- First establish whether it is a real drop or a measurement problem. Half of them are measurement.
- The shape of the fall tells you the cause: cliff, slide or seasonal dip.
- Check your own changes before blaming an algorithm update.
- Do not make five changes at once. You will never know which one worked.
Traffic falls are alarming and usually diagnosable within an hour if you work through the possibilities in the right order. The order matters more than the tools: most wasted diagnostic time is spent on sophisticated analysis of a drop that turned out to be a broken tracking tag or last year’s August.
Step 1: is it real?
Check Search Console clicks alongside analytics sessions. If Search Console is stable and analytics fell, the problem is tracking: a broken tag, a consent banner change, a new plugin. This is the single most common cause of a "drop" and it costs nothing but the check.
| Search Console | Analytics | What it means |
|---|---|---|
| Stable | Fell | Tracking problem. Nothing is wrong with the site |
| Fell | Fell | A real drop. Continue to step two |
| Fell | Stable | Usually a Search Console property or filter change, not a site problem |
| Impressions stable, clicks fell | Fell | You still rank but fewer people click: a new feature, an AI answer, or a competitor with better titles |
Step 2: what shape is it?
A cliff overnight points to something technical or a manual action: a noindex deployed, robots.txt changed, the site down during a crawl, a migration gone wrong, or a penalty visible in Search Console.
A slide over weeks points to competition, content decay or an algorithm update rolling out.
A dip matching last year is seasonality. Compare year on year before doing anything.
| Shape | Most likely causes | Where to look first |
|---|---|---|
| Vertical fall, one day | noindex, robots.txt, outage, migration, manual action | Search Console messages, then the robots meta on a live URL |
| Steep fall over 3-5 days | A core update rolling out, or a bulk redirect gone wrong | Update timing, then the redirect map |
| Slide over 4-8 weeks | Content decay, a competitor improving, gradual link loss | Query comparison, page by page |
| Sawtooth, recovering each time | Server errors or crawl budget, not ranking | Response times and 5xx errors during the dips |
| Same dip as last year | Seasonality | Year-on-year comparison, then stop |
Step 3: check what you changed
Pull the change log for the two weeks before the drop: deployments, plugin updates, theme changes, content edits, redirects, new tags. Most sudden falls are self-inflicted and recent. Rolling back is often faster than diagnosing further.
The most common single cause of a vertical fall. Check the robots meta tag and the X-Robots-Tag header on a live URL, not on staging.
The rankings belonged to the old URLs. If they now 404, the fall is permanent until the redirects are in place.
Treated as a soft 404 and worse than leaving the pages gone. Redirects must go to the closest equivalent page, one hop each.
Quiet, gradual, and easy to miss: pages remain live but consolidate onto the wrong URL. Spot-check canonicals on each template.
Step 4: check the technical basics
- Pages report in Search Console: did a group of URLs move into not-indexed?
- robots.txt and meta robots on the affected templates.
- Canonicals pointing somewhere unexpected.
- Server response times and errors during the period.
- Redirects: chains, loops, or a bulk redirect to the homepage.
- hreflang, if the site is multilingual: a broken pair can swap which language ranks.
- Sitemaps: are they still listing the URLs that matter, and returning 200?
Step 5: narrow it down
In Search Console, compare the periods before and after by page and by query. Where the loss is concentrated names the cause:
| The loss is concentrated on | Most likely cause | The response |
|---|---|---|
| One template or directory | Technical: indexing, canonical, or speed on that template | Fix the template, then request validation |
| Informational pages, evenly | A core update, or answers now shown in the results | Improve depth and first-hand value on the best of them |
| One query where a rival now ranks | Competition | Improve that page. Nothing else needs changing |
| Branded queries | Reputation, a name change, or a rival bidding on the brand | Not an SEO problem in the usual sense |
| Everything, evenly | A site-wide technical change or a manual action | Search Console messages first, then the change log |
“A drop concentrated on one template is a technical problem. A drop spread evenly across every page is something you did to the whole site. The distribution tells you more than any tool will.”
Step 6: act narrowly
Fix the single most likely cause, document it, and wait two weeks. Changing five things at once means you learn nothing and risk making it worse. If the cause is an algorithm update, the response is improving the affected pages rather than chasing rumours about what the update targeted.
Before touching anything. A hypothesis you recorded is falsifiable; one you remember is not.
The single most likely cause, on the affected pages only.
You will need it to read the recovery curve four weeks from now.
Recovery from technical fixes takes days to weeks. Impatience here is what turns one problem into three.
What not to do first
Four responses are common, expensive, and almost never the right opening move:
Without a manual action for unnatural links, this removes value more often than it recovers it.
A site-wide rewrite destroys the evidence. You will never know what the cause was.
If existing pages stopped performing, more pages of the same kind will not fix it.
Reasonable eventually, but it resets the change log at exactly the moment you need it.
Frequently asked questions
How long does recovery take?
Technical fixes, days to weeks. Quality-related declines, a quarter or more, and only with genuine improvement.
Should I disavow links after a drop?
Almost never, unless there is a manual action for unnatural links. Disavowing good links by mistake makes things worse.
Is a 10% fluctuation worth investigating?
No. Week-to-week noise of that size is normal. Investigate sustained changes over four weeks or sudden falls of a third or more.
Impressions held but clicks fell. What happened?
You still rank and fewer people click. Usually a new feature above you in the results, an AI answer covering the question, or a competitor with a better title. Rewrite titles and descriptions for the affected queries before assuming a ranking problem.
Traffic fell right after a core update was announced. Is that the cause?
Check your own change log first. Updates are announced often enough that the timing coincides with ordinary site changes more often than people expect.
Do we need to tell Google we fixed it?
For indexing problems, use the validation button in the Search Console report - it prompts a recrawl. For everything else, no: the fix is discovered on the next crawl.
Traffic fallen and no explanation?
We diagnose in this order and tell you what happened, whether or not it is something we can fix.
Request a diagnosisRelated: SEO services.