What Happens When You Move a 1,500-Page Website

Updated since publication. After this went live we plotted our own decline week by week instead of month by month, and it corrected part of the story - including a conclusion we'd been fairly confident about. The section on what we blamed has been rewritten rather than quietly patched, because the correction is the most useful thing in here.

In April we replatformed leadpages.com. About 1,500 pages moved to a new codebase, and the site changed address at the same time, from www to the bare domain.

Everyone who has done a migration knows the script. Traffic dips, you hold your nerve for a few weeks while the crawlers re-learn the site, and it comes back.

Ours didn't come back.

In March we'd had our best average search position in sixteen months. By July, daily organic visits had fallen by nearly two-thirds from that peak. And the strangest part, which took us three months to see clearly: year over year, our rankings were improving the whole time. Most of our pages ranked better in July than they had a year earlier. The traffic fell anyway.

This is the story of what that turned out to be, the seven explanations that died on the way, the diagnosis we got wrong twice, and the checklist I wish we'd had in April.

What everyone said it was

The explanation arrived quickly, and it was reasonable. A Google core update had rolled out in June. Everyone said the same thing: it's the update, it will settle.

I didn't buy it. My reasoning felt airtight: our decline had started back in April with the migration, so an update in June couldn't be the cause of something that began two months earlier. The update explained when we noticed. It didn't explain when it began.

That reasoning was wrong, and it was wrong because I was reading the wrong chart.

Monthly figures are what most people look at, and they lag badly. A month that contains both good weeks and terrible weeks averages into something mild, and the average keeps sinking for weeks after the actual damage has stopped. When we finally plotted the same data week by week, the shape was nothing like the smooth April-onward slide we'd been arguing about. It was three separate drops:

• Early March into April: down about a quarter. This one is ours. It lines up with the migration.
• April: flat. Nothing moved for the entire month.
• Through May: down about another quarter.
• Two weeks in mid-June: down more than half. Larger than the other two drops combined.

That mid-June fall is the single biggest event in the whole series, and it lands exactly where the core update did. The team's first instinct was closer to right than the tidier theory I replaced it with.

So the trap I set out to describe is real, but it caught us twice. The first time, everyone accepted the explanation that fit the moment we noticed. The second time - the more embarrassing one - I replaced it with a cleaner story of my own, one where a single cause explained everything, and then stopped checking. Both mistakes had the same fix: plot it weekly and look at the actual shape before you commit to a narrative.

One more thing worth reporting, because it changes what to do next: it has stopped. The last five weeks are flat, and the most recent is the highest of the five. Whatever this was, it is no longer in progress.

Why we stopped waiting

Here is the part where I'm supposed to say we called in the experts. We don't have any. Like most companies our size, we have no SEO team - no agency on retainer, no specialist in-house. For a while, that was part of why waiting felt like the only plan.

Two things ended the waiting.

The first was what organic actually carries. For us it isn't an audience number, it's signups and trials - real pipeline. Watching visits fall is uncomfortable. Watching the channel that feeds new customers die quietly while you wait for it to "settle" is not survivable.

The second was realising we didn't need a hire to run the investigation. We ran it with AI agents. One pulled the performance history of every page on the site out of Search Console. One traced what had happened to every URL from the old site - where it went, what redirected, what vanished. One read the live results pages for our most important searches and recorded who actually ranks there now. Every explanation anyone had offered got turned into a claim that could be tested, and then tested against live data.

What came back changed what we thought the problem was - twice, as it turned out.

What the data actually showed

Three facts, all verified, that refused to fit together:

Clicks were down 45% year over year, comparing the same April-to-July window, with the loss concentrated on our highest-traffic pages rather than spread evenly.

Impressions were flat. Google was showing us roughly as often as ever.

Rankings were up year over year. Everywhere. Of the pages that survived the migration, 86% ranked better than a year earlier - the median page improved fifteen positions.

Better seats. Smaller room. A site does not lose half its clicks while out-ranking its own previous year because of a quality problem. Something structural had happened, and something in the results pages themselves had changed. Both turned out to be true.

One caveat we missed the first time, and it matters if you are running this analysis on your own site. That ranking improvement is a year-over-year comparison. Zoom into the two weeks of the June fall on its own and the picture reverses: our average position got worse through it, and our impressions dropped sharply with it. Both readings are correct, they just answer different questions. "Are we better than last year?" and "what happened to us in June?" have opposite answers, and if you only run the first one you will conclude nothing is wrong with your rankings when something briefly was.

Seven explanations, tested, dead

The audit turned every theory we'd heard into a testable claim. Seven died in one day.

1. "The lost URLs are the cause." We did lose pages in the move - over a thousand addresses stopped earning. Their total contribution to the decline: about three percent. Real, and nowhere near enough.
2. "The site exists on two addresses and Google is confused." Every old address redirects in one hop. Canonicals clean. Sitemap clean. Dead.
3. "The www-to-bare-domain move never finished." Google's own inspection tool: the old address is recorded as redirected. The consolidation is complete. Dead.
4. "The homepage's authority got split." Link tools show links pointing at both addresses, but Google follows them through the redirect. Dead.
5. "The whole site got demoted." Demoted sites do not out-rank their previous year on 86% of pages. Dead.
6. "One section broke and is dragging the rest." The year-over-year ranking improvement is uniform across every section. Dead.
7. "The pages stopped being relevant for the money terms." The pages most affected are literally titled for those terms. Dead.

Note what is not on that list: the core update. We tested seven theories hard and left the one everybody started with largely unexamined, because we thought the timeline had already ruled it out. It hadn't.

What it actually was

Three things, not two - and not in the order we assumed.

The largest single event: the June core update. More than half the fall happened in two weeks, and those two weeks line up with it. This is the part we do not control and cannot fix directly; the only response to a quality re-ranking is better pages, which is why the repair work that followed focused on consolidating thin, overlapping content rather than chasing redirects alone.

The part that was ours: the migration moved the house and left no forwarding addresses.

When the site changed domains, the domain-level redirect went in - anyone visiting the old address reached the new one. What didn't ship was the page-level map: this old URL means that new URL, for every one of 1,500 pages. In the week of the move, six lost pages were caught and four of them were rebuilt by hand from archived copies. Nobody could audit how many weren't caught, because the map didn't exist.

Authority in search flows through those individual connections, and through internal links - which had also been pruned when the new site launched with a slimmer navigation. The systematic repair, nearly 400 page-level redirects plus consolidation of overlapping pages, shipped three months after the move. Whether it rebuilds everything that was lost is a question the next two months will answer - search moves slowly, and we are measuring rather than assuming. What is certain is the cost of the gap: for three months, Google was re-learning a site whose pages had quietly disowned their own history.

The one that outlasts both: the results page changed shape underneath us.

AI-generated answers now sit above the results on seven of our ten most important search terms. Sitewide, our click-through rate fell from 0.85% to 0.39% while our year-over-year position improved. The same seat earns half the clicks it used to, because an answer now sits between the searcher and the results.

That third cause is the one that matters beyond our own story, because no redirect map fixes it and no update recovery reverses it. Ranking well is no longer the same thing as being seen.

The part we were most afraid of

The reason we stopped waiting was fear for signups and trials. So here is the strangest finding of the whole audit: they never moved.

Since we started measuring the channel in June, organic trial starts have held steady, week after week, while visits kept falling. The clicks we lost were overwhelmingly low-intent - informational searches, curiosity traffic, visits that were never going to become customers. The visitors with intent kept finding us.

I want to be careful with that reassurance, because it has an expiry date. Visibility is where next year's customers come from, and you cannot shrink your way to growth. But if you are staring at a falling traffic graph right now: check your trials and signups by channel before you panic. The graph that pays the bills may be telling a different story than the one that scared you.

The checklist I wish we'd had in April

If you are planning to move a large site, this is the list our three months bought:

1. Export every URL that earned a click in the last 12 months from Search Console before you touch anything. That list is the site, as far as search is concerned.
2. Build the page-level redirect map before the move, not after. A domain-level redirect is not a migration plan. Every URL needs a specific destination.
3. Ship the redirects with the migration. Ours arrived three months late, and that gap is where the damage lived.
4. Audit your internal links. A slimmer navigation is a design decision and an authority decision at the same time. Know which links you are deleting.
5. Baseline everything the week before: positions, clicks, and signups per channel. You cannot diagnose what you didn't measure.
6. Watch position and click-through separately. A single traffic number cannot distinguish "we got worse" from "the results page changed." Ours couldn't.
7. Plot it weekly, never monthly. This is the one that cost us most. Monthly averages hid the shape of our own decline and let us argue for months about a cause that the weekly chart would have settled in an afternoon.
8. Put a deadline on "it will settle." Pick a date in advance - six weeks is generous - after which waiting becomes auditing.
9. Hunt for forgotten copies of the old site. We found two complete, crawlable copies of our old site still live on internal subdomains, telling Google the old address was the real one. They had been there the entire time we were repairing the migration. Both are fixed now, and neither would ever have been found without going looking.
10. Separate what you caused from what happened to you. A migration and an algorithm update in the same quarter will blur into one story, and the blurred version leads to the wrong fix. Date each drop before you explain any of them.

Frequently asked questions

Why did my traffic drop after a website migration?

The most common cause is redirects that exist at the domain level but not the page level: the old domain forwards, but individual pages lost their specific successors, and the authority they earned does not transfer. Pruned internal links and leftover crawlable copies of the old site compound it. Check redirects URL by URL, not domain by domain.

How long should traffic take to recover after a migration?

If the page-level redirects shipped with the move, most sites resettle within four to eight weeks. Set that expectation as a deadline in advance. If traffic has not begun recovering by then, stop waiting - audit the redirect map URL by URL, because the most likely cause is structural, and it does not fix itself.

Can rankings improve while traffic falls?

Yes, and it is increasingly common. Our pages ranked fifteen positions better on median year over year while clicks fell by nearly half. AI-generated answers and richer results absorb clicks that used to reach websites, so position measures your seat while clicks measure the size of the room. One caution from our own data: that pattern held year over year, but during the sharpest two weeks of our decline our position got worse too. Run both the long and the short comparison, because they can disagree.

How do I tell whether a migration or an algorithm update caused my drop?

Plot weekly clicks, not monthly, and mark the date of each. A migration shows up as a step at the launch date; an update shows up as a step at the rollout date. Monthly averages blur both into one slope and will let you argue for months. If you see more than one step, you probably have more than one cause.

How do you migrate a website without losing SEO?

Inventory every URL that earned a click in the last year, build a page-level redirect map to specific successors, ship the redirects with the migration rather than after it, preserve your internal links, and baseline positions, clicks and signups per channel before you move. Then set a deadline in advance for when waiting becomes auditing. The full checklist above is the long answer.

Do I need page-level redirects if the whole domain already redirects?

Yes. A domain-level redirect gets visitors to your new homepage. It does not tell search engines where each specific page went, and the authority each page earned over years does not transfer without that mapping. Every URL that ever earned a click needs its own destination.