Why did my website traffic suddenly drop, and who fixes it?

A sudden fall in organic traffic almost always traces to one of six causes. Naming which one you have is most of the battle, because each points to a different fix and a different person to call. The six:

• Bad data: the drop is not real; your tracking or analytics broke.
• A technical fault: pages dropped out of Google’s index.
• A manual action: a human reviewer at Google penalized your site.
• An algorithm update: Google changed how it ranks.
• A migration or redesign: a site change quietly damaged something.
• Falling demand: fewer people are searching for your topic.

That step matters more than it sounds. All six causes produce the same shape on a graph, and nothing on the screen tells a fault costing an afternoon from one costing a year. Say you spend a whole quarter improving content when a plugin quietly blocked your pages from being indexed. That quarter is wasted, and this mix-up happens often.

The expensive mistake is rarely the drop itself. It is what you buy during the panicked fortnight that follows.

Before anything else, confirm the drop is real

Verify the measurement before paying anyone to repair the site. A large share of reported drops turn out to be reporting faults rather than traffic faults, and every one of them costs nothing to rule out. If your analytics has stopped counting visitors who are still arriving, no amount of search work will move the number.

Four checks settle the question, and the order matters.

1. Compare two systems, not one. Look at Search Console impressions and clicks beside your analytics sessions across the same dates. If Search Console looks normal while analytics collapsed, the counting is the problem, not the traffic.
2. Find out what changed on the site just before. A theme update, a revised consent banner, a tag manager edit, or a developer handover can wipe out the tracking code entirely without anyone noticing.
3. Check you are reading the right property. Google’s own guidance names the property definition as a leading cause of apparent losses, noting that “Mixing up http and https is probably the most common reason” a site appears to have lost traffic (source: Google Search Console Help on traffic drops).
4. Confirm the pages are still in the index. As the same guidance puts it, “If Google can’t find and index your site, all your numbers will drop, so start by checking your site coverage in the Google index.”

The pattern worth memorizing is the split between those two systems. Steady impressions alongside collapsed sessions points to an analytics problem almost every time. Impressions and clicks falling together confirms the drop is real and comes from the search side, and then everything that follows applies.

Once you have confirmed the fall is real, the shape of the graph itself narrows down what caused it.

The shape of the fall narrows the cause

Before naming a cause, look at how the line actually fell. A drop that happens overnight, one that slides slowly across months, one limited to a single section, and one that repeats every year are four different problems. The shape rules out possibilities more reliably than it confirms them. Google organizes its own diagnostic documentation around exactly this idea.

That last row is worth knowing before you spend anything. Google is direct that a drop tied to an algorithmic change does not mean a defect. It states that “there might not be anything fundamentally wrong with your content” (source: Google Search Central on debugging drops in Search traffic). It also names the ordinary case that needs no response at all, noting that demand shifts “either due to a new trend, or seasonality throughout the year.”

Shape narrows the field, but it does not name the cause. For that you need to understand how search visibility works in Singapore, plus one more step.

The six kinds of drop, and how to tell them apart

Six causes account for nearly every sudden organic decline. Each carries a single check that sets it apart from the other five. Work down the table until a row matches what you can actually see in your own accounts. The rightmost column deserves reading twice, because it decides who you call rather than just what you read next.

The third row is the one owners fear most and meet least often. A manual action is a human decision rather than a mechanical one, since Google “issues a manual action against a site when a human reviewer at Google has determined that pages on the site are not compliant with Google’s spam policies” (source: Google’s manual actions report documentation). You are notified when it happens, and the notice appears inside your own account. That makes it both the fastest of the six to confirm and the slowest to reverse. What that reversal actually involves belongs to our guide to recovering from a Google penalty or manual action.

The gap between the third and fourth rows confuses almost everyone. Google draws it plainly enough, explaining that it detects violations “both through automated systems and, as needed, human review that can result in a manual action” (source: Google’s spam policies). Only the human version generates a notice. An algorithmic loss arrives entirely silently, which is why owners so often hunt for a penalty that was never issued against them.

In the rescue work AuthorityLighthouse takes on, owners almost always arrive having already decided what happened. The cause they have settled on is often not the one the evidence supports. The most common version is a business braced for a penalty when the timeline points squarely at a site change nobody had classified as a site change. That is why the opening conversation is about dates rather than search. It is also why the migration row deserves checking before the more dramatic possibilities. Some turbulence after a move is expected rather than alarming. Google advises that “with any significant change to a site, you may experience ranking fluctuations while Google recrawls and reindexes your site,” and that “a medium-sized website can take a few weeks for most pages to move in our index” (source: Google’s documentation on moving a site with URL changes). A decline that keeps deepening well beyond that window is a different matter. Protecting a site through such a move is a discipline of its own, covered in our guide to migrating without losing SEO.

The bottom row deserves separate attention, because it is the category now expanding. Pew Research Center analyzed 68,879 Google searches conducted by 900 US adults during March 2025. It found that users clicked a traditional result on 8% of visits where an AI summary appeared, compared with 15% of visits where none did (source: Pew Research Center). A business holding its rankings perfectly can still watch clicks fall for reasons that have nothing to do with its own site.

You can run the checks in the table yourself in about an hour.

How to find your own answer in an hour

You can reach a probable cause in a single sitting, without technical skill, using accounts you already own. The goal is not to repair anything. It is to enter the next conversation knowing which of the six rows describes your situation, so you can judge whether the person across the table agrees with the evidence.

Work through these in order and record what you observe at every step.

1. Find the date the decline began. Narrow it to a single day where you can, because every later step is measured against this date, and a vague “sometime in June” eats most of the hour.
2. Open the manual actions report. It states either that no issues were detected or exactly what was detected. That makes it a clear yes-or-no answer worth taking early, rather than going on speculating about penalties.
3. Compare impressions against clicks across twelve months. Both falling together points to a ranking or indexing problem. Impressions holding steady while clicks fall points to changing demand or a reshaped results page.
4. Check index coverage for your most valuable pages. If pages that used to earn traffic are no longer indexed, you have found the problem. It usually traces to a configuration setting rather than any judgment about quality.
5. List every change made to the site in the fortnight before that date. Launches, redesigns, plugin updates, developer work, and content deletions all belong here, because most declines are explained at this step and nowhere else.
6. Check whether a confirmed Google update landed on your date. If the timing lines up and the decline was broad rather than narrow, you are in the algorithmic row rather than the technical one.

Step four is where most one-hour investigations end. When previously indexed pages are simply gone, the cause is nearly always a configuration setting rather than a verdict on your content, and the fix is just as quick. Our guide to what to do when your site is not indexed at all covers that branch in full.

That hour tells you what you are dealing with. It does not tell you what fixing it will involve, which is the next question.

What a competent rescue engagement looks like

A genuine rescue begins with a written diagnosis you could act on without the agency that produced it. Only after that does it propose repair work. Anyone quoting a monthly retainer before working out what actually happened is selling a subscription rather than a repair. The diagnosis is a separate, bounded piece of work, and it deserves to be priced that way.

The sequence matters far more than the individual durations. Setting the diagnosis before scoping the work is what stops a technical fault being treated with a content retainer. That swap is the most expensive mistake in this business, so the diagnosis and the quote should never arrive together.

From AuthorityLighthouse’s audit and rescue work, the opening days of an engagement are almost all reconciliation rather than search work. You assemble the change log, the deployment history, and the account records into a single timeline, until the date of the decline has a plausible explanation sitting beside it. Owners routinely brace for a large recovery invoice and find the diagnosis costs far less than they feared, because the answer is usually waiting in a timeline nobody had bothered to assemble. Sometimes the honest recommendation is a smaller invoice than the one the owner was already expecting.

Where the repair phase is ordinary ongoing work rather than a one-off fault, it belongs inside the same scope as what SEO services actually cover. If the diagnosis lands on the algorithmic row, expect a longer horizon. Google describes core updates as “significant, broad changes to our search algorithms and systems” that it makes several times a year. Treating them as a recurring category rather than chasing each one individually is the more productive posture, which we set out separately for Google’s algorithm updates in Singapore (source: Google’s guidance on core updates).

That is the work involved. The question owners actually ask is how long before the traffic comes back.

How long recovery takes, and when it doesn’t come back

Recovery times range from days to many months, depending entirely on the underlying cause. Two of the six causes may never reverse completely. An honest answer separates the mechanical faults, which recover fully and quickly, from the judgment-based ones, which recover partly and slowly if they recover at all.

Google sets expectations on the slow rows itself. On manual actions it warns that “most reconsideration reviews can take several days or weeks, although in some cases, such as link-related reconsideration requests, it may take longer than usual to review your request.” On algorithmic recovery it is blunter. It allows that some changes take days, but that “it could take several months for our systems to learn and confirm that the site as a whole is now producing helpful, reliable, people-first content.” Its debugging guidance adds a practical note that saves owners a lot of anxiety: “In general, you’ll likely want to wait a few weeks to analyze your site in Search Console again.”

The bottom row is the one nobody wants to hear, and so the one worth taking most seriously. Analysis of Search Console data across 64 publisher sites by Define Media Group found organic search clicks down 42% against a pre-AI-Overviews baseline by the fourth quarter of 2025 (source: Search Engine Land). Where the traffic that left was answer-seeking rather than buying, trying to win it back is usually money wasted. The more productive response is reweighting toward the commercial searches that still turn into customers. What that reweighting actually returns over a real engagement is visible in our published case studies, where the multi-year records tell a very different story from the six-month ones.

One more caution concerns redesigns. A rebuild ordered to fix a decline the design never caused usually deepens the problem, and the recovery path afterward becomes a separate project entirely. Our guide to recovering traffic after a redesign covers it.

Timelines only matter once the right person is handling the problem, which is the last question to settle.

Who to call, and what to refuse while you’re panicking

Route the problem by its cause rather than by whoever answers the phone first. Four of the six causes are not agency work at all. Paying a retainer to fix a developer’s problem is the standard way a modest fault becomes a year of spending. Match your row against the table before making any call.

Four things are worth refusing outright in the fortnight after a decline, and each is common enough to name on its own.

• A guaranteed recovery, or a guaranteed date. Nobody controls Google’s results, so a promise about them tells you about the seller rather than about your prospects.
• A retainer signed before a written diagnosis. The diagnosis is exactly what tells you whether a retainer is even the right purchase.
• A rebuild sold as a cure. Rebuilding an undiagnosed site destroys the evidence and adds a second cause on top of the original one.
• A link cleanup package bought without a manual action. If your account shows no manual action, you are buying a remedy for a condition nobody has diagnosed you with.

That list matters because a decline is the moment a business is easiest to sell to, and the offers arriving in that window are rarely the good ones. The broader pattern is catalogued separately under the red flags in a bad SEO offer. And if this episode has convinced you the current arrangement is not working, what to establish before choosing an SEO agency in Singapore is covered in its own guide.

Frequently asked questions

Should I fire my agency over a traffic drop?

Not on the decline alone, because two of the six causes have nothing whatever to do with them and a third involves Google acting across an entire category simultaneously. Judge them on their response instead. A competent agency identifies the cause within a fortnight, presents the evidence supporting it, and states plainly which portion was theirs to prevent. An incompetent one arrives with a proposal before any diagnosis, or explains the decline using terms you cannot independently verify. The response is the test, not the decline itself.

Was the drop my fault?

Usually it is nobody’s fault in any useful sense, and the question deserves setting aside because pursuing it slows the diagnosis considerably. The most common causes are a configuration setting changed during ordinary maintenance, a supplier’s work that nobody reviewed afterward, or a decision at Google affecting thousands of sites the same day. The exception involves a site where warnings remained visible for months and were consistently ignored. Establish the date and the cause first, then assign responsibility afterward, when it can be done accurately.

My rankings look the same but traffic is down. What is happening?

You are almost certainly looking at a demand-side drop rather than a ranking one. It happens when the result page changes shape around you, when an answer appears above the results, or when fewer people search your terms than last year. Confirm it by comparing impressions against clicks over twelve months: if impressions hold steady while clicks fall, your position is intact and the click behavior changed. That is a channel question rather than a repair job.

How much should a diagnosis cost?

Treat it as a defined project rather than a monthly fee, and expect it to cost meaningfully less than a retainer month. What you are buying is a written cause with the evidence attached, delivered inside about two weeks, and useful even if you hire somebody else to do the repair. Be wary of a free diagnosis attached to a twelve-month commitment, since the incentive there is to find a cause that requires exactly the service being sold.

Can I just wait and see whether it comes back?

For one of the six causes waiting is genuinely correct, and for the remainder it becomes expensive. A seasonal decline reverses independently, and an algorithmic drop occasionally settles across subsequent updates without any intervention whatever. A technical fault, an indexing failure, and migration damage all deteriorate the longer they remain unaddressed, because the affected pages continue losing ground while nobody is watching them. The safe compromise involves spending the hour establishing which row describes you, then deciding whether waiting constitutes a strategy or an accident.

Does a smaller drop matter, or only a big one?

Size matters less than shape and scope. A 15% fall concentrated on the pages that generate inquiries is a serious commercial problem, while a 40% fall spread across pages nobody was buying from may cost you very little. Check which pages lost the traffic before deciding how alarmed to be, and weight the ones that were producing calls, bookings, or sales. Businesses regularly panic over the wrong number and miss a small one that mattered.

Should I delete or rewrite the pages that lost traffic?

Not before you know why they lost it, because deletion removes the evidence and rewriting the wrong pages wastes a quarter. If the cause was technical, the pages are fine and the setting is not. If the cause was an algorithmic judgment about quality, improving them genuinely helps, but that means substantive improvement rather than cosmetic edits and republished dates. Establish the cause first; the right action follows from it rather than the other way around.

How do I stop this happening again?

Two habits prevent most repeat drops, and neither is technical. Keep a written log of every change made to the site, with dates, so the next fall can be explained in an afternoon rather than a fortnight. And review Search Console monthly rather than only in a crisis, because most of the six causes announce themselves in falling impressions or coverage errors weeks before revenue notices. Businesses that keep both records recover faster, mainly because they spend far less time working out what happened.

Written by the AuthorityLighthouse editorial team with AI-assisted drafting, reviewed by Sunny Tay, Editor. Updated August 2026. Primary sources include Google Search Central, Google Search Console Help, Pew Research Center, and Search Engine Land. Observations about diagnosis and rescue engagements are drawn from AuthorityLighthouse’s own audit and client recovery work. All facts independently verified against cited documentation current as of August 2026.

This article shares general guidance on search visibility for Singapore businesses, drawn from AuthorityLighthouse’s client work. Results shown are from real, named client engagements; every business starts from a different baseline, and no outcome is guaranteed.