In short
Work in order: confirm the drop is real rather than a tracking artefact, establish exactly when it started, then determine whether it is technical, algorithmic, seasonal, or competitive. The start date is the most informative single fact — a sharp drop on one day points to a technical change or a manual action, while a gradual slide over weeks points to an algorithm update or competition.
First, confirm it is real
A surprising share of traffic emergencies are measurement problems.
Did the tracking break? A deploy that dropped the analytics snippet, a consent banner change, a tag manager container that stopped firing. Check whether the drop appears in server logs or Search Console as well — if only one source shows it, suspect the source.
Are you comparing the right periods? Month-to-date against last full month always looks like a decline. So does any week containing a holiday.
Did filtering change? A new bot filter or an IP exclusion can remove traffic that was never real, which looks alarming and is an improvement.
Rule this out first. It takes ten minutes and prevents days of investigating nothing.
Establish exactly when
The start date is the most informative fact available. Narrow it to the day.
A cliff on a single day points to something that happened that day: a deploy, a DNS change, a robots.txt edit, a manual action, or a broad algorithm update.
A steady decline over weeks points to something gradual: competitors improving, content ageing, or a slow reassessment of the site.
A drop confined to one section points at that section, and immediately narrows the search.
Then line the date up against your own change log and against known algorithm update dates. Correlation is not proof, but a drop beginning on the day you deployed is worth examining first.
Segment before theorising
A total is uninformative. Break it down.
By channel. Organic, direct, referral, paid, social. A drop confined to one channel is a much smaller problem than a drop across all of them — which usually means tracking or a site-wide technical fault.
By page. Is it everything, one section, or a handful of pages that used to carry disproportionate traffic?
By device. Mobile-only drops point at mobile rendering, speed, or a layout change.
By country. Sudden geographic changes can indicate a redirect rule or a hosting problem.
By query. In Search Console, compare periods and sort by impression change. Losing impressions means visibility fell; holding impressions while losing clicks means the listing changed — a title rewrite, a new AI Overview, a competitor with better snippets.
That last distinction matters and is frequently missed: impressions steady with clicks down is a different problem from impressions down, and needs a different fix.
The technical causes, checked in order
These are the fastest to rule in or out, so do them first.
Accidental noindex. A staging setting deployed to production. Check the meta tag and the HTTP header on your important pages.
robots.txt blocking. One careless line can remove a section.
Broken redirects after a change. Redirect chains, loops, or redirects to the wrong target.
Server errors. Sustained 5xx during crawling causes real drops.
Canonical mistakes. Every page canonicalising to the homepage will deindex a site quietly.
Certificate or DNS problems. Brief for users, longer-lasting for crawlers.
A speed regression. A heavy script added site-wide can move rankings and conversion together.
Search Console's page indexing report shows most of these directly, and comparing it against a month ago usually identifies the cause in minutes.
Manual actions and algorithm updates
Check for a manual action in Search Console. Explicit, rare, and if present it is the whole answer.
Check for an algorithm update around your start date. If a broad update landed, and the drop is spread across the site rather than concentrated, that is likely the cause.
Algorithm-related drops are not fixed by tactics. They reflect a reassessment of quality relative to competitors, and recovery comes from making the site genuinely better — which is slow, and usually confirmed only at a subsequent update. Anyone offering a quick recovery is selling something.
If your drop coincides with a large volume of similar pages published shortly before, consider whether scaled content is the underlying issue; site-level quality assessment means weak pages can drag down strong ones.
The non-emergencies
Seasonality. Compare to the same period last year, not last month.
A one-off spike expiring. A post that did well is not a permanent baseline.
Losing a single big referrer. One link that was sending most of your traffic is a concentration risk, not an SEO problem.
Competitors improving. If your rankings are stable and traffic fell, the results page around you may have changed — new AI Overviews, more ads, a competitor with a better snippet.
What to do about it
Once you have the cause, the response follows: fix technical faults immediately; wait and improve after an algorithm update; do nothing for seasonality; and for competitive erosion, improve the specific pages that lost ground rather than the site in general.
What not to do: change many things at once. If you fix six things simultaneously and traffic recovers, you have learned nothing and cannot repeat it. Change deliberately and record what you changed and when.
If traffic has dropped and the cause is not obvious after the checks above, book a call.
Common questions
How do I know if a traffic drop is real?
Check whether it appears in more than one source — analytics, server logs, and Search Console. If only one shows it, suspect that source: a missing analytics snippet, a consent banner change, or a new bot filter can all look like a collapse. Also confirm you are not comparing a partial period against a complete one.
What is the fastest way to diagnose a traffic drop?
Establish the exact start date, then segment by channel, page, device and query. A cliff on one day points to a technical change or manual action; a gradual slide points to an algorithm update or competition. A drop in one channel is a far smaller problem than one across all of them.
What's the difference between losing impressions and losing clicks?
Losing impressions means your visibility fell — you are appearing less often. Holding impressions while losing clicks means you still appear but the listing changed: a rewritten title, a new AI Overview above you, or a competitor with a better snippet. They need entirely different fixes.
What technical issues cause sudden traffic drops?
An accidental noindex deployed from staging, a robots.txt line blocking a section, broken or looping redirects after a migration, sustained server errors during crawling, canonical tags pointing everything at the homepage, DNS or certificate problems, and site-wide speed regressions. Search Console's page indexing report surfaces most of them.
How do you recover from an algorithm update?
Slowly, and by making the site genuinely better rather than by applying tactics. Algorithm drops reflect a reassessment of quality relative to competitors, and recovery is usually confirmed only at a subsequent update. If the drop followed publishing a large volume of similar pages, thin templated content may be the underlying cause.
