The short answer
A redesign does not preserve search visibility simply by copying text. You must inventory current URLs and performance, decide what stays, map each old page to an equivalent destination, test redirects, canonicals, hreflang and measurement data, then monitor indexing after launch. Risk falls when design, content, development, SEO and business teams validate every stage together.
- Inventory before deleting
- Map every old URL
- Test signals and conversions
- Launch with stop criteria
- Monitor by language and page type
Why can a redesign affect search rankings?
A redesign often changes several systems at once: design, templates, navigation, content, URLs, CMS, hosting and measurement. Google must recrawl pages, interpret new relationships and confirm which addresses replace the old ones. Temporary fluctuations can occur, but lasting losses are not inevitable. They mainly arise when useful pages disappear without an equivalent, redirects are wrong or signals become contradictory.
Start by defining the project type. A visual modernisation with unchanged URLs has fewer variables than a change in site structure. A domain, protocol or CMS migration adds further dependencies. Google recommends changing one thing at a time where possible and preparing precise mappings between old and new URLs. This discipline also aids diagnosis: if everything changes on the same day, the cause of a drop becomes difficult to isolate.
The useful question is therefore not “can we guarantee zero fluctuation?”, but “what evidence will let us detect and correct an issue quickly?”. The answer combines inventory, pre-launch tests, decision criteria and post-launch monitoring.
| Type of redesign | Relative risk | Main checks | Best choice according to need |
|---|---|---|---|
| Same URL, new design | Low to moderate | Content, templates, performance | Targeted visual refresh |
| New structure and URLs | High | Mapping, redirects, links, hreflang | Repositioning or simplification |
| New CMS or hosting | Moderate to high | Rendering, statuses, canonicals, performance | Technical modernisation |
| New domain | Very high | All checks and Search Console ownership | Justified rebranding |
Build an inventory before designing
The inventory is the current site’s memory. Export all URLs found in the sitemap, CMS, analytics tools, Google Search Console, logs if available and an internal crawl. Add pages receiving external links, conversions or seasonal traffic. No tool sees everything: the sitemap may omit an old landing page, while the crawl misses an orphan page still visited from Google.
For each URL, record the HTTP code, title, canonical, language, depth, template, traffic, queries, inbound links, conversions and business owner. Data need not be perfect to be useful; it must make decisions explicit. Then classify each page: retain, improve, merge, redirect, remove or review.
The inventory must also cover forms, downloads, measurement events, consented pixels, feeds, internal search engines and legal pages. A successful redesign extends beyond pages appearing in the menu. In a project supported by the Websites & redesigns service from GVISION Studio, GVISION can connect this mapping to UX needs, content and technical constraints rather than treating SEO as a final check.
Preserve intent, not just words
A new page can copy an old page’s text while losing its function. Intent depends on the topic, but also format, depth, audience, navigation and next action. A service page reduced to three paragraphs on a general page no longer provides the same answer. Conversely, merging two redundant pages can improve clarity if the destination genuinely covers both needs.
Work by role: commercial page, guide, case study, category, resource, contact or help. Compare the main query, secondary questions, internal links and expected conversion. Keep the elements explaining why the page gets visits or enquiries, then improve their presentation. Do not mechanically protect all old content: protect what delivers verifiable value.
For a multilingual Belgian site, check that versions offer equivalent value. An important FR page without an NL equivalent must not be redirected to a generic Dutch page for convenience. Decide whether to create the translation, keep the page temporarily or provide a localised alternative that truly meets the same need.
Build a URL-by-URL mapping table
The mapping table is the migration’s central document. Every old indexable URL receives a decision and, when it changes, a single new URL. The destination must be the closest equivalent, not automatically the homepage. Redirecting everything to the homepage offers little help to users and may be interpreted as a logical error.
Add the old URL, new URL, expected status, reason, language, owner, test date and result to the table. Flag difficult cases: content mergers, removal without replacement, parameters, PDFs, campaigns, subdomains and pages requiring login. The table supports development, acceptance testing and post-launch support.
Avoid generating rules solely from the new menus. URLs have a history: old campaigns, slash variants, capital letters, translated paths or files still linked by partners. Normalise patterns without creating overly broad redirects. A manual sample remains necessary even when rules are generated automatically. Finally, keep the table versioned to identify who changed a destination and why.
Choose and test the right redirects
For a permanent move, Google considers permanent HTTP redirects, including 301 and 308, a strong signal towards the new URL. The exact code depends on the platform, but behaviour must be permanent, direct and stable. An old page should reach its final destination in one hop. Chains slow crawling, complicate diagnosis and can break when an intermediary disappears.
Test rules in a representative environment, then externally after launch. Check the code, Location header, final destination, language and absence of loops. Check a sample of priority pages, but also all URLs through an automated test. Important resources — images, PDFs or indexed scripts — deserve their own decision.
A redirect is no substitute for a good page. If the destination no longer meets the intent, consider retaining an explanatory page or returning a consistent removal status. Keep redirects long enough for users, bookmarks and historical links. Google’s documentation on redirects and search serves as a technical reference, but destination relevance remains an editorial decision.
Align canonicals, hreflang and languages
Each localised URL must declare itself as its own canonical when it is a distinct page. FR, NL and EN variants are linked through hreflang; they are not canonicalised to French. A contradictory canonical asks Google to group pages while hreflang says they are alternatives. The result becomes unpredictable.
Check complete groups: each variant references the others and itself. The x-default value may point to the selection version or intended default page. If TranslatePress manages these tags, do not duplicate them manually in the theme or SEO plugin. Test rendered HTML, not just interface settings.
Slugs can be localised, but the redirect table must then manage each language separately. Also check menus, language selectors, canonicals, sitemaps and structured data. A common error is translating the interface while leaving internal links pointing to old French URLs. Walk through real Dutch- and English-speaking visitor journeys to confirm end-to-end migration consistency.
Protect staging without blocking the public site
Staging must stay out of the index while remaining accessible to authorised people and tools. Authentication is generally more robust than a noindex tag alone. Robots.txt controls crawling, not necessarily whether a URL appears in results; Google states that a blocked URL may still appear without a snippet if known from elsewhere.
The classic trap happens at launch: noindex, HTTP protection, a robots rule or a staging canonical remains active. Add these items to a cutover checklist and test them from an external connection. Conversely, do not make staging publicly indexable just to allow crawling. Use appropriate acceptance-testing access.
The architecture forhosting, security and maintenance also influences migration: certificates, DNS, cache, CDN, backups, monitoring and rollback capability must be prepared. GVISION recommends appointing one person authorised to initiate cutover and another to validate the result. This separation reduces improvised decisions when several teams act simultaneously.
Organise technical and editorial acceptance testing
Acceptance testing starts before launch day. On a representative URL set, check HTTP codes, titles, descriptions, H1s, canonicals, hreflang, structured data, links, images, alt attributes, forms and conversion events. Compare old and new rendering to spot missing information or actions. Business validation complements the crawl: a robot cannot tell whether a quote request, registration or booking actually completes.
Also test less visible states: 404 errors, empty results, form confirmation, consent refusal, internal search, printing, downloads, keyboard use and mobile zoom. Each issue receives a severity level, owner and decision. A defect list without owners is not acceptance testing.
The photograph in this section illustrates the validation meeting between development and marketing. Keep evidence: crawl exports, screenshots of critical templates, form results and event comparisons. They speed investigation if a metric changes after launch. Acceptance testing is not intended to demonstrate perfection, but to establish what was checked, accepted or deferred.
Carry over content, metadata and internal links
First carry over content that satisfies searches and supports conversions. Keep useful factual information, evidence, questions and calls to action, then adapt the structure to the new design. A more elegant component does not compensate for removing a paragraph that answers the query. Check titles and descriptions page by page; a generic value copied site-wide dilutes meaning.
Internal links convey context and organise discovery. Update links in content, menus, breadcrumbs, footers and related components. They must point directly to new URLs without going through a redirect. Identify orphan pages and ensure strategic pages remain accessible from relevant hubs.
Do not forget media. Keep files still being searched for or redirect them to a useful equivalent. Provide dimensions, compression, alternative text and captions according to context. If names change, update references in structured data and social cards. Finally, review on mobile: a block hidden by CSS may exist in code but be unusable for the reader.
Treat performance and accessibility as release criteria
A redesign can improve the experience while making the site heavier. Measure Core Web Vitals before and after using lab data and, where available, real-world data. The web.dev reference includes LCP for loading, INP for responsiveness and CLS for visual stability. Compare equivalent page types rather than a misleading overall average.
Define budgets: page weight, image size, number of third-party scripts, server time and behaviour on a realistic mobile device. Optimise images, fonts, caches and components before launch. A new theme does not justify unlimited resource inflation.
Accessibility must be tested through automation and human checks. WCAG 2.2 covers keyboard navigation, visible focus, contrast, text alternatives, labels and errors, among other areas. An automated tool cannot validate the relevance of a reading order or message. Include these criteria in the definition of “ready”, alongside critical forms and templates. This reduces costly post-publication fixes and improves usability for far more visitors than formally recognised scenarios alone.
Preserve analytics, Search Console and conversions
An apparent decline may come from broken tracking rather than actual loss. Document current tags, the consent manager, events, goals, campaign parameters and internal exclusions. On staging, use a separate measurement environment or debug mode to avoid polluting production data.
Before cutover, record a baseline: entry pages, organic clicks, impressions, queries, conversions, error rate and performance by page type and language. Avoid promising a precise level after migration; these data detect gaps rather than guarantee a ranking. Check that event definitions remain comparable, otherwise the before/after curve no longer has the same meaning.
After launch, submit the new sitemap in Search Console and track coverage, errors, selected canonicals and performance. A sitemap helps discover URLs but does not guarantee indexing. Keep the old sitemap or a list of old URLs during the checking phase if that helps monitor redirects. Always cross-check signals: Search Console, analytics, logs, monitoring and enquiries received.
Prepare cutover and rollback
The launch plan must follow a clear timeline. Freeze non-essential changes, make backups, validate the URL table, lower DNS TTL if necessary, publish, purge caches, run critical tests, then authorise communications. Every step has an owner, time, evidence and stop condition.
Rollback is not always a simple button. Returning to old code when the database, content or DNS has changed can create further inconsistency. Define what can be undone, within what timeframe, by whom and with which data. For a limited SEO issue, correcting the rule or page may be safer than a full rollback. For a payment or login failure, stopping may be immediate.
The localised infographic summarises five milestones: inventory, mapping, acceptance testing, controlled launch and monitoring. Use it as support, not a replacement for the detailed plan. Arrange launch-day cover and a single decision channel. Sales and support teams must know where to report a broken URL or impossible conversion without scattering information.
Monitor for 30, 60 and 90 days without waiting for a crisis
The first hours are for checking availability, forms, priority redirects, robots, canonicals, sitemap and analytics. Over the following days, analyse 404s, chains, non-indexable pages, canonical choices and language-specific gaps. After several weeks, observe click, impression, conversion and engagement trends by page group rather than one overall figure.
Create a correction register with impact, hypothesis, change, date and result. Do not change several variables simultaneously without need: you would lose the ability to learn. Alerts must lead to a defined action. A rise in 404s from unknown bots does not have the same priority as a heavily visited old service page without a destination.
At 30 days, correct mapping and measurement errors. At 60 days, examine underperforming content and internal links. At 90 days, decide which redirects and monitoring become operational. Common mistakes include launching on a Friday evening, forgotten noindex, blanket homepage redirects, contradictory canonicals, an old sitemap and lack of ownership. Simple governance prevents recurrence.
Frequently asked questions
Does a redesign always lose rankings?
No. Fluctuations are possible during recrawling, but thorough preparation reduces the risk of lasting loss.
Should you keep all old URLs?
No. Retain useful URLs or redirect them to a relevant equivalent. Removal without replacement can return a consistent status.
301 or 308: which should you choose?
Both indicate a permanent move. Choose according to the platform and above all test that the redirect is direct, stable and correct.
How long should you keep redirects?
Long enough for search engines, users, bookmarks and external links. Strategic redirects often become a lasting rule.
Can staging be blocked using robots.txt alone?
It is not ideal. Authentication provides better protection; robots.txt prevents crawling but does not guarantee that a known URL disappears.
When should you submit the new sitemap?
After launch and URL validation. Submit it in Search Console and monitor errors and selected canonicals.
What should you check first after launch?
Availability, forms, key redirects, noindex/robots, canonicals, hreflang, analytics, sitemap and 404 errors.
Conclusion: treat the redesign as a measurable migration
A safe redesign connects goals, inventory, mapping, acceptance testing and monitoring. It does not promise unchanged results; it organises evidence and corrections. To turn this plan into scope, responsibilities and a schedule, you can prepare your redesign with GVISION.
Official sources
Published on . Updated on .
Let’s talk about your project
Want to apply this advice in your organisation? Let's review your needs and next steps together.
Talk to GVISION



