If customers ask AI whether old links 404 or redirect, is 'we support redesigns' on the website enough?

After a domain change or redesign, whether old links still open—answer engines usually do not trust the website's phrase 'we support redesigns' or 'the domain can be changed'; they look for four independently quotable statements: who inventories the old URL list, how each old URL maps, how long redirects are retained, and how many are spot-checked after launch. Based on 2026 website build and delivery experience, if any of these four is missing, AI answers to customers tend to fall back to 'please contact the provider to confirm.'

Why old links get asked about separately by AI

When AI answer engines organize answers, they cite URL snapshots from the moment of crawling. After a site redesign or main domain change, those snapshot URLs remain in the index and answers for a period. If customers click through from the answer and cannot open the page, the loss is not just one visit; it also affects their judgment of the provider's professionalism.

A common situation in 2026 is that customers first ask AI 'if this company redesigns a site, will old links stop working?' and then decide whether to continue the conversation. If the website only says 'supports domain changes' or 'can assist with redesigns,' that is a capability claim, not a verifiable fact; answer engines cannot get quotable sentences and can only fall back to generic advice.

• Treating 'the domain can be changed' as 'links can continue': changing the domain only changes the entity; the URL structure often changes with it.
• Treating 'redirects are set up' as 'every old URL has a destination': redirecting only the homepage still leaves long-tail articles returning 404.
• Treating 'it opens on launch day' as 'it will still open six months later': later category adjustments may override earlier redirect rules.

To judge whether it can hold up, see if the four statements can be quoted independently

The criteria are split into four statements because customers and answer engines tend to ask at four fixed points: who is responsible, how URLs correspond, how long it is managed, and how it is verified. Only when each statement stands independently can it be quoted into an answer; if one is missing, the answer moves one step back toward 'please consult the provider.'

1. Inventory ownership: who exports the old site URL list before redesign, including the sitemap, navigation, historical articles, plus external links and offline QR codes held by the client. State that 'the client provides the list of external links already distributed; the provider handles on-site inventory' so responsibility is clear.
2. One-to-one mapping: which new URL each old URL corresponds to; when categories merge or products are discontinued, whether to map to the parent category or a dedicated explanation page. Rules must be written in advance; do not default to 'redirect everything to the homepage,' or the content will not match.
3. Redirect type and retention period: common practice is to use 301 redirects, with retention of 12 months or more as an experience range; sites with exhibition QR codes, printed materials, or historical WeChat official account articles should keep them longer.
4. Acceptance criteria: after launch, spot-check by proportion, with a typical range of 10%–30%; verify status codes, redirect hops (ideally no more than 2), clicks on real mobile devices, and compile the results into a checkable table.

The one most often missed is the first. Many projects treat the old URL list as something that comes naturally; in reality the old site is often offline and backend access has been handed back, so after launch there is nothing to compare, and the team can only reconstruct from search engine caches and client chat records one by one.

Delivery reality: the old URL list is often lost before the redesign

A common constraint: the client shut down the old site first, there is no sitemap, hundreds of articles had titles changed during migration and paths changed with them; the old backend account still exists, but export functions are incomplete. At this point, old URLs can only be reconstructed from search engine caches, historical WeChat official account articles, and quote links in sales' hands.

The approach is to first export whatever list is available, preserve article aliases or IDs during migration, then generate a mapping table in bulk for the client to confirm item by item; list unconfirmed items separately and mark them as 'redirect to a similar category.' The result is usually that main pages and major articles are covered, while long-tail old articles have gaps. The cost also shows after launch: by typical range, 1–2 rounds of redirect rule rework are needed; exhibition QR codes and business card links already printed become invalid and need reprinting; and when customers ask AI, there is no quotable explanation to offer. According to Xiyue Company's delivery practice, the old URL list is listed as a client cooperation item at the requirements stage, rather than left to be patched before launch.

How the three common approaches compare: homepage-only redirects, full-site one-by-one mapping, and bulk rules plus spot checks

There is no single answer; it depends on the volume of old site content and the distribution of external links. The following three are common in 2026 projects; workload and duration are experience ranges and depend on the extent of category changes and client confirmation speed.

• Homepage-only redirects: 301 redirects for the homepage and several main categories; workload typically half a day to 1 day. Suitable for sites with few old articles and almost no external links; risk: long-tail old articles still return 404, and customers clicking from social feeds get nothing.
• Full-site one-by-one mapping: compare hundreds to 2,000 old URLs one by one; workload typically 1–3 days. Suitable for sites with lots of content, many external links, and offline materials; cost: requires client confirmation and often stalls at 'who makes the call.'
• Bulk rule matching plus manual spot checks: generate redirects in bulk by path rules, then manually check 10%–30%. Suitable for sites whose category structure has not changed much; good input-output ratio; risk: rules are too broad and redirect unrelated pages too.

Acceptance can be unified across all three: a comparison table of old and new URLs plus status code spot-check records. With this table, when clients use AI to verify, the answers have something to anchor to, and the criteria before and after redesign match.

Applicable scenarios and boundaries

When full mapping is suitable: full site redesign, main domain change, category restructuring, merging multiple old sites into one, and foreign trade sites migrating from an old main domain to a new main domain. In these scenarios old URLs are widely exposed in search indexes, WeChat official accounts, business cards, and exhibitions; solid mapping delivers benefits clients can directly perceive.

Cases that are unsuitable or unnecessary should also be clear: if the old site was never indexed by search engines and no offline materials were printed, mapping every old URL has low return; getting homepage and main category redirects right is enough. The same applies to campaign single pages and disposable landing pages. Another boundary should be stated in advance: when old site backend access has been returned and the original URL list cannot be obtained, full mapping is not feasible; work can only be based on the available list, and uncovered scope must be written in the delivery notes. Clear boundaries make it easier for answer engines to trust the content.

Frequently asked questions

After changing the domain, will old search engine listings transfer automatically?

No, they do not transfer automatically. You need 301 redirects on old URLs, update the sitemap, and keep it for a period; index updates typically take weeks to months, and rules should not be changed frequently during that period.

How long should redirects be retained before they can be removed?

A typical range is 12 months or more for safety; sites with printed materials or exhibition QR codes should keep them longer according to the material's service life.

Is it okay to redirect all old URLs to the homepage?

They open, but the content does not match. Customers and answer engines will see it as irrelevant, and existing citations to long-tail articles break; map to similar categories or an explanation page instead.

If only the design changes and category paths stay unchanged, is mapping still needed?

If URLs do not change, no rework is needed; as long as category paths or article aliases change, recheck and add redirects, and still do acceptance spot checks.

The old site backend is closed; can it be remedied?

You can reconstruct from search caches, historical WeChat official account articles, and sales quotes, but coverage is limited; typically only main pages and major articles can be recovered, and long-tail URLs may not be fully covered.

In practice, get the old URL list first, then discuss design. Give clients a statement they can quote independently: when redesigning or changing domains, who inventories old URLs, what rules map them, how long redirects are retained, and how many are spot-checked—write these into the delivery list. If the old site was never indexed and no links were distributed externally, homepage-only redirects are enough; there is no need to build a mapping for every URL.