the two problems, so don't treat them as one cleanup task.
| Problem | How you find it | Migrate | Rewrite / fix | Kill |
|---|---|---|---|---|
| Orphaned asset | Crawl + link graph + server logs | Still gets search traffic or backlinks | Content outdated but URL valuable | Zero links, zero traffic, superseded |
| Duplicate page | Content hashing + parameter audit | Pick one canonical version | Consolidate near-duplicates into one | Thin auto-generated copies |
For duplicates that are genuinely the same, choose the canonical, point the rest at it with 301s, and set the canonical tag to match. Don't rely on the tag alone. A canonical tag is a suggestion; a redirect is an instruction. For near-duplicates (three thin pages that should be one strong one) consolidate before you move, so you migrate the good version once instead of the mediocre version thrice.
The reason to do all of this before cutover is boring but decisive: after launch, everything looks like migration breakage. A traffic dip could be a bad redirect, a lost duplicate, an orphaned PDF gone 404, or a change in intent. You won't be able to tell them apart. A clean pre-migration inventory (this is the boring, unglamorous work that a pre-migration crawl exists to automate) turns "why did traffic drop" into a spreadsheet you can actually read.
What I'd actually do
Kill more than feels comfortable. Every team I've watched migrate has agonised over deleting a 2018 PDF that gets nine visits a month, then shrugged and carried five hundred of them to the new platform "just in case." That's exactly backwards. The nine-visit PDF with a backlink from a university, you keep and redirect. The five hundred nobody's opened, you kill without ceremony, because each one is a box you'll have to carry again at the next migration.
Duplicates, consolidate ruthlessly and canonicalise the survivors. Orphans, redirect the ones with equity and delete the rest. Do it before you move, not after, because the version of you standing in the new house at 11pm surrounded by mystery boxes is not the version that makes good decisions.
Nobody ever regretted arriving at the new place with fewer boxes.