Usually because the improvement removed whatever was doing the ranking. A rewrite, a redesign and a consolidation each change the text, the links and sometimes the URL at once, so the landing page that ships answers a narrower question than the one it used to win.
What did the rewrite take out?
Words, almost always. The supporting detail that matched long-tail queries: the questions block, the specification table, the paragraph about a use case nobody on the team thought mattered.
A page that has been live for a year does not rank for one term. It picks up hundreds of query variants, and most of them attach to sentences no one wrote deliberately. A conversion-led rewrite treats that text as clutter, because by the standards of a conversion test it is clutter. Cutting it is a real improvement to the page and a real reduction in the surface the page can match. Both things are true at once, which is why this failure keeps happening to teams who did nothing careless.
The check is mechanical. Pull the query list for that URL from the performance report for the eight weeks before the change, sort by impressions, and read the bottom half rather than the top. Those are the queries the deleted paragraphs were answering. If they are gone from the current list, you know what you removed.
Did the page start answering a different question?
Often, yes. A new headline, title and opening paragraph can shift which intent the page matches, and on-page SEO signals that all point somewhere new will move the page to a different set of results.
The common version of this: a page that was built around pricing gets repositioned around the product category, because the category language tested better with sales. The page now competes against category pages with far more authority behind them, and it has stopped being the obvious answer to the pricing question it used to own outright. Nothing is broken. The page is in a harder race.
A mature page earns its traffic across hundreds of queries nobody targeted. An improvement judged only on the primary term can trade away every one of the others and still look like a win in the test.
Did the URL or the links pointing at it change?
This is the expensive version, and it is easy to miss because it lives outside the page. A relaunch that changes the URL without a redirect, or routes it through a chain of redirects, starts the new address from very little.
The same applies to internal links. If the navigation, the related-content block or the footer was rebuilt in the same release, the page may now be reachable only from the sitemap, which makes it an orphan page in practice even though it is still published. External links are worse to recover: off-page SEO signals point at the address they were created for, and if that address no longer resolves cleanly, the page authority built up over years is sitting at a dead end.
Which symptom points at which cause?
| What you see | Usually means | First check |
|---|---|---|
| Impressions hold, clicks fall | Title and description rewritten | Compare the current snippet against the old one |
| Impressions fall across many small queries | Supporting copy was cut | Query list before and after the change |
| One primary term drops, the rest hold | The page was repositioned onto a different intent | What the headline and title now claim |
| Everything drops at once, same week | URL change, redirect problem or lost internal links | URL inspection on the new address |
| Decline arrives over several weeks | Gradual recrawl of a page that got weaker | Last crawl date against the release date |
Is the decline real, or is it the comparison?
Check the window before you accept the number. A relaunch compared against its own launch period, a seasonal peak, or a four-week window that includes a campaign will show a fall that has nothing to do with the rewrite.
The page also needs to be recrawled and re-evaluated before its new version is being judged at all, and that does not happen the day you deploy. For the first stretch after a change, you are looking at a mix of old and new assessments, which is the worst possible moment to draw a conclusion.
Reverting in week one. A partial recrawl is not a verdict, and a revert resets the clock again, so you end up with two changes to untangle instead of one.
Does page experience account for any of it?
Rarely on its own, and it is worth ruling out quickly. A redesign can make a page heavier even while it looks cleaner, and page experience problems hurt the people who arrive more than they move the ranking.
What matters here is the interaction between the two. If the new version is slower and the content is thinner, you will see the click-through hold while the behavior on the page gets worse, and the ranking follows later for reasons that are hard to separate. Fix the weight because it costs you conversions, not because you expect it to restore the traffic.
Do you revert, patch, or wait?
The answer depends on which of the four causes you confirmed, and they are not equally urgent. A redirect fault or a lost internal link is worth fixing today, because nothing recovers while the address is broken. Cut copy can be added back in a form the new design can live with, usually as a questions block or a detail section below the conversion path. A repositioned page is a strategy decision rather than a bug, and the real question is whether the term you walked away from was worth more than the one you moved toward.
So pick one before the next release: restore the address, restore the detail, or accept the trade and say out loud which query you decided to stop winning.