Website migration SEO checklist

Establish what "no loss" means in numbers, before anything changes. Everything measured after launch is compared against what is captured here, so a benchmark taken late is a benchmark that proves nothing.

1. Get read-only access to Search Console and analytics

   Shared

   Both properties, before any other work starts. Access requests routinely take longer than the person requesting them expects, and every later step depends on this one.

   Output Verified access to both properties

2. Export 12 months of URL-level performance

   SEO

   Clicks and impressions per page, not per query. Freeze the file with the export date in its name. This is the only dataset that makes a post-launch comparison possible.

   Output Dated benchmark export

3. Crawl the current site in full

   SEO

   Every indexable URL with status code, canonical, meta robots, title, headings, word count, internal link count and structured data recorded per URL.

   Output Production crawl export

4. Pull the referring domain data

   SEO

   Which pages carry external links, from any backlink tool. This list overlaps far less with the traffic list than people expect, and the difference is where equity hides.

   Output Pages ranked by referring domains

5. Request server log access

   Engineering

   Ask early even if it turns out to be unavailable. Logs are the only source that shows what crawlers actually do rather than what you infer they do.

   Output Log access, or a documented decision that it is unavailable

6. Record the indexable URL count

   SEO

   A single number, dated. After launch, comparing this against the new count is the fastest way to detect index bloat arriving with the new platform.

   Output Baseline indexable URL count

7. Tier URLs by priority

   Shared

   Combine organic clicks, referring domains and revenue where available. The top tier gets one-to-one mapping; lower tiers can be handled by rule.

   Output Tiered URL inventory

Gate — do not start the next phase until this is true A dated, frozen benchmark exists: URL-level clicks and impressions for the last 12 months, the complete indexable URL inventory, and a full crawl of the current site.

Decide where every URL goes. This is the phase that determines whether the migration works, and it is the one most often compressed when the build runs late.

1. Review the proposed URL structure against the current one

   Shared

   Before mapping begins. If the new structure creates avoidable mapping ambiguity, this is the last point at which saying so is cheap.

   Output Structure findings with named owners

2. Write pattern rules for the bulk of URLs

   SEO

   Most URLs map cleanly by rule. Writing the rules first turns the remainder into a visible, countable exception list rather than a vague concern.

   Output Rule set with match counts

3. Test the rules against the real inventory

   SEO

   Not against the pattern — against the actual URL export. Every rule has exceptions, and they are always in the oldest content nobody remembers.

   Output Rule coverage report with exceptions

4. Resolve every exception by hand

   Content

   Pages whose category no longer exists, merged product variants, retired campaigns. Each needs a destination, an owner and a deadline.

   Output Exception map with owners

5. Decide the deliberate removal list

   Content

   Pages with no relevant destination return 410 Gone rather than redirecting somewhere irrelevant. Make this a recorded decision, never a default.

   Output Documented 410 list

6. Specify where redirect rules will live

   Engineering

   Edge, application, or both. Edge rules respond faster and survive deploys; application rules suit logic needing database lookups. Decide and document the split.

   Output Redirect architecture specification

7. Plan the internal link rewrite

   Engineering

   Internal links must point at final destinations in the source, not rely on redirects. Include absolute URLs stored in content database fields.

   Output Internal link rewrite scope

Gate — do not start the next phase until this is true Every URL in the priority tiers has a named destination and a status code. The count of unmapped URLs is zero, not "nearly zero".

Prove the new site presents what the old one presented. The parity diff finds the problems nobody thought to test for, which is reliably where the damage is.

1. Protect staging with HTTP authentication

   Engineering

   Not robots.txt. A Disallow rule does not prevent indexing of a URL that has been linked from anywhere, and staging URLs get pasted into tickets and chat.

   Output Staging behind authentication

2. Crawl staging using auth credentials

   SEO

   Configure the crawler with the credentials rather than removing the protection. Taking auth off so a tool can crawl is how staging sites get indexed.

   Output Staging crawl export

3. Diff staging against production, field by field

   SEO

   Titles, descriptions, H1s, canonicals, meta robots, word counts, internal link counts, structured data types, image alt coverage, hreflang.

   Output Parity diff with variance counts per field

4. Compare indexable URL counts

   SEO

   A large unexplained increase means the new platform generates URLs the old one did not. That is index bloat arriving with the build, and it is cheaper to prevent than to undo.

   Output URL count comparison

5. Verify rendering for every template

   SEO

   Fetch raw HTML and confirm body content, internal links and canonical tags exist without JavaScript execution. Do this per template, not per page.

   Output Rendering report by template

6. Test priority redirects against staging

   SEO

   Every URL in the priority tiers, individually, for final status code and hop count. Chains and loops fail the gate outright.

   Output Redirect test results

7. Confirm the noindex removal plan

   Engineering

   Staging carries noindex. Establish exactly how it is removed at cutover and who verifies its absence in production, in writing, before launch day.

   Output Documented noindex removal step with an owner

Gate — do not start the next phase until this is true Every unexplained difference in the parity diff is either fixed or accepted in writing, and every priority URL resolves in one hop to a 200 on staging.

A monitoring shift, not a strategy session. Everything on this list should already have been decided; today is about verifying that what was decided actually shipped.

1. Lower DNS TTL 48 hours ahead

   Engineering

   Drop to 300 seconds so the cutover propagates in minutes and a rollback is fast. This has to happen two days early to be useful, which makes it easy to miss.

   Output TTL confirmed at the registrar

2. Verify redirects on the live host immediately

   SEO

   As soon as DNS resolves. Staging tests do not prove production behavior — edge configuration frequently differs between the two.

   Output Live redirect verification

3. Check robots.txt in production

   SEO

   The staging version shipping to production is one of the two most expensive launch-day failures. It takes thirty seconds to check.

   Output Production robots.txt confirmed

4. Check meta robots and X-Robots-Tag in production

   SEO

   The other expensive failure. A site-wide noindex can begin removing pages within days. Check headers as well as tags — X-Robots-Tag is invisible in page source.

   Output Production directive audit

5. Submit updated sitemaps

   SEO

   New sitemap URLs submitted in Search Console, old ones removed. Confirm sitemap URLs match canonicals exactly, trailing slash included.

   Output Sitemaps submitted and accepted

6. File Change of Address if the domain changed

   SEO

   Only after the sitewide redirect is live and verified, and only for a genuine domain move. It does not apply to protocol or structure changes.

   Output Change of Address filed

7. Watch for crawl errors and server load

   Engineering

   Crawl activity spikes after a migration as search engines rediscover the site. Confirm the infrastructure handles it without timing out.

   Output First 48 hours monitored

8. Know the rollback trigger and who can pull it

   Shared

   Agreed in advance, in writing. Deciding what constitutes a rollback condition during an incident produces the wrong decision.

   Output Rollback criteria with a named decision-maker

Gate — do not start the next phase until this is true Sampled priority redirects return one hop to a 200 on the live host, and production carries no accidental noindex, Disallow or stale canonical.

Measure against the frozen benchmark on a matched cohort of URLs. Week-over-week comparison in the first fortnight mixes recrawled pages with pages Google has not looked at yet, and tells you almost nothing.

1. Compare a matched URL cohort, not the site total

   SEO

   Follow each pre-migration URL to its destination and compare that page's performance. If A 301s to B, A's performance is now B's performance.

   Output Cohort comparison against benchmark

2. Re-test the redirect map against production

   SEO

   Releases after launch introduce chains. A rule added in week three often runs before the migration rules and quietly adds a hop to everything.

   Output Re-test results with any new exceptions

3. Track the Googlebot hit distribution

   SEO

   In server logs, watch requests shift from old URLs to new ones. This is the only direct evidence of how discovery is actually progressing.

   Output Crawler request distribution over time

4. Watch indexation counts in both directions

   SEO

   Old URLs should fall out; new URLs should be discovered. A rise in total indexable URLs against the baseline means bloat rather than growth.

   Output Indexation trend against baseline

5. Identify losing cohorts by template and section

   SEO

   Losses concentrate in identifiable groups — a template that lost its canonical, a section that lost internal links. Grouping makes the cause visible.

   Output Loss analysis grouped by template and section

6. Rank remaining issues by traffic at risk

   SEO

   Not by severity label. A critical issue affecting forty unvisited URLs should not outrank a moderate one affecting a revenue-driving section.

   Output Prioritized remediation list

7. Keep old domain redirects live

   Engineering

   For at least twelve months where a domain changed, and ideally indefinitely. Letting the registration lapse discards accumulated signal permanently.

   Output Registration on auto-renew with a named owner

Gate — do not start the next phase until this is true Googlebot requests predominantly hit new URLs, and cohort performance has returned to the pre-migration range or the remaining gap is explained and owned.