Guide · Jun 25, 2026 · 10 min read · by the Keystone Search team

Diagnosing and fixing a traffic drop

The graph drops. Yesterday it was fine, today organic traffic is down thirty percent, and the questions start arriving faster than you can answer them. Is the site broken? Did we get penalised? Is this an algorithm update? Did a competitor do something? The instinct in this moment is to do something, anything, immediately. That instinct is usually wrong. The most common way a recoverable traffic drop becomes a permanent one is panic: someone makes a hasty change based on a guess, then another change to fix the first, and within a week the site is in a worse state than the drop alone would have caused.

This guide is the opposite of panic. It is a calm, ordered checklist for figuring out what actually happened before you touch anything. The goal is diagnosis first, treatment second. By the time you finish reading you should know how to tell the four main causes of a traffic drop apart, how to gather the evidence that distinguishes them, and how to avoid the self-inflicted damage that makes everything worse.

First, confirm the drop is real

Before you diagnose a problem, make sure you have one. A surprising share of "traffic drops" turn out to be measurement artefacts, and chasing a phantom wastes days. Run through these checks before anything else:

• Is your analytics tracking actually working? A broken tag, a botched site deployment that dropped the tracking code, or a consent banner change can make traffic appear to vanish when visitors are still arriving normally.
• Are you comparing like with like? A drop from Friday to Sunday is not a drop, it is the weekend. Compare against the same day last week and the same period last month, not against an arbitrary high point.
• Is the drop in a specific channel? A fall in "organic search" is a different problem from a fall in total traffic. Isolate the channel before you assume it is an SEO issue.
• Is it filtered or sampled data? Large date ranges and heavy filtering can introduce sampling that distorts the picture. Check the raw numbers.

Only once you have confirmed that real organic visitors are genuinely down, against a sensible baseline, in a specific channel, is it worth moving on. If the data is broken, fixing the tracking is the whole job and there was never a traffic problem at all.

The four suspects, and how they behave differently

Nearly every genuine organic traffic drop traces back to one of four causes, and the good news is that they leave different fingerprints. Learning to read those fingerprints is most of the battle. The four suspects are an algorithm update, a technical problem, a seasonal or demand shift, and a manual action. They are not equally likely, and they do not look alike.

An algorithm update tends to be gradual over a few days, affects a broad set of pages rather than a single template, and usually coincides with chatter across the wider search community. A technical problem is often sudden and sharp, frequently tied to a specific date when something was deployed, and tends to hit a structurally defined set of pages, everything under one directory, or every page using one template. A seasonal shift is smooth, predictable, and visible in your year-over-year data; if your traffic falls every December, December is not a crisis. A manual action is rare, arrives with an explicit notice in your search console, and is the only one of the four that comes with a message telling you exactly what happened.

The single most useful question to ask early is: did this happen all at once, or over several days? Sudden and sharp points toward technical. Gradual over days points toward an algorithm update. That one distinction routes you down the right investigative path and saves you from looking for an algorithm penalty when the real cause is a robots.txt file someone changed on Tuesday.

Investigating a suspected technical problem

Technical causes are the most common and, mercifully, the most fixable, because they are concrete. Something on the site changed and broke the relationship between your pages and the search engine. Work through this list, roughly in order of how often each one is the culprit:

1. Check whether pages are being accidentally blocked. A stray "Disallow: /" in robots.txt, or a noindex tag pushed live by mistake, can deindex large parts of a site within days. This is the single most common technical cause and the first thing to rule out.
2. Look for a recent deployment or migration. Did the site move to a new platform, change its URL structure, or ship a redesign right before the drop? Migrations that lose redirects are a classic cause; old URLs that earned rankings now return errors and the equity evaporates.
3. Verify pages are returning the right status codes. Pages that should be live returning 404 or 500 errors, or redirect chains that loop, all bleed traffic. Crawl the site and look for status codes that should not be there.
4. Confirm the search engine can still render your pages. A change to how the site loads content, particularly a shift to client-side rendering that the crawler cannot execute, can make content invisible even though it looks fine in your browser.
5. Check canonical tags and indexing signals. A misconfigured canonical pointing every page at the homepage, or a sitemap suddenly listing the wrong URLs, sends confusing signals that suppress rankings.

The advantage of technical causes is that the fix usually restores the traffic, and often quickly. Remove the accidental noindex, restore the lost redirects, fix the status codes, and the rankings tend to come back as the site is recrawled. If your investigation lands here, you are in the best-case scenario. Many of these issues overlap with the work covered in a proper technical SEO foundation, and a site that had those foundations in place is far less likely to suffer a self-inflicted technical drop in the first place.

Investigating a suspected algorithm update

If the drop was gradual, broad, and coincided with wider community reports of ranking volatility, you are likely looking at an algorithm update. This is a fundamentally different situation from a technical problem, and it calls for a different temperament. There is no single broken thing to fix. The search engine changed how it weighs quality, relevance, or trust, and your pages now sit differently relative to competitors.

The most important thing to understand about algorithm updates is that the recovery is not a switch you flip. You do not "fix" an update the way you fix a broken redirect. Instead, you work out what the update appears to reward, honestly assess where your content falls short of that bar, and improve it, knowing that recovery often only comes with the next update cycle, weeks or months later. This is slow, and it is frustrating, but rushing it makes things worse.

To investigate, segment the drop. Did it hit your whole site evenly, or did certain types of pages suffer while others held steady or grew? Updates rarely treat a site uniformly. If your thin, templated pages fell while your genuinely useful content held, the update is telling you something specific about quality, and the response is to improve or remove the weak pages rather than to tinker site-wide. If a particular topic area collapsed, perhaps your expertise or coverage on that topic was shallow compared to competitors who now rank above you. Many post-update recoveries come down to honestly confronting pages that were never that good and dealing with them properly.

One honest caution: after an algorithm update, the search community fills with confident theories about exactly what changed and exactly what to do. Most of these are guesses. Treat them as hypotheses to test against your own segmented data, not as instructions. The teams that recover are the ones that respond to what their data shows, not to the loudest theory of the week.

Ruling out seasonality and demand shifts

Not every drop is your fault, and not every drop is a problem. If people search less for what you offer, your traffic falls even though your rankings are perfectly intact. A tax-software site loses traffic in May. A swimwear retailer loses traffic in October. A topic that was briefly in the news fades when the news cycle moves on. None of this means anything broke.

The way to rule this in or out is year-over-year comparison and a look at search demand itself. If your traffic for a query is down but the total search volume for that query is also down by a similar amount, your share of the pie has not changed, the pie just shrank. That is seasonality or a demand shift, and the correct response is usually to do nothing reactive, perhaps to plan content around the next upswing, and to resist the temptation to "fix" a problem that does not exist. Mistaking a seasonal trough for a penalty and frantically changing the site is a genuinely common and entirely avoidable error.

Demand can also shift permanently rather than cyclically. If a product category is in long-term decline, or buyer behaviour has moved to a different kind of query, no amount of SEO will conjure back searches that people have stopped making. Recognising this is uncomfortable but important: it redirects effort toward where the demand actually went instead of toward propping up a fading term.

Manual actions and the rarest cause

A manual action is the one cause you do not have to diagnose, because the search engine tells you directly. Check the manual actions section of your search console. If there is a notice there, that is your answer and your path forward is defined: read exactly what the notice describes, fix the underlying issue thoroughly, and submit a reconsideration request. Manual actions are rare for legitimate sites, and they almost always stem from something a site genuinely did, unnatural link schemes, cloaking, thin affiliate pages built to manipulate rankings. If you receive one, the honest work is to fix the actual behaviour, not to argue.

If there is no notice in your search console, you do not have a manual action, full stop. Plenty of panicked teams convince themselves they have been "penalised" when what they actually have is an algorithm update or a technical fault. The word "penalty" gets used loosely, but a true manual penalty is documented and visible. The absence of a notice rules it out and sends you back to the other three suspects.

A calm working order, and what not to do

Putting it together, here is the sequence that keeps you from making things worse. Confirm the drop is real and isolate the affected channel and pages. Check your search console for a manual action, which is the fastest cause to rule in or out. Ask whether the drop was sudden or gradual to choose between the technical and algorithmic paths. Check year-over-year data to rule out seasonality. Then investigate the most likely suspect thoroughly before changing anything.

Throughout, document what you find. Screenshot the rankings, save the crawl, note the exact date of the drop and any deployments around it. This record is what lets you tell, later, whether a change you made actually helped. The most damaging anti-pattern in traffic recovery is making several changes at once in a panic; if traffic then recovers, you cannot tell which change worked, and if it gets worse, you cannot tell which change hurt. Change one thing, observe, then change the next.

Finally, accept the timelines honestly. A technical fix can restore traffic in days. An algorithm-related recovery can take a full update cycle, months, not weeks. Seasonality recovers on its own schedule, which is to say next season. Anyone who promises to reverse an algorithm-driven drop by next week is selling something. The calm, evidence-led approach is slower than the panicked one in the first hour, but it is far faster to a real recovery, because it solves the actual problem instead of layering new ones on top of it.