Diagnosing and Recovering From Migration Traffic Drops

Two questions get conflated after a migration, and they need answering in order. Did measurable search performance actually change? And did the migration introduce a defect? Neither a drop nor a recovery should be assumed. Establish the measurement first, then investigate the cause.

Google's wording on what to expect is conditional, and the modality matters: If you change the URLs of existing pages on your site, you may experience ranking fluctuations while Google recrawls and reindexes your site. As a general rule, a medium-sized website can take a few weeks for Google to notice the change; larger sites can take longer. That is a statement that fluctuation is possible during reprocessing - not that most migrations lose traffic, and not that an observed loss will return on its own. The same guidance names the trigger for investigating: If you see a drop after moving and it's not recovering, check the site move troubleshooting section for common mistakes when migrating a site with URL changes.

So the shape of the work is: establish that the drop is real, rule out the causes that have nothing to do with your migration, then work Google's documented failure list before your own hypotheses.

Step 1: Establish That the Drop Is Real

Before diagnosing a drop, confirm you are looking at one.

• Check your own measurement first. A migration frequently breaks analytics, consent handling or conversion tracking, and a reporting failure looks exactly like a traffic collapse in the dashboard everyone watches. Compare Search Console against your analytics platform. If Search Console clicks held while analytics sessions fell, prioritize measurement and post-click delivery checks - analytics implementation, consent handling, landing-page responses, redirects, and how each tool defines a session. Any of those can produce the mismatch, so it narrows the search rather than naming the cause.
• Allow for preliminary data. Search Console's own caveat: The newest data can be preliminary, meaning it's still being collected and might change in the next few hours. Do not open an incident on the most recent day's numbers.
• Check whether it is Google's reporting rather than your site. Google's debugging guidance names a reporting glitch as one possible pattern and points to the Search Console Data Anomalies page - The drop might be related to a change in the data processing or a logging error.
• Separate impressions from clicks. Google's framing: If both impressions and clicks dropped, check the list of the most common reasons that could have caused it. If your impressions remain the same but your clicks drop, you might not be generating the best page title and snippet that you could, and so users don't understand the content of your page, or perhaps other sites had a more appealing rich result. Note the modality - a weaker title or snippet is one possibility Google offers, not the diagnosis. Broadly stable impressions with falling clicks is a CTR-pattern investigation: compare position, queries, pages, devices, countries and search appearance across equivalent periods; look at what changed in your titles and descriptions and at how the live result actually renders; and check whether competing results or SERP features became more attractive. A migration that rewrote metadata is a strong candidate here, but so is a position change inside rounded reporting, or a shift in query mix.

Pass rule for this step: you can state, from two independent sources, the size of the drop, the date it began, and whether impressions, clicks or both are affected.

Step 2: Rule Out the Causes That Are Not Your Migration

The migration is the most recent change, so it attracts the blame. Google's debugging guidance covers several categories of cause, and a site move is only one of them. The rest, each with the check that either excludes it or promotes it to its own investigation:

Pass rule: each plausible non-migration category is either excluded with a named check, or promoted to its own investigation. "It's probably the migration" is not an exclusion - and neither is "it lines up with an update."

Step 3: Establish the Shape of the Drop

Three questions, answered before any fix is attempted. Each one orders your checks - none of them proves a cause on its own.

1. When did it start, relative to cutover? Use timing to prioritize, not to conclude. A decline beginning on cutover day makes availability, redirects, robots and indexing directives, routing and tracking the high-priority checks. A decline appearing two or three weeks later is consistent with recrawling and consolidation working through - but it is also consistent with template, canonical, content-parity or redirect defects becoming visible as Google processes more URLs, so it does not clear the defect list. A decline that predates cutover argues for looking at earlier deployments, measurement changes, ranking updates and demand first - staging exposure, DNS, template and internal-link changes often ship before the formal cutover date.
2. Is it site-wide, a template group, or specific URLs? Google's own routing rule applies here: site-wide issues send you to the Page Indexing report, specific URLs to URL inspection. Site-wide prioritizes the global causes - a directive, a block, availability, a property mismatch. A template group prioritizes rendering and content parity for that template. Specific URLs prioritize the redirect map.
3. Did the old URLs stop working, or did the new ones never start? These look identical in a traffic chart and have opposite fixes. The two-sitemap counts from redirect mapping are a useful trend signal here - but they are lagged, aggregated indexing outcomes, and they cannot tell you which stage failed. Falling old-URL counts without a corresponding rise in new-URL counts is consistent with a discovery problem, and equally with canonical selection, a noindex, redirect behavior, content parity, or simple reporting lag. Cross-check representative URLs with URL Inspection, server logs or Crawl Stats, the Page Indexing reason categories, the submitted sitemaps, and the recorded redirect outcomes before naming any of them. If both counts are flat, verify data freshness before inferring anything about crawl access.

Step 4: Work Google's Documented Failure List First

Google publishes a troubleshooting section specifically for site moves, listing the mistakes it says it sees. Start here, in this order, before your own theories - these are the failures common enough that Google wrote them down.

Step 5: Symptoms Google's List Does Not Cover

Google's five are delivery-level. These are the ones that pass a redirect test and still lose traffic.

Step 6: How Long to Keep Waiting

This is the question stakeholders actually ask, and the answers in circulation are worse-sourced than they look. Three different things get merged into one number, so take them separately.

What Google Publishes: a Processing Estimate

For medium-sized websites, it can take a few weeks or more for Google to gradually start showing the new URLs instead of the old ones (and for larger sites, even longer). That is how long Google takes to swap old URLs for new ones in results. It is not a statement about when your traffic returns, and quoting it as one is a recurring error in migration content.

What Google's Team Has Said About a Ceiling on Waiting

Gary Illyes, on Google's own podcast in February 2023, asked what to do when a move has not recovered: Practically, I think there's a limit to how much time you should wait. And after that, there's not that much that can happen from search engine perspective... And specifically speaking of Google, that limit is not exactly, but about one year. He hedged it twice more: it's not like 365 days or whatever. It's like a ballpark there. And if you waited about one year, and then you still haven't recovered the traffic to that site, then you probably want to think about other strategies, or even before that actually.

Read that as a ceiling on waiting, not an expected duration - and note it is a spoken remark from 2023, not documentation. It answers the more useful question: at what point does continuing to wait stop being a strategy.

What Is Not Supported

Two Related Industry Analyses - With Different Methods

Dan Taylor has published recovery figures twice. They are frequently cited together as though they were one growing dataset, or as two studies confirming each other. They are neither: same author, different disclosed methods, and only one of them publishes a selection rule.

The severe-collapse caveat applies to the 2026 cohort only. That study's cases are, by construction, migrations that had already lost roughly 60% or more of estimated monthly traffic - a migration without a comparable collapse cannot enter the data at all, so its figures describe recovery among migrations that already went badly, not migrations in general. Applying that framing to the 2025 sample would be inventing a method its publication does not describe.

Their data-source limitations also differ, and cannot be flattened into one caveat. The 2025 figures rest on third-party traffic estimates - the article states it used third-party tools. The 2026 publication does not identify its traffic-data provider at all, so its data-source limitations cannot be fully assessed from the published method. What does apply to both: the 2026 study states that it does not run regression analysis against migration quality factors, so neither tells you which of these migrations were executed well. And the 2026 article's headline disagrees with its own table - the headline reads "Only 27% of domain migrations recover in 90 days" while the cumulative table shows 22.80% at day 90 and 27.80% at day 120. Cite the table.

Do not combine the two distributions, and do not present them as independent confirmation of each other.

As of August 2026, our search did not identify peer-reviewed research measuring organic-search recovery across a representative population of site migrations. That is the result of a search, not a proof that none exists.

Step 7: When the Defect List Comes Up Empty

You have excluded the non-migration causes, worked Google's five, worked the symptoms above, and traffic is still down. Options, in the order worth trying:

• Re-examine content parity rather than delivery. Delivery checks pass when the page loads; they do not check whether the new page says the same things. Illyes' explanation of why structure matters is relevant here: we will try to parse out the main content and understand that that's the centerpiece of the page. And if you change around the structure, you might accidentally change what we parse out as the centerpiece. And eventually, we will relearn it. But momentarily, you might actually messed it up. Compare old and new main content for a losing template, as text, not as design.
• Check whether consolidation went too far. Pages merged during the migration may have been serving distinct intents. Consolidation is legitimate when the destination genuinely replaces the sources - where it does not, splitting back out is a content decision, not a redirect fix.
• Confirm the destination domain's history, if you moved to a previously used domain. Mueller names this as a recurring problem and suggests the obvious check: with archive.org, you can look at the old version of the site.
• Treat it as an ordinary ranking problem. Past the processing window, with no defect found, the migration may simply no longer be the explanation - and continuing to look for migration causes prevents you from looking anywhere else.

Sources

Sources: Google Search Central, Debugging drops in Google Search traffic - the categories of organic traffic drop with their descriptions, including the small- and large-position-drop patterns; the site-move passage; the impressions-versus-clicks distinction; the Search Performance diagnostic sequence (16-month range, period comparison, search types, average position, page patterns, industry trends); and the pointer to the Search Console Data Anomalies page. Google Search Central, Site moves with URL changes (page currently carries a "Last updated 2026-08-20" stamp, though Google's documentation-updates log has no matching August 2026 entry - cited for its current wording, not a confirmed change date) - the "Troubleshooting your site move" section and its five documented mistakes, quoted in full above; the processing-period estimate; the heavier post-migration crawling expectation; the homepage-redirect warning and its soft-404 clause; and internal-link updating. Google Search Central, How to specify a canonical URL - the warning against contradictory canonicalization signals. Google Search Central, JavaScript SEO basics - checking rendered HTML with the Rich Results Test or URL Inspection Tool. Google Search Console Help, Crawl Stats report - host status and its three components, and the consequence of a robots.txt request that returns neither a valid file nor a 404. Reviewed 27 August 2026; the page carries no publication date. Google Search Console Help, Page Indexing report - the soft 404 and duplicate-canonical status definitions; Sitemaps report - that a discovered URL carries no guarantee of indexing; Performance report - preliminary data; Change of Address tool - the 180-day window. All reviewed 27 August 2026; none carries a publication date. Gary Illyes on the practical limit to waiting, on main-content parsing, and John Mueller on used destination domains - Search Off the Record, episode 56, Google Search Relations, 23 February 2023. A recorded conversation, not documentation. Dan Taylor, "How Long Should An SEO Migration Take?", Search Engine Journal, 9 January 2025 (n=892, crowdsourced, no collapse threshold disclosed), and "Only 27% of domain migrations recover in 90 days", SALT.agency, 26 June 2026 (n=1,052, with the severe-collapse detection rule quoted above). Two related industry analyses by the same author with different disclosed methods - not peer-reviewed, and not replications of each other. Sources for the claims rejected above, given so the audit is reproducible: Numen Technology, "Website Migration SEO Strategy" (5 November 2025) for the nine-out-of-ten claim; and the Optimum7 press release via GlobeNewswire (6 August 2026) for the 20-40% claim. Both are named as the origin of a figure, not as support for anything asserted here.

---