The addresses stayed the same. The new website went live on a Thursday. By the following Monday, a few of the pages that used to sit on the first page had slipped, and someone in the room said the migration must have gone well because the URLs had not changed. I hear that sentence a lot, and I no longer treat it as a conclusion.
On 9 November 2026, Search Engine Land published an article on website migrations that makes the same point from the other direction: a migration can still hurt the search results when the URLs themselves do not change. Rendering, canonical tags, internal links and HTTP responses can all change in the new build, and those are the things Google uses to crawl a page and to choose it. The address in the bar is only one part of that job. Google's own documentation had already been saying a version of this. I would start there, not with a recovery date, because I do not have one to give you.
Why isn't the address the whole story?
Google's guide to moving a site with URL changes is written for a different job, a change of domain or of path, but one sentence in it applies to any large rebuild. "With any significant change to a site, you may experience ranking fluctuations while Google recrawls and reindexes your site." A new template, a new CMS, a move off WordPress onto a custom React site, all count as a significant change even when every public address stays put. The same guide also says to change only one thing at a time: plan the changes one after the other, and if you want a new domain, a new CMS and a new layout, do each one separately. If you change the platform and the look in one night, and the rankings move, you have mixed two causes and you will spend the next month arguing about which of them did it.
The address is the promise you make to a bookmark and to a backlink. It is not a complete description of the page. Google still has to fetch it, decide which copy is the official one, follow the links, and read the content in the first HTML response or after rendering the JavaScript. Those four jobs can be done differently by a new site at the old address, and a drop after a launch is one of those jobs, not a mysterious penalty.
I wrote about why a site disappears from Google when the problem is a missing page, a noindex or a crawl block. A redesign drop is quieter than that. The pages are still in the index. They just do a slightly different job when Google next looks at them.
If you have also been watching the November 2026 core update, do not fold the two together. The update started on 10 November and may take up to two weeks. A launch in the same window needs its own before-and-after snapshot, because otherwise you will blame the rebuild for a movement that started on every other site, or blame the update for a noindex you shipped on Friday. I wrote about what to do during that core update separately. This article is about the launch.
Four things that quietly change
The first is rendering. A WordPress theme puts the article, the heading and the menu in the first HTML file. A new front end, especially a custom React site, can send an almost empty shell and fill the page in the browser. Google's JavaScript SEO basics say Search may queue a page for rendering, that this "typically takes a few seconds, but it can take longer," and that not every bot can run JavaScript at all. That "typically" is Google's word, not a clock I would bet a launch on. Google recommends server-side rendering or pre-rendering so the first response already contains the content. If the new site hides the text until the script runs, some crawlers will see a blank page at an address that used to contain the article. That is not a URL change. It is a change in what the URL returns. Open the source.
The second is the canonical tag. On many WordPress sites a plugin writes it. In a rebuild, that job moves into the templates, or it is forgotten, or it still names the test host, or it disagrees with the www and non-www versions. Search Engine Land's 9 November piece notes that canonicals can disappear in a migration even when the public URLs stay the same. I have seen that on staging copies that went live with the staging address still in the tag. Google then has to guess which copy to keep. Guessing is not a plan.
The third is internal links. A new menu can look perfect to a person and still be uncrawlable if the links exist only after a script runs, or if they are buttons that never become ordinary hrefs. Important service pages can also drop out of the footer and the sitemap in the same week, because someone tidied the new information architecture and forgot that those pages were how Google found the rest of the site. The page is still at the old address. The paths to it have changed.
The fourth is the status code. A missing URL that used to return 404 can, on a new single-page app, return 200 and a "page not found" screen. Google then has to infer that the page is gone, and it is slower and less certain than a real 404. Look at the status. The opposite happens too: a test-site noindex or a robots.txt Disallow, copied into production, tells Google not to index pages that are live and returning 200. I would rather catch both of those on the test host than explain them after launch. Catch it there.
None of these four needs a new domain to go wrong. They go wrong at the same addresses.
Check the test site before you launch
Google's guide to changing your web hosting is written for a hosting or CDN switch that leaves the public URLs alone, not for a redesign. The test advice still holds when the new website sits on a temporary hostname before launch. Google says to allow public testing on a temporary host, to add a noindex rule so the test copy is not accidentally indexed, and to remove any robots or noindex blocks from the new copy when you are ready to switch. That is the moment to do the boring work, not the Friday after launch.
On the test host I look at the raw HTML of the home page, the main service pages and a few articles, not only at the pretty preview in the browser. I want the heading, the main text and the main menu in that first file, not only after the script has run. View source on a phone-sized window as well as on a desktop, because a header that collapses into a script-only menu can hide every service link in the first response. I check the canonical on each of those pages, and I check that it names the future public address, not the test host, and not a leftover www or non-www variant you are about to redirect. If the site is in more than one language, I check that the language tags still point at the matching public URLs, not at the temporary host.
Then I walk the links. I click through the menu and the footer with JavaScript off, or I use a crawler that does not run scripts, and I write down any important URL that has disappeared from the links. Write the misses down. I compare that list with the sitemap on the test host, and with the list of indexed URLs I export from Search Console before the switch, so the baseline is a file, not a memory. I request a handful of old addresses that should now be gone, and I want a real 404 or a real redirect, not a 200 with an error template. I fetch robots.txt on the test host and I write down every Disallow I will need to remove, or to keep, on the morning of the switch. A block that was sensible on staging is a silent drop in production if it travels with the files.
Do this on the test copy.
Do not discover it on Monday.
If you are leaving WordPress for a custom React site, the jobs that a plugin used to do, the canonicals, the sitemap, sometimes the status codes, have to be built into the new templates. We check those before the switch, as part of the web design work, not as a surprise in week one. We keep the public addresses. We only lift the noindex when the four checks pass. An existing WordPress site can be redesigned in place when the structure already holds. The platform is not the first question. The four checks are. That is how we protect the rankings during a redesign: not by promising they will look the same on day three, but by refusing to treat "same URLs" as a plan, and by doing this work on the test host, in front of the switch, while there is still time to fix a tag or a missing link without an inquest. If you already have a new site in staging, send the test address to contact@magnuszgr.com and we will tell you what we would still check before it goes live.
What should you watch in the first two weeks?
This window is my practice, not a rule Google publishes for a same-URL rebuild. Google does say that ranking can fluctuate while it recrawls a changed site, and in the URL-change guide it talks about a few weeks or more for the new addresses to show instead of the old ones. I would not treat that second figure as a promise that a same-URL drop will recover on a particular day. I do not know that, and I will not pretend to.
What I do in the first two weeks is narrower. I watch the Page indexing report for a rise in "Crawled - currently not indexed", for noindex, for not found, and for duplicate pages without a user-selected canonical. I look at crawl stats to see whether Google is actually fetching the new HTML. I inspect a sample of the URLs that used to bring in the most clicks, and I compare the first HTML with what I signed off on the test host. I also keep the Performance export I took before launch and I wait for a full week of data after the switch before I call a movement a trend.
Day three is noise.
If something in those reports points at a fault, a canonical still naming the test host, a menu missing in the HTML, a 200 on a deleted URL, I fix that fault. I do not rewrite the articles to "recover." Rewriting mixes a third change into a launch you are still trying to read.
SEO work after a redesign is mostly this: reading those reports, fixing the four jobs if they slipped, and leaving the healthy pages alone until the recrawl has had a chance to finish.
When is a drop normal, and when is it a fault?
A short wobble in positions, with the pages still indexed, the HTML still complete, and the four checks clean, can be the fluctuation Google already warned about. I would wait, keep the snapshot, and look again after a week, then after two. I cannot tell you, from outside your Search Console, whether a given drop is that wobble or a fault you shipped at launch. The reports are what settle it. Guessing from a ranking tracker alone is how a team "fixes" a healthy page and leaves a staging canonical in the header.
A drop is a fault when the test-site noindex is still in production, when the content is missing from the first HTML, when important pages have no incoming links in the crawl, or when deleted addresses now return 200. Those are fixable, and they are fixable without changing the public URLs.
If the four checks were clean at launch, I would wait, keep the snapshot, and look again after a week, then after two, rather than rewrite the pages to chase a Tuesday position. The offer I can make is the work before the switch, and the reading of the reports afterwards, not a date on which the numbers will come back.

