A migration loses rankings when the signals search engines rely on change faster than they can process them: URLs, redirects, internal links, structured data. None of that has to happen. Here’s what we actually check.
Before touching anything
- A full crawl of the current site, capturing every indexable URL, not just the ones in the main navigation. Blog posts, old landing pages, and PDFs get missed constantly, and each one needs a home in the new structure.
- Traffic and backlink data pulled per page, so the pages actually earning rankings and links get identified before anyone decides what to keep or cut.
- Two to four weeks of baseline analytics, organic traffic, keyword positions, and crawl errors, recorded before launch. Without this, there’s no way to tell a real post-launch problem from normal noise.
The redirect map is the actual deliverable
Every old URL gets mapped to its single closest equivalent on the new site, built as a real document before development starts, not improvised at launch.
- One old URL, one specific new URL. Redirecting everything to the homepage is the most common mistake we see, and it tells search engines nothing about where the content actually went.
- A permanent (301) redirect, not a temporary one. A temporary redirect signals the old URL might come back, which is the wrong signal for a page that’s genuinely gone for good.
- No redirect chains. Each old URL should point directly to its final destination in one hop. A chain of three redirects is three separate points where something can break.
- Redirects stay live for the long term, a year at minimum. Old backlinks and bookmarks don’t disappear on a schedule, and removing a redirect early breaks every one of them at once.
Staging and launch
The staging site needs to be genuinely invisible to search engines, password-protected or blocked with a noindex directive, since an indexed staging copy creates duplicate content competing with the real site.
At launch, we confirm four things before calling it done: every redirect actually resolves to the right page, canonical tags point to the new URLs, the XML sitemap reflects the new structure with nothing that no longer exists, and analytics tracking is firing correctly. If the domain itself changed, both the old and new properties get verified in Search Console, so crawl and indexing data stays visible across the transition.
The weeks after launch matter more than launch day
Search engines take time to process redirects and transfer ranking signals, and that window is when real problems are still fixable cheaply.
- Daily checks for the first few weeks: organic traffic, indexed page count, crawl errors, and keyword positions for the pages that mattered most before the move.
- Temporary fluctuation is normal. A gradual dip that stabilizes within a few weeks isn’t a problem. A sustained drop, or a sharp one right after launch, is worth investigating immediately rather than waiting to see if it resolves on its own.
- A second full crawl of the live site, checking for 404s, broken redirects, and any page that didn’t carry over its title tag or meta description in the move. Comparing this crawl against the pre-launch one directly surfaces the specific pages that changed status or lost a redirect, rather than someone scanning the results by eye.
Where AI actually helps
We use AI (Claude Code) directly in this workflow, in two specific places.
- Drafting the redirect map. For a large site, the redirect map itself is often the most time-consuming part of the whole migration, matching a few dozen pages by hand is straightforward, matching a few thousand isn’t. AI generates a first draft by matching each old URL’s content and structure against the closest page on the new site, which we then review, rather than building the mapping from a blank spreadsheet.
- Pre-launch content diffing. AI compares each old page’s title tag, meta description, and body content against its new counterpart to flag ones that got dropped in the rebuild, the exact mistake listed below as one of the most common.
None of this replaces the judgment migrations still need: someone still has to decide whether a page is worth keeping, which new page is the true closest equivalent, and whether a ranking drop is real or normal fluctuation. AI turns the manual-matching work into a review job instead of a build-from-scratch job, it doesn’t remove the decision.
What actually goes wrong
The same handful of mistakes account for most of the damage we’ve seen on migrations we didn’t build from scratch:
- Skipping the initial audit, so pages with real traffic or backlinks get dropped without anyone noticing until rankings fall.
- On-page elements like title tags and meta descriptions lost in the move, even when the redirects themselves are correct.
- Staging left indexable, creating duplicate content that confuses which version should rank.
- Monitoring stopped after the first few days, when ranking signals are still being processed weeks later.
If you’re planning a migration and want a second opinion on the redirect map before it goes live, that’s a conversation we’re happy to have.