Measuring whether AI models mention your client’s brand
A growing share of buyers now ask a language model for recommendations before they ever open a search results page. For considered...
A redesign that loses thirty percent of organic traffic is not a redesign. It is a downgrade the client paid for, and it takes six to nine months to recover from if it recovers at all.
It is also almost entirely preventable. Rankings do not disappear because a site looks different. They disappear because URLs moved without redirects, because pages that were earning traffic got deleted for looking dated, or because the new build shipped without the technical signals the old one had. All three are process failures rather than technical ones.
The short version:
Here is the sequence we run on every rebuild and migration.
Do this in week one, before designs, because the data influences what gets designed.
Crawl the existing site and export every URL that returns a 200. Pull twelve months of Search Console data at page level: clicks, impressions, and the queries each page ranks for. Pull analytics for the same period, including conversions per landing page, not just sessions.
Now join them. You want one spreadsheet with a row per URL showing traffic, conversions, and top queries. This document is the single most useful artefact in the whole project, and most rebuilds are quoted without it existing.
Two things become obvious immediately. First, a small number of pages are doing most of the work, and they need protecting. Second, there are usually pages earning steady traffic that nobody internally remembers writing, which is exactly the kind of page that gets deleted during a redesign because it does not fit the new navigation.
Every URL gets one of four labels, and no URL gets left unlabelled.
Keep the URL. Same address, new design. The safest outcome and the one to aim for on anything earning traffic. Changing a URL that is ranking well in order to make it fit a tidier structure is a trade almost nobody should take.
Move it. New address, 301 redirect from the old one. Fine when the structure genuinely has to change, provided the redirect goes to the closest equivalent page and not to the homepage.
Merge it. Several thin pages become one strong page, with all the old URLs redirecting to it. Often the right call for sites with years of overlapping content, and it usually improves rankings rather than risking them.
Remove it. Only for pages with no traffic, no conversions and no inbound links. Serve a 410 rather than redirecting it somewhere irrelevant.
The failure mode here is redirecting large numbers of pages to the homepage because it feels safer than a 404. It is not safer. It signals that the content is gone and it wastes the equity the page had.
These are the details that get missed when a rebuild is treated as a design project.
Title tags and meta descriptions do not automatically survive a platform change, and the new theme’s defaults are almost never what the old pages had. Export them alongside the URL list and reapply them, then improve them deliberately rather than by accident.
The same applies to structured data. If the old site had product, article, FAQ or organisation schema and the new one does not, you have removed something search engines were using. Canonical tags need checking on every template, particularly on paginated archives and anything with query parameters. Heading hierarchy matters more than most designers expect: an h1 that became a styled div because it looked better is a real loss.
Also check what the new site does with images. A rebuild that ships without alt text on templates that previously had it will lose image search traffic, and image search is a meaningful share of traffic for product and equipment businesses.
Take the full list of old URLs, point it at the staging environment, and crawl it. Every single one should return a 301 to a 200, in one hop.
Two specific things to look for. Redirect chains, where an old URL goes to a second URL that then goes to a third, which dilutes the signal and slows the page. And redirect loops, which are more common than they should be and take a page out of the index entirely.
Do this while the site is still on staging. Fixing a hundred redirects the week before launch is a normal afternoon. Fixing them after launch means every one of those URLs served a 404 to a crawler first.
Launch day is not the end of this work.
Submit the new sitemap immediately and confirm it is being read. Watch Search Console index coverage daily for the first fortnight, then weekly. You are looking for a rise in crawl errors, pages dropping out of the index, and any spike in 404s that suggests a URL nobody accounted for.
Expect a dip. A small drop in weeks two and three is normal while search engines reprocess the site, and it usually recovers by week five or six. What is not normal is a decline that continues past a month, or a specific high-value page that vanishes. Both are signals to go back to the URL map rather than wait it out.
Tell them the dip is coming, before it happens, with a number and a timeframe attached.
A client who was warned reads a temporary decline as evidence you know what you are doing. The same client, unwarned, reads it as the redesign having damaged their business, and that conversation is very difficult to recover from even when the traffic comes back.
It is the cheapest risk management on the entire project, and it costs one paragraph in the launch email.
With a complete redirect map, expect a small dip in weeks two and three and recovery by week five or six. A decline continuing past a month indicates a real fault, usually an unmapped URL set or missing technical signals, and should be investigated rather than waited out.
No. Redirect each old URL to its closest equivalent page. Bulk redirects to the homepage are treated as soft removals and waste the ranking equity those pages held. Pages with no traffic, conversions or links should return a 410 instead.
No. Titles, meta descriptions, schema and canonical tags all need exporting and reapplying deliberately. New theme defaults are almost never what the ranking pages had, and this is one of the most common silent losses in a replatform.
We build under your brand, on a US contract, with one Houston-based point of contact. Send a brief and judge us on the estimate.