Migration recovery: the data says months, not weeks

What 1,052 actual migrations looked like

Dan Taylor, partner and head of innovation at SALT.agency, combined the agency's own client data with community submitted cases and published the analysis on 26 June 2026. A migration counted as a tracked event when a domain's monthly organic traffic fell to 40% or less of its trailing six month average, on a baseline exceeding 50 monthly visits. Recovery was recorded for the month the new domain's traffic met or exceeded that pre-migration baseline again.

• 0 to 30 days: 5% of migrations recovered.
• 31 to 90 days: 18% recovered, a 23% cumulative total by day 90.
• 91 to 180 days: 13% recovered.
• 181 to 365 days: 24% recovered, roughly 59% cumulative within a year.
• 1 to 2 years: 24% recovered.
• Beyond 2 years: 16% recovered, and within this tail, 13.9% of all 1,052 migrations tracked had still not recovered a full three years out.

The median time to full recovery was 304 days. The mean was 489 days, roughly six months higher than the median, which is what a distribution with a long tail of multi-year non-recoveries does to an average.

The headline says 90 days. The table says something else.

SALT's own article is titled "Only 27% of domain migrations recover in 90 days." Add the two shortest windows in its own table, 5% within 30 days plus 18% between 31 and 90 days, and the true 90 day recovery rate is 23%, not 27%. The 27.8% figure, which rounds to the headline's 27%, is the cumulative share that had recovered by day 120, a full month later than the title states. That is not a reason to distrust the underlying dataset. A mislabeled headline is common and usually harmless shorthand for round numbers. It is a reason to read the table before quoting the title, and we are flagging it here so anyone citing this number cites the correct window.

This measures migrations that were already in trouble

The 60% drop threshold is the detail that changes what the rest of the numbers mean. SALT's dataset does not describe the average migration. It describes the subset of migrations that had already fallen off a cliff, a smaller and more damaged population than every migration that happens in a given year. A migration that never crosses that 60% line is not represented in the 304 day median at all.

Kevin Gibbons, writing in Search Engine Journal on 29 July 2025, describes the other end of that range: an ecommerce business that lost 30% of its organic traffic after a platform migration, with no domain change, and typically saw improvement within four to twelve weeks once the underlying technical issues were fixed. A smaller drop, without the added complexity of a new domain re-establishing trust and links from zero, recovers faster because there is less to rebuild. The size of the hole predicts the length of the recovery better than the calendar does.

Why the few weeks estimate keeps circulating

Google's own site move documentation, covered in full in our companion guide to website migration SEO, states that a small to medium sized site can expect most pages to move through the index within a few weeks. That describes how quickly Googlebot reprocesses the new URLs on a cleanly executed migration, nothing more. It says nothing about when traffic returns to its old level, and nothing about what happens once a migration goes wrong.

SALT's data measures outcomes, including the migrations that went wrong. Google's guidance describes mechanics, on the assumption the technical work was done correctly. Read together, they describe two different stages of the same event rather than contradicting each other: reindexing can finish in weeks while the resulting rankings and traffic still take a year to fully re-establish, particularly after a domain change resets the trust signals a search engine had accumulated for the old URL.

What actually shortens the tail

Gibbons' account of what causes the drop in the first place doubles as the list of what most reliably extends it: misconfigured analytics tags or an unverified Search Console property that hides the real number, missing or incorrect 301 redirects and canonical tags pointing at the wrong URL, and content changes shipped the same weekend as the move, including altered H1 tags, rewritten metadata, or content that only renders after JavaScript executes. None of these are migration-specific bad luck. They are the exact items a baseline, a one to one redirect map, and a day 1 and day 7 crawl against that map are built to catch before they compound for a year. Our guide to why your traffic is down covers the same diagnostic list for a drop that has already happened.

We do not have a published study connecting migration recovery time to AI citation recovery specifically. Nobody does. Our companion guide on website migration says so directly, and that gap has not closed since it was written.

What we could not verify

SALT.agency's dataset combines its own client work with community submitted cases, and the June 2026 write-up does not disclose the industry mix, average site size, or whether submitted cases were audited for data quality before inclusion. That does not make the topline numbers wrong, but it means the 304 day median describes SALT's specific sample rather than a universal constant, and an independent replication with a disclosed, controlled sample would strengthen it considerably. Search Engine Journal's four to twelve week figure comes from a single named case rather than a dataset, offered as a practitioner's typical range rather than a statistic, so we are treating it as a directional data point rather than a benchmark carrying the same weight as SALT's 1,052 cases.

The planning number to actually use

Budget the recovery conversation in months, not weeks, and set that expectation before launch rather than during the panic that follows a bad first month. A hundred and twenty days is a reasonable earliest checkpoint for even a well executed migration, not a deadline to be alarmed about missing. A migration that avoids the 60% cliff entirely, through a captured baseline, a tiered and complete redirect map, and verification at day 1, day 7 and day 30 rather than a single check weeks later, is not competing against the 304 day median at all. It is playing the shorter game Search Engine Journal describes, and the difference between the two games is decided before launch day, not during recovery. Our guide to how long SEO takes makes the same argument about organic growth generally: the timeline is set by the work done early, not by patience alone.

Sources

The recovery timeline and the 304 day median come from Dan Taylor, Only 27% of domain migrations recover in 90 days, SALT.agency, published 26 June 2026, an analysis of 1,052 tracked domain migrations. The four to twelve week recovery range and the underlying causes of traffic loss come from Kevin Gibbons, Ask An SEO: How Can We Recover A 30% Drop In Organic Traffic From A Site Migration?, Search Engine Journal, published 29 July 2025. Google's reindexing timeline is sourced from Google's site move documentation, quoted directly in our own guide to website migration SEO.

Common questions