How to Tell if Your Traffic Drop Is Caused by a Google Update
An update is the explanation everybody reaches for and one of the less common causes. This guide rules out the ordinary reasons first, then gives you a real method for the update question, including what to do while a rollout is still running.
It Probably Was Not An Update
Most people arriving here have already decided an algorithm update took their traffic. Frequently it did not. Checking the less exciting explanations first costs an hour and saves the months that businesses lose rebuilding sites in response to something that never happened.
Why an update takes the blame. It has a name.
An update is the only explanation with a date, a label and an industry writing about it. A shorter month has none of that, so it never gets suspected even when it is the whole answer.
Why the order of checking matters. Cost.
If an update caused it, the response is slow improvement over months. If a developer blocked a section of your site, the response is a ten minute fix. Assuming the expensive answer first is how the cheap one goes unnoticed.
What this page owns. One question only.
Whether an update caused the fall. Whether the fall is real in the first place belongs to our material on measuring performance, whether it is a manual action belongs to the risky practices material and whether a site change caused it belongs to the migrations material. Each is a separate question with a separate answer.
How to use what follows. In order.
The blocks run from the cheapest check to the most involved. Work down them rather than jumping to the timing comparison, which is where everybody wants to start.
Rule Out The Ordinary Explanations First
Five things account for a great many reported drops. None of them is an update. Our material on measuring performance owns this properly, since the prior question is whether the drop is real at all, so treat this as the short version.
A shorter month. The commonest of all.
February against January is three days lighter before anything else happens. Comparing calendar months rather than equal periods manufactures a decline that does not exist.
Seasonality. Compare against last year.
A roofer in July and a roofer in November are different businesses as far as search demand goes. The comparison that tells you something is the same period twelve months earlier.
A tracking change. Nothing fell except the recording.
A tag removed during a redesign, a consent banner changed, a filter added to a report or a property switched over. The traffic carried on. The measurement stopped.
A site change nobody mentioned. Ask before you investigate.
A developer, a plugin update or a hosting move can take pages out of the index without anybody connecting the two. This is worth ruling out before every other test on this page. Our migrations material covers it in depth.
A competitor improving. Nothing happened to you.
Somebody published a better answer, earned coverage or fixed a site that was holding them back. Positions are comparative, so their gain shows up in your reporting as a loss.
Then Check The Timing
An update driven fall begins on a date. Finding that date and comparing it against the published record of rollouts is the first real test in this investigation. It is a great deal more reliable than anybody's impression of when things felt wrong.
Find the date first. Before you look at any update record.
Work out when the decline started from your own data rather than deciding which rollout you were affected by and looking for evidence. Doing it the other way round produces the answer you went in with.
Where the published dates live. Two places.
Google posts named ranking rollouts to its Search Status Dashboard as they begin and marks them when they finish. It also maintains a ranking release history page recording them. Those are the only dates worth building an argument on.
How to line them up. Annotate the chart.
Mark the rollout start and completion dates on your own reporting. A fall that begins within a day or two of a rollout starting is a strong indication. A fall that began three weeks earlier is somebody else's problem being borrowed.
What a match does not prove. Cause.
Plenty of things happen on the same day as a rollout, including things you did. A date match makes an update the leading candidate rather than the verdict, which is why the remaining tests exist.
Gradual Beats Sudden
How the fall looks on a chart is one of the cheapest and most informative tests available. A broad reassessment is applied progressively, so it tends to produce a slope. A single day collapse points somewhere else entirely.
What a rollout looks like. A slope with noise in it.
Positions move in both directions across the period, so the decline arrives in steps rather than at once. Some pages drop then partially recover before the thing has even finished.
What a cliff looks like. Technical.
Traffic that goes from normal to almost nothing inside a day is usually a site problem. Pages blocked from indexing, a canonical pointing at the wrong place, a redirect gone wrong, an expired certificate or a server that stopped responding to crawlers.
What a staircase looks like. More than one cause.
Two distinct steps a fortnight apart usually means two separate events. The Search Status Dashboard recorded a spam rollout beginning on 24 March 2026 and a core rollout beginning on 27 March, which is exactly the situation where confident attribution stops being possible.
Why this test comes early. It takes two minutes.
Look at the graph before you read a single industry article. The shape will tell you which half of this page you should be reading.
What Volatility Measures Actually Are
Several companies publish a daily score describing how much search results are moving. When one of those scores spikes, the industry reports an update. Understanding what those measures can and cannot tell you is worth more than watching them.
What they do. Track movement across large keyword sets.
Positions for tens of thousands of searches are recorded daily and the amount of change is expressed as a single figure. A high figure means results moved a lot that day across whatever was being tracked.
What a spike is evidence of. Something broad happened.
It is reasonable evidence that a change was applied across many results, which is genuinely useful when Google has said nothing. It is inference drawn from observation rather than an announcement.
What they cannot tell you. Anything about your site.
The keywords being tracked are not your keywords and the sites being tracked are not your site. A measure showing extreme movement is compatible with your own positions not having shifted at all.
Why we describe these generically. House standard.
We do not name measurement products anywhere on this site, partly because the category matters more than any particular one and partly because naming tools tends to substitute for explaining the method. What follows is what to do with the signal.
What To Do With That Signal
The useful question is whether the movement is industry wide or yours alone. That single distinction sends the investigation in one of two completely different directions. It is the most valuable thing any of these measures can contribute.
How to check it yourself. Look at your competitors.
Take ten searches that matter to your business and record who ranks for them now. If the entire set has reshuffled, something broad happened. If everybody else sits exactly where they were and only you moved, nothing broad happened to you.
If everybody moved. You are in a reassessment.
The work becomes comparative rather than corrective, which is what recovering from an algorithm update covers. There is no fault to find, so looking for one wastes the effort.
If only you moved. Look at your own site.
Something specific happened, which points at a technical change, a content change, a manual action or a link problem. Those live in the migrations and risky practices material rather than here.
Why this order beats reading commentary. It is about you.
Ten minutes checking your own searches tells you more than a fortnight of industry articles about an update that may never have touched your sector.
Did It Affect Your Whole Site Or One Part
Where the fall sits within your site is the most diagnostic thing you can look at. A decline spread across everything and a decline confined to one corner have almost nothing in common. Telling them apart takes one report.
Broad and even. Points at a core update.
Service pages, guides, location pages and the home page all softening together is the signature of a reassessment applied across the board. Nothing was singled out because nothing needed to be.
Confined to one section. Points at something specific.
A fall living entirely in one directory, one page type or one group of searches is usually a change to those pages, a technical fault in that part of the site or an update addressing one particular behaviour.
How to slice it. Four ways.
Compare the period before and after by page, by search term, by device and by country. Search Console will do all four. The one that splits cleanly into winners and losers is your answer.
A single page falling. Almost never an update.
One page losing everything while the rest of the site holds is a page level cause. Something changed on it, something changed pointing at it or a better answer arrived for the search it served.
Check What Replaced You
Looking at what now ranks where you used to tells you what the reassessment rewarded. It is more informative than any measurement product and it costs nothing beyond the willingness to read a competitor's page properly.
Why nobody does it. It is uncomfortable and slow.
Reading a chart takes seconds. Reading five competitor pages and admitting what they do better takes an afternoon. The afternoon is where the actual answer is.
How to do it. Ten searches, written down.
Take the searches you lost, note who occupies the top results now and record what those pages have that yours does not. Depth, first hand knowledge, a named author, evidence, better structure or simply a page built for that exact question.
What you are looking for. A pattern across the winners.
One competitor beating you is chance. All five sharing a characteristic your pages lack is the change described in something you can act on.
Why this beats a site audit. Different question.
An audit tells you what is technically imperfect on your site. A reassessment was not about that. Comparison against what won is the only reliable guide to what actually moved.
Confirmed Or Not
Many changes are never confirmed by Google. An unconfirmed one is still a real change and waiting to be told officially before you respond wastes weeks that your own data has already accounted for.
What confirmation adds. Certainty about the cause.
A named rollout on the dashboard gives you dates you can rely on. That is genuinely valuable for the timing test in block three and it is the only thing confirmation contributes.
What it does not add. Anything about your site.
Confirmation that a rollout happened says nothing about whether it touched you. Businesses regularly attribute a fall to a confirmed rollout their sector was never involved in.
Why most changes go unmentioned. They are not events.
Google's documentation on core updates states that smaller ones happen continually and are not announced because they are not widely noticeable. How often Google updates its algorithm covers what that means for a business.
What to do about the uncertainty. Record everything.
Export your position and click data, note the dates and keep a record of what the site looked like. That record turns an unconfirmed change into something you can still investigate properly later.
Then Do Nothing For A While
If a rollout is still in progress, stop. Positions move throughout one, so any measurement taken during it describes a state that will not exist by the end and any change you make during it can never be assessed.
What Google publishes about timing. An estimate.
Named rollouts usually arrive with an expected period attached, described as an estimate rather than a commitment. Some finish sooner and some run longer, which is why we never quote a duration as fact.
How to know it has settled. The dashboard says so.
Google marks a rollout complete and updates the ranking release history. Give it a settled period after that, then compare a full stable stretch against a full stable stretch from before.
What the panic response costs. The evidence.
Pages deleted, content rewritten and links disavowed while the ground is moving leaves you unable to tell what the update did and what you did. That is the commonest self-inflicted damage in this whole subject.
What to do meanwhile. Prepare rather than react.
Keep working on improvements you had already planned, gather the comparison material and read what recovery genuinely involves before you commit to anything. The full series sits on the Google algorithm updates guide.
Work out what
happened before
you start
fixing it.
Most of the damage done after a traffic drop is done in the first fortnight, by businesses acting on a diagnosis nobody had actually made. We would rather tell you it was a shorter month than sell you a recovery project you do not need.
How we work through one of these:
If the answer turns out to be something we do not handle, we will tell you that instead.
Every guide.
One position.
How updates work, the history of what each one changed, whether an update caused your drop, what recovery genuinely involves, what helpful content means now and how to build a site that rollouts leave alone.
