Traffic drops
Why your Google rankings dropped.
The instinct is to start changing things. Almost every expensive recovery we have been called into began with a fortnight of changes made before anybody knew what had happened.
Why did my Google rankings drop?
A ranking drop has a small number of possible causes, and Google publishes the order to check them: a technical problem stopping pages being served, crawled or indexed; a manual action or a security issue; a change somebody made to the site; a core update or other change to Google's systems; a change to the results page itself, such as an AI summary absorbing the clicks; or simply less demand for what you sell this month.
Diagnose in that order using the Search Console performance report, comparing date ranges and segmenting by search type, country, device, query and page, because the segment where the loss is concentrated usually names the cause.
Before you change anything
A drop is information. Changing five things at once destroys that information, because afterwards you cannot tell which change helped, which hurt, and whether anything you did mattered at all.
Give yourself one working session to diagnose before touching the site. Most causes announce themselves within twenty minutes if you look in the right order, and the ones that do not are usually the ones where patience is the correct response anyway.
Two practical notes on the data. Search Console reports lag by a couple of days, so the last two bars of any graph are always incomplete and always look like a decline. And a drop in rankings, a drop in traffic and a drop in enquiries are three different events. Confirm which one you are actually looking at before you decide what caused it.
The order below follows Google's own published guidance for debugging search traffic drops. It is worth using precisely because it is theirs: it rules out the cheap causes before you spend anything on the expensive ones.
The diagnostic order
Each step either finds your cause or eliminates a whole category. Do not skip ahead to the interesting explanations, because the boring ones are more often correct.
Confirm the drop is real
Open the Search Console performance report and compare the last 28 days against the previous 28, not against a single good week. Then look at the same period last year. A shape that repeats annually is seasonality, not a fault. Exclude the final two days, which are incomplete, and check whether anything changed in your analytics setup, since a broken tag produces a graph identical to a catastrophe.
Check the site is reachable and indexable
This is the category that produces the sharpest, most alarming graphs and has the cheapest fixes. Use URL Inspection on a page that has lost visibility and read what Google says about it. Check robots.txt, check for a noindex tag left behind by a deployment, check your server for errors and check that the certificate has not expired.
- A staging noindex pushed live is the single most common cause of a sudden total loss
- A robots.txt file overwritten during a deployment blocks crawling of whole directories
- Server errors or timeouts during a crawl cause pages to be dropped over the following days
Check the manual actions and security reports
Both live in Search Console and both are usually empty. A manual action means a human reviewer has applied a penalty and tells you which policy was breached. A security issue usually means the site has been compromised, which is common on neglected content management systems and which suppresses everything until the site is cleaned and a review requested.
See the remaining steps: The diagnostic orderHide the remaining steps: The diagnostic order
Find out whether anything changed on the site
Ask everyone with access, then check independently. Redesigns, migrations, plugin updates, a content management system upgrade, a new cookie banner, a developer tidying up redirects, or a page somebody deleted because it looked untidy. A migration that lost its redirect map is the most damaging version of this and the most recoverable, provided somebody still has the old URL list.
Segment the loss to locate it
This is the step that does the most work and the one people skip. In the performance report, break the same period down by search type, by country, by device, by query and by page. A loss confined to one country, one device or one section of the site points somewhere completely different from a loss spread evenly across everything.
- Site-wide and even suggests a technical cause or a systems change
- Confined to one section suggests a change to that section or its content quality
- Confined to mobile suggests a rendering, speed or layout problem on phones
- Confined to image or video search suggests a change to how that surface is displayed
Separate lost position from lost clicks
Compare impressions with clicks and average position. If impressions and position hold steady while clicks fall, you did not move: the results page changed around you. An AI summary, a new ad block, a larger map pack or a product carousel can absorb the click without touching your ranking. That is a real loss, but the response is different and no amount of technical work fixes it.
Check whether demand itself moved
Search interest rises and falls, and some of it is nothing to do with you. Compare your query-level impressions with the same period last year, and check whether the terms losing volume are seasonal. A roofing business in a mild winter and a tax practice in July are looking at the calendar, not at a penalty.
Only then consider a systems change
If everything above is clean and the loss is broad, gradual and content-related, a core update or another change to Google's systems is the remaining explanation. There is no switch to flip. Google's guidance is consistent and unglamorous: there is no single fix, and the route back is making the content genuinely better against its own published questions about helpfulness and experience. Recovery is measured in months and sometimes waits on a subsequent update.
Write down what you found, what you changed and on which date. The next drop, and there will be one, is far easier to diagnose against a written history.
Where a drop can enter the system
Three stages, and a failure at any one of them looks the same from the outside.
Most alarming drops are stage-one or stage-two failures: the page stopped being reachable, or it stopped being kept. Those are the ones that appear overnight as a cliff on the graph and are usually fixed in an afternoon once identified.
Stage-three losses look different. They are gradual, partial, and concentrated in particular queries or sections. If the graph slopes rather than falls, stop checking the server and start reading the pages that lost ground against the pages that took their place.
Five shapes, and what each one usually means
The shape of the line narrows the cause faster than any tool. Match yours to one of these before going further.
A cliff, overnight, across everything
Almost always technical or a manual action. Something was deployed, blocked, deleted or penalised on a specific day. Find the date on the graph, then find what happened on that date. Deployment logs and change history are more useful here than any SEO tool.
A slope, over several weeks, across many queries
Usually a systems change or accumulated competitive loss. Check the dates against known core update windows, then read the pages now ranking above you. This is the slow, unsatisfying category and the one where reacting quickly does the most damage.
One section falls, the rest holds
Something specific to that section: a template change, a batch of thin pages published, a set of URLs that moved, or a content quality judgement applied to one topic. The rest of the site holding steady is itself the diagnosis, because it rules out everything site-wide.
Impressions hold, clicks fall
You did not move, the results page did. An AI summary, more ads, a bigger map block or a feature that answers the question without a visit. Confirm by searching the affected queries in a signed-out browser and looking at what now sits above you.
One country or one device only
Points at something specific rather than something general: a geotargeting or hreflang change, a localised competitor, a rendering fault on mobile, or a speed regression that only affects phones. Segment first, then test on the affected device rather than on your desktop.
Three different things people call a ranking drop
Naming which one you have decides who should be working on it.
| What actually fell | Where you see it | Who fixes it |
|---|---|---|
| Position for specific queries | Average position in Search Console | SEO and content |
| Clicks, with position unchanged | Clicks against impressions | Content, titles and ad strategy |
| Sessions, with clicks unchanged | Analytics against Search Console | Tracking and tag configuration |
| Enquiries, with sessions unchanged | Form and call volume | The website and the follow-up process |
| Revenue, with enquiries unchanged | Your own sales records | Sales, pricing and capacity |
Four of those five have nothing to do with Google. Working out which row you are in takes ten minutes and saves a month of the wrong work.
Sometimes you did not move, the page moved
This category of loss is real, is not your fault, and cannot be fixed by improving the page. What it can be answered with is breadth: presence in the map pack, presence in the AI answer where the question is one you genuinely answer, paid coverage where the economics work, and content aimed at searches that still produce clicks because they need a decision rather than a fact.
It also changes what you should be reporting. If your measurement stops at sessions, this shows up as failure. If it reaches calls and booked work, you can see whether the business actually lost anything.
What not to do in the first fortnight
Every item here is something we have been asked to undo. None of them helped, and several made the recovery slower.
- Do not rewrite or delete large numbers of pages before you know what happened
- Do not change your title tags site-wide as a reaction
- Do not disavow links reflexively, since Google states most sites never need to and a careless file removes good links
- Do not migrate platform or redesign while you are still diagnosing
- Do not add structured data in the hope it counts as a fix, because it does not raise rankings
See the full checklist: What not to do in the first fortnightHide the full checklist: What not to do in the first fortnight
- Do not buy links to recover, which is the behaviour Google's spam policies were written for
- Do not judge a recovery on three days of data, or on the two incomplete days at the end of the report
- Do not fire the agency before you have the written diagnosis, because you will need it either way
Patience is an actual tactic here. Some drops reverse on their own when a change is refined or rolled back, and a site that has been rewritten twice in the meantime cannot tell.
Related answers
The questions that usually follow a drop.
Not showing up at all
Separating an indexing failure from a competitive one from a profile problem.
Not showing up at allWhat technical SEO is
The crawling, indexing and serving layer where the fastest fixes usually live.
What technical SEO isHow long SEO takes
Realistic recovery timelines, and why months is the honest unit.
How long SEO takesRanking number one
Why positions move constantly, and what can honestly be promised about them.
Ranking number oneThe full local SEO guide
The programme a local business is defending when visibility moves.
The full local SEO guideLost visibility and need the diagnosis rather than a sales pitch?
We run Google's own diagnostic order across your site and send you what we find, in writing.
Straight answers.
How long should I wait before acting on a drop?
Diagnose immediately, act deliberately. Technical faults should be fixed the moment they are found, because every day of a noindex tag costs you more. Content and quality responses should wait until you have a segmented picture, which usually means a couple of weeks of data.
The two incomplete days at the end of every Search Console report are not a trend. Neither is a single bad Tuesday.
How do I know whether it was a core update?
Compare the dates on your graph against Google's announced update windows, and check whether the loss is broad, gradual and spread across many queries rather than sharp and site-wide. Core update losses look like a slope, not a cliff.
If it was one, Google's own position is that there is no specific fix and no penalty to lift. The route back is making the content genuinely more useful against its published questions about helpfulness, experience and originality, and recovery can take months.
We redesigned the site and traffic fell. What now?
Find the old URL list, from a backup, an old sitemap, an analytics export or an archive crawl, and check every one for a redirect to the right new page. Missing and incorrect redirects are the leading cause of post-migration losses, and they are fully recoverable.
Then check that the new templates did not drop headings, body content, internal links or structured data that the old pages carried. A cleaner design frequently ships with less content on the page, and that alone can move rankings.
Can a competitor cause my rankings to drop?
By publishing something better, yes, constantly and legitimately. By pointing spam links at you, very rarely: Google states its systems ignore the large majority of low-quality links without any action from you, and the disavow tool is intended for cases where you know links were bought or you have a manual action.
A careless disavow file that removes genuine links does more harm than the spam it was aimed at. Treat it as a last resort with evidence, not a first response to anxiety.
My rankings look the same but traffic is down. What happened?
Look at impressions against clicks. If both position and impressions held while clicks fell, the results page changed around you rather than under you: an AI summary, more ads, a larger map block or a feature that answers the query without a visit.
Also check your analytics tag, your consent banner and any recent cookie or tracking change. A measurement fault produces a graph that looks exactly like a traffic collapse, and it is a surprisingly common explanation.
Where this comes from.
Primary documentation and published research behind the guidance on this page.
- Google Search Central: Debugging drops in Google Search traffic (opens in a new tab)Google's decision tree for diagnosing a ranking or traffic loss.
- Google Search Console Help: About Search Console (opens in a new tab)
- Google Search Central: Creating helpful, reliable, people-first content (opens in a new tab)Google's published self-assessment questions for content quality and experience.
- Google Search Central: Spam policies for Google web search (opens in a new tab)Includes the link spam policy that covers bought and exchanged links.
Talk to the team
A short call, a look at how the business currently shows up, and a straight answer on what we would do first.