If a page lost rankings, work through the causes in order: confirm the drop is real, check indexation and robots rules, check for a manual action, then compare the SERP, the page, and its links. What follows is a worked diagnostic walkthrough, not a client case study.
SerpInsight is a new business. We don’t have a shelf of client rescues to point at, and we’re not going to invent one just because this post sits in the Case Notes category. So read it as a reasoning exercise: the sequence we’d work through if you handed us a URL that used to sit at position 4 and now sits at 19.
Order matters more than people give it credit for. The instinct is to jump straight to content and backlinks, because those are the interesting problems. Resist it. The cheap checks near the top rule out the expensive explanations further down, and plenty of drops turn out to be dull — a noindex left after a staging push, or a canonical pointing at the wrong URL. Dull is good news. Dull gets fixed in an afternoon.
What actually counts as a page that lost rankings?
A page has genuinely lost rankings when its average position across a group of tracked queries falls and stays down for at least a week, and impressions or clicks fall with it. A two-position wobble on a single day is noise. A ten-day slide from the top three to the bottom of page two, across a whole cluster, is worth investigating.
Pin down the drop date first. Not “sometime in March” — the specific day the line bends. That date is the most useful evidence you have, because nearly every cause below leaves a date-shaped fingerprint: a deploy, a migration, a rollout, a link that disappeared. Without a date you’re guessing. With one, you can line the drop up against your deployment log and CMS revision history, and the culprit often falls out on its own.
Is the drop real, or an artefact of how you’re measuring it?
Verify the drop in Search Console before you trust a rank tracker. Trackers report one location, one device and one language; Search Console reports what real users saw. If Search Console impressions and clicks are flat while your tracker shows a collapse, the problem is the tracker.
There’s a subtler version that catches experienced people out. Average position is a mean. If a page starts picking up impressions for two hundred new long-tail queries where it ranks 40th, the mean drops sharply even though nothing moved for the queries that mattered. Segment before you conclude anything. The other traps are mundane: a tracker retargeted from city to country, a property switched from domain to URL-prefix, or a comparison running 28 days against 7.
Is the page still indexed?
Run the URL through Search Console’s URL Inspection tool and request the live test as well as the indexed version. If the page isn’t indexed, nothing further down this list matters. Indexation is the cheapest check with the highest hit rate, which is why it comes third rather than tenth.
The failure modes repeat themselves. A noindex reinstated by a plugin update. A robots.txt disallow that blocks recrawl, so the ranking decays with no visible error. A canonical pointing at a parameterized duplicate after a template refactor — the sneakiest of the lot, because the page looks perfect in a browser and only the source tells you otherwise. Then the accidental 404 after a slug change, a redirect chain that lost its final hop, and servers returning a 5xx to Googlebot while serving humans normally.
Is there a manual action or a security issue?
Open the Manual actions and Security issues reports. They’re two clicks and they’re either empty or they’re not. A manual action changes the entire remedy: you’ll need cleanup plus a reconsideration request, and no amount of content editing will move the page until it’s lifted.
Manual actions get blamed far more often than they get issued, but partial matches applied to a single directory are easy to miss if you only glance at the summary. Hacked content is the other thing this pair of reports surfaces. If you find an unnatural-links action and you’ve been buying placements, removal or disavowal plus honest disclosure is the route back — one of several reasons we say plainly that paid links carry risk you can reduce but not eliminate.
Did the SERP change rather than the page?
Search the query and look at what’s actually there now. If an AI overview, a video carousel, a local pack or a stack of forum threads has taken the top of the results, your organic position may not have moved at all — the real estate above it did. Position and visibility stopped being the same number years ago.
Intent shift is the harder version. A query that used to return buying guides starts returning product listings because the dominant user behind it changed. Your guide didn’t get worse; it stopped matching. You can spot it by looking at the format of the pages now ranking rather than their domains. If eight of the top ten switched format, the fix is a different page type, not a better version of what you have.
Did somebody change the page?
Pull the page’s revision history and diff it against the version that was ranking. Content edits, redesigns and internal-link removals are the most under-diagnosed cause of a single page sliding, because whoever made the change wasn’t thinking about search when they made it.
We’d look for an H1 moved into an image, an FAQ block shifted into late-rendered JavaScript, a tidy-up that deleted nine hundred words of specifics for being “too long”, a comparison table replaced with a graphic. Internal links deserve their own pass: a page linked from forty posts through a sidebar module, now linked from three, has lost a real signal, and nobody connects a widget swap with a ranking change six weeks later. Keeping old crawls around is half of what a deliberate internal linking approach is for.
Did the page lose external links?
Compare referring domains for that specific URL against a snapshot from before the drop date. Links disappear quietly. Sites get redesigned, old posts get pruned, and a link insertion (also called a niche edit) on a blog that later changes hands can vanish without anyone telling you.
Decay is constant, so the question is never “did we lose links” but “did we lose the ones carrying this page”. A URL resting on four strong referring domains is fragile in a way a URL with sixty isn’t. Migrations make it worse: if the old URL still holds the links and the redirect got flattened to the homepage during a cleanup, that equity is stranded. We wrote more about natural attrition in our piece on how long placements typically stay live.
Are site-wide problems showing up on one page?
Check crawl stats, server response times, Core Web Vitals and whether the site recently published a large volume of thin pages. Site-wide issues rarely sink one URL alone, but they lower the ceiling for everything and turn a close call into a loss.
Realistic candidates: a hosting migration that tripled response time, a bot-management rule challenging Googlebot, or a mass of programmatic location pages published last quarter. Honestly, page speed gets blamed for far more ranking drops than it causes — it’s a tie-breaker between close results, not an explanation for a fifteen-position fall.
Did a core update land on that date?
Line the drop date against the published core and spam update windows. If the slide begins inside a rollout and affects a whole section rather than one URL, you’re looking at a broad re-assessment of the site, not a broken page.
What we wouldn’t do is panic-edit during a rollout. Positions move around inside them, and changes made mid-flight tell you nothing because you can’t attribute the outcome. Wait for the announced completion, then group the losing pages and look for what they share. Recovery usually arrives with a later update rather than a fortnight after your edits, and it isn’t guaranteed at all.
Which symptom points to which cause?
Most drops fit one of a small number of shapes, and the shape tells you where to look first.
| Symptom pattern | Most likely cause | How to confirm | Fix |
|---|---|---|---|
| Clicks and impressions both hit zero overnight | Deindexed — noindex, robots block, 404 or bad canonical | URL Inspection, live test plus indexed version | Remove the block, resubmit, watch for recrawl |
| Tracker shows collapse, Search Console flat | Measurement artefact — tracker location, device or a new SERP feature | Filter Search Console to the same query set and dates | Fix the tracker config; nothing to fix on the page |
| One URL falls, site steady, date matches a deploy | The page changed — content, template or internal links | Diff against Wayback and CMS revisions | Restore what was removed, then re-request indexing |
| Position holds but clicks fall | A SERP feature or heavier ad load absorbing the clicks | Check the live SERP and CTR by position | Rewrite title and description; target the feature |
| A whole section slides over three to ten days | Core update re-assessment | Match the drop date to published rollout windows | Rebuild the weakest pages; expect slow recovery |
| Slow decline across six to twelve months | Content gone stale while competitors refreshed theirs | Compare against the current top three for depth and recency | Rewrite properly rather than tweaking dates |
| Sitewide fall with a Search Console message | Manual action or a hack | Manual actions and Security issues reports | Clean up fully, then file a reconsideration request |
| Drop begins the week of a migration | Redirect or canonical errors; links stranded on old URLs | Crawl the historical URL list and check every target | Repair redirects one-to-one to the replacement pages |
What does the full checklist look like, in order?
Run these top to bottom and stop when you find something. Working out of order is how people spend three weeks rewriting a page that was blocked in robots.txt the whole time.
- Pin down the exact drop date from Search Console, not from memory.
- Confirm the drop for a fixed query set, not an averaged dashboard number.
- Check indexation with the URL Inspection live test, and read the raw source for noindex, canonical and hreflang errors.
- Check robots.txt and the server response for Googlebot’s user agent.
- Open Manual actions and Security issues.
- Look at the live SERP: new features, new competitors, format shifts.
- Diff the page against the version that was ranking.
- Count internal links to the page and compare with an older crawl.
- Compare referring domains to that URL before and after the drop date.
- Verify redirects from historical URLs resolve one-to-one.
- Review crawl stats, response times and recent bulk publishing.
- Line the drop date up against published update windows.
- Only then compare your page, honestly, against the pages that replaced it.
Nothing here is a diagnosis of a real site
This page is method, not measurement. We haven’t attached outcome numbers to it because we don’t have a client drop to report, and inventing one would make the rest worthless.
What if someone simply wrote a better page?
Sometimes there’s no bug. Indexation is clean, no manual action, no update, no lost links, the page hasn’t been touched in a year — and two new URLs above you are genuinely more useful. That’s the diagnosis. The fix is to make a better page, and no shortcut hides behind it.
“Better” is worth defining, because it usually gets treated as a vibe. It means answering more of the sub-questions the query implies, using specifics where yours uses generalities, showing something a competitor can’t copy from a template, and getting to the point faster. Open the two pages side by side and ask which one you’d send to a friend. If it isn’t yours, you have your answer. If you’d rather someone else ran this sequence and handed you the findings in order, that’s what a structured technical and content review is for.
Key takeaways
- Establish the exact drop date first — nearly every cause leaves a date-shaped fingerprint.
- Confirm the drop in Search Console against a fixed query set before trusting a rank tracker.
- Indexation, robots rules and canonicals explain many sudden drops and take minutes to rule out.
- A falling average position with flat clicks often means new long-tail impressions, not a real loss.
- Content edits, redesigns and removed internal links are the most under-diagnosed single-page cause.
- Section-wide slides matching a rollout window are re-assessments; don’t panic-edit mid-rollout.
- Sometimes a competitor published something better, and the only fix is a genuinely better page.