If your numbers in SEMrush, Ahrefs or Google Search Console have taken a noticeable hit over the past few months, you are not imagining it. I have been seeing the same pattern across multiple sites in different industries, and in almost every case, the anxiety about the numbers runs well ahead of what is actually happening.
The reality is messier than a single explanation. At least four separate things are happening at once: measurement corrections that made inflated data look like a traffic crash, a structural reshaping of the search results page that has been compressing clicks for two years, a pair of March 2026 algorithm updates targeting specific types of content, and Google becoming far more selective about which pages it indexes at all.
Understanding which of these is affecting your numbers, and in what proportion, changes completely what you should do about it.
Want to work out what is actually driving the drop on your site? Get in touch.
Your data was already wrong before the recent drop
Some of what you are seeing is a correction, not a real decline. Two separate measurement events in eight months explain a significant portion of the apparent drop.
September 2025: Google removed the num=100 parameter
Between 12 and 18 September 2025, Google quietly disabled an undocumented URL parameter called num=100. This parameter had allowed SEO tools (including Ahrefs, Semrush, and most rank-tracking platforms) to pull up to 100 search results per query in a single request, rather than loading 10-result pages one at a time. It was never officially supported, but it had become the backbone of how visibility and impression data were collected at scale across the industry.
When Google switched it off, dashboards changed overnight. A study of 319 websites found that roughly 88% of sites saw a drop in impressions, and about 77% lost unique ranking terms in their tracking data.
I watched this happen in real time across every site I monitor daily in Search Console. Impressions fell off a cliff. Average position shot up as if the site had suddenly jumped to the top of every result (almost like winning the lottery … on paper). And yet clicks did not follow. In many cases, clicks dropped too. That combination (better positions, fewer impressions, no gain in clicks) is exactly what you would expect when automated bulk-loading stops generating phantom data, not when a site actually improves its rankings.
Average position appeared to improve because the deep-page results that had been generating automated impressions were no longer counted. The rankings that remained were the ones real users were actually seeing.
The critical point is this: actual clicks and sessions remained largely unchanged. Conversion didn’t change. What disappeared were phantom impressions created by tools bulk-loading 100 results at a time and data generated by bots, not real users. Ahrefs and Semrush had to retool their collection methods, and for several weeks, the historical trend lines showed a step-change that reflected a measurement correction, not a ranking loss. Many keywords ranked on pages 2-10 disappeared from reporting.
April 2026: Google confirms GSC over-reporting since May 2025
Then, in April 2026, Google confirmed a separate issue. Search Console had been over-reporting impressions since 13 May 2025 … nearly eleven months of inflated data from the one tool most SEOs and marketers treat as closest to the reliable truth.
The fix is framed as greater transparency (a more accurate description is that Google is correcting data that should have been right in the first place).
So, the GSC impressions drop you see right now has two distinct causes: the September 2025 num=100 cleanup, and the April 2026 GSC recalibration. Neither represents a real collapse in search visibility (both were measurement problems only).
To be clear about something: Google does not owe anyone a perfect tool. Search Console is free, and google.com is Google’s website. They can update it, change how it works, or adjust what it reports without asking permission or issuing advance notice. All we can do is observe.
What this means in practice is that treating any platform’s data as the absolute truth was always a mistake, and this episode just makes that harder to ignore. Paid tools estimate. Google has the actual data and still got reporting wrong twice in under a year.
If you are comparing your current impressions to anything before September 2025, you are not comparing like for like.
Interpreting SEO data properly has always required scepticism. Use data as a directional signal and verify it against behaviour. When a number drops suddenly, ask first whether the metric changed or whether the measurement did.
What Google has done to the search results page
Separate from the data problems, the environment your content ranks in has genuinely changed. And I want to be clear, again: I am not against any of these changes. They can design their results page however they want, the same way any business can update its product. What we can do is understand what changed, adapt, and keep working. What we should not do is pretend the SERP is the same as it was three years ago.
The rollout of AI Overviews
Google announced Search Generative Experience (SGE) at Google I/O in May 2023 as an opt-in experiment. In May 2024 it was rebranded as AI Overviews and launched to all U.S. users. By August 2024 it had rolled out to the UK, India, Japan, Indonesia, Mexico and Brazil. By May 2025 it was live in over 200 countries and 40 languages.
In March 2025, Google began testing AI Mode (where an entire results page is AI-generated) and by May 2025 had started rolling it out to U.S. users.
AI Overviews now appear in roughly 13% of all queries globally, with prevalence above 30% in some verticals. A UK SERP study found them appearing in nearly one in five UK searches, with particularly high presence in health, finance, and local services. For e-commerce queries specifically, BrightEdge data shows AI Overview presence grew from 14.4% to 20.7% of terms in just 23 days in September 2025.
The impact on organic results sitting below AI Overviews has been measured independently by multiple research teams. Every single study found lower click-through rates:
| Source | CTR impact when AI Overviews appear |
|---|---|
| Seer Interactive (2025) | Organic CTR dropped 61% — from 1.76% to 0.61% — across 25.1M impressions |
| Ahrefs (updated Dec 2025) | Position-1 CTR reduced by 58% on AI Overview queries vs matched non-AIO queries |
| Authoritas | Top organic CTR down ~79%; desktop traffic down 56.1%, mobile down 48.2% |
| BrightEdge (one year of AIO data) | Impressions up 49% since May 2024, but click-throughs down nearly 30% |
| Ideava meta-analysis (12+ studies) | Every study found CTR decline; reductions ranged from 15% to 89% depending on niche |
I have been watching this pattern across too many sites to treat it as a coincidence. Rankings hold. Sometimes, average positions even appear to improve slightly. But clicks fall anyway.
The question worth asking when you see a drop is not just “did we lose rankings?” … it is “did clicks disappear while positions stayed?” If the answer to the second question is yes, AI Overviews and SERP feature expansion are almost certainly the bigger factor, not an algorithm penalty (not because you did something wrong).
How Google changed its ad labels
Google’s ad labelling has been progressively softened over the past decade, with each iteration making paid results harder to distinguish from organic ones:
- Pre-2013: ads had coloured background shading that visually separated them from organic results
- 2013–2014: the background was removed, and a small yellow “Ad” label was introduced
- 2016: the yellow label was replaced with a green one that matched the colour of organic URL text, making the two look almost identical
- 2017–2019: the green label was iterated to a white label with a green border, then to more subtle mobile treatments
- 2019: a new mobile design introduced a black “Ad” label alongside favicons for both paid and organic results, reducing the visual difference further
The UK’s Competition and Markets Authority and the European Commission have both raised concerns about how commercial content is presented in Google search results. By Q1 2026, ads appear inside 25.5% of AI Overview results, up from around 5.17% in early 2025. Google is not just monetising the classic results page … it is extending that monetisation into the AI layer sitting above it.
SERP features pushing organic further down
Try searching for any moderately competitive informational query and look carefully at what sits between the first organic result and the second. Often, it is a People Also Ask block, then an AI Overview, then possibly a Discussions section. The second organic result ends up so far down the page that it sits closer to what used to be position nine or ten in the old layout, when the main variables were paid ads and organic listings. Now imagine ranking third.
This is not a stable or consistent layout. Google changes it constantly … by query, by intent, by day, and by personal signals, including whether you are logged in and your search history. That variability makes results harder to predict and harder to report on accurately, because the same ranking can produce very different visibility depending on what Google decides to insert above it on any given search.
A 2025 study analysing 1,000 high-volume UK keywords found that 98.7% of Google UK searches now display at least one SERP feature or enhanced element. STAT’s 2024 SERP features research found that these features account for roughly 65% of total visible attention on the page.
Research by Yard Digital shows a position-two organic result can appear anywhere from around 300 pixels down the page to 1,700 pixels down, depending on how many modules stack above it. The most common features occupying that space include:
- Local packs, present in 29.7% of all UK searches, rising to 90–95% in service categories where the three-pack fills the entire above-the-fold view on mobile
- Featured snippets, present in approximately 19% of queries
- People Also Ask boxes, present in around 8.5% of searches
- Shopping results, appearing in 11.6% of all UK searches, with much higher rates in e-commerce.
Even queries without AI Overviews saw organic CTR fall 41% year-on-year in one dataset, meaning the broader SERP environment is eroding clicks independently of AI. Understanding what affects your organic traffic now requires accepting that the environment your content ranks in has changed fundamentally since most benchmarks were set.
Google’s March 2026 core and spam updates
On top of the structural SERP changes, Google ran two significant updates in March 2026 that had a direct impact on rankings, particularly for sites producing content at volume.
The spam update launched on 24 March 2026 and completed in roughly 20 hours (the fastest spam update on record). Google described it as a standard global update enforcing existing spam policies. A few analysts pointed to a focus on hacked content, cloaking, and pure spam patterns rather than broad content quality. It was targeted rather than sweeping.
The core update started on 27 March and completed on 8 April (a 12-day rollout). Google described it as “a regular update designed to better surface relevant, satisfying content for searchers from all types of sites.”
What early data shows is less routine. Sites with AI-generated content at scale, programmatic city and service pages with near-duplicate copy, and affiliate or review sites with thin, summary-style roundups saw the hardest hits. Some affected sites reported drops of 60–90% in organic visibility within 72 hours.
The pattern was consistent with what has been building since March 2024: spam update first to strip obvious abuse, then a core update to recalibrate quality and relevance. Winners had the same profile each time … strong authorship signals, genuine depth, first-hand experience in review content, and sites that had already removed or improved legacy thin content after earlier update cycles.
You can find more background on how this pattern developed in the Google November 2024 core algorithm update breakdown.
What Google now does with content it considers unhelpful
This is the part that matters most for the medium term, and the most under-discussed part of the current environment.
From helpful content update to core signal
The September 2023 Helpful Content Update introduced a site-wide classifier that could demote entire domains with a high proportion of unhelpful pages. In March 2024, Google folded that signal directly into its core ranking systems. Google said this integration was intended to reduce low-quality or unhelpful content in search results by 40%. Since then, unhelpful content is no longer assessed by a separate system (multiple core ranking signals now evaluate it simultaneously on every update).
The definition of “unhelpful” has expanded considerably since 2023. By 2025–2026, Google’s targeting is more specific:
- AI-generated content at scale with no human editing, original perspective, or first-hand experience
- Affiliate and review content that is summary-only, with commercial links outweighing genuine evaluation
- Programmatic pages: large networks of near-duplicate location, city, or micro-topic pages built from templates
- Off-topic content published to chase search volume rather than build genuine topical authority
- Outdated or unmaintained archives that dilute overall site quality without serving any real user need
The site-wide nature of the signal matters. A site does not just lose ranking for individual weak pages … Google can demote the domain as a whole when the proportion of unhelpful content is high enough. Publishing large volumes of thin content does not just leave those pages unranked. It can drag down your stronger pages with it.
Two things I have observed personally that are worth flagging.
The first is about internal linking. If you have strong, well-ranking pages that link out to thin or low-quality content elsewhere on the site, those links can work against the page they come from. Google evaluates where a page points, not just where it receives links from.
In several cases I have worked on, removing internal links from strong articles to weaker pages (and/or removing the weak pages entirely) improved the ranking of the stronger content. It is not always the sole reason for a recovery, but the pattern has been consistent enough that I no longer treat it as a coincidence. Linking to a weak page does not automatically block your strong page from ranking, but, for sure, it does not help.
The second is about scale. I removed hundreds of pages from sites I manage (outdated posts, off-topic articles, content that was published to hit a volume target rather than serve a real need). The result, when combined with publishing fresh, well-structured content in the right areas, has been a faster and more predictable improvement than most people expect. You do not always need months to see movement.
On a site Google already trusts, quality content published in the right context can rank in under 24 hours. The site also needs to be in good health.
The difficult truth is that a lot of businesses are sitting on hundreds of pages that are quietly pulling their domain quality down, and nobody has told them. If you were told that you need time to rank, that may be true. But it is also possible that the content already on your site is the thing holding back the content you are trying to rank. A content audit is worth doing before publishing anything new.
Not sure where your site stands? Get in touch, and I can take a look.
What “Crawled – currently not indexed” actually tells you
Many site owners are currently seeing a spike in this status in Search Console and it deserves a plain explanation.
“Crawled — currently not indexed” means Google visited and processed the URL but chose not to add it to the index. It is not a technical error, it is a quality judgement. Google considers the page unworthy of index space, and it cannot appear in search results for as long as that status holds.
The content types that tend to accumulate here are consistent across multiple analyses:
- Thin or low-value pages: short posts, boilerplate location pages, auto-generated content with no unique value
- Duplicate or near-duplicate pages: parameter variations, tag and category archives, copied manufacturer descriptions
- Poorly linked pages: URLs buried deep with few or no internal links, orphaned from the rest of the site
- Off-topic or low-engagement content that attracts neither links nor user engagement (think of that article the business wanted to publish, but nobody is actually searching for it…).
From around 27 May 2025, multiple sites reported spikes in this status correlated with pages offering low uniqueness or thin value, even where those pages were technically accessible with no errors. Entire sections of sites, including old blog archives, tag pages, and thin category pages, disappeared from Google’s index without any manual action. Google is now far more willing to simply not index marginal URLs, especially when a site has a high ratio of them.
This is not a penalty. But it is a clear signal, and it compounds. Every page Google decides not to index consumes crawl budget without contributing authority. A site carrying many such pages is sending quality signals that affect how Google treats everything else it chooses to rank.
The SEO metrics that actually matter in this environment include your index coverage. How much of your published content Google considers worth keeping is now a meaningful indicator of overall site health, not a background technical detail.
Is SEO dead? What the actual numbers say
No. But “not dead” is not the same as “unchanged,” and both things need to be said.
Organic search still drives roughly 53% of all website traffic on average across the web, based on BrightEdge data. That figure is skewed by the large number of sites with no paid activity at all.
For businesses running active paid search campaigns, the paid share is typically far higher and, in some cases, exceeds organic. The point the data makes is not that paid is insignificant, but that organic remains the dominant acquisition channel at scale. The global SEO software market was valued at approximately $84.9 billion in 2025 and is projected to reach $96.4 billion in 2026. Investment in SEO is not declining.
What has shifted is how search demand converts into clicks to external websites:
- BrightEdge shows Google search impressions up over 49% since May 2024, but click-through rates are down nearly 30%
- SparkToro’s 2024 zero-click study found that 58.5% of U.S. Google searches result in zero clicks, with only about 36% going to the open web
- Graphite’s large-scale dataset shows Google referring approximately 16% less organic traffic to external websites than at the start of 2025
- Gartner estimates AI-powered assistants will handle around 25% of global search queries by 2026
Search demand is growing. The share of that demand that routes users to external websites is shrinking, and both things are true simultaneously.
The framing I find most useful is this: AI has not replaced search. It has expanded how and where search happens. The same person who used to Google a question is now sometimes asking Perplexity, ChatGPT, or another AI tool instead. That is not a new audience; it is the same audience on more surfaces. And the technical foundations, content depth, and brand signals that earned you visibility on Google are the same foundations that surface brands across AI search tools.
Whether organic search is still worth investing in has a clear answer in the data. But what that investment looks like needs to reflect the current landscape, not 2022.
Follow the money: who is still hiring for SEO
If SEO were dying, you would not see the companies building the AI tools that are supposedly replacing it hiring senior SEO leads at quarter-million-dollar salaries.
In February 2026, Anthropic posted an SEO Lead role at a base salary of $255,000–$320,000 USD. The role covers organic search strategy, technical SEO infrastructure, and conversion optimisation across claude.ai, docs.anthropic.com, and anthropic.com … described explicitly as “a core growth lever, not a side project.”
Stripe is hiring an AEO/GEO Marketing Manager at $143,400–$215,200 USD. These are not legacy businesses reluctantly keeping an SEO team alive. These are the companies building the AI products people claim are killing search.
The broader hiring data tells the same story. A 2026 analysis of 3,900 SEO job listings found that Director, VP, and Head-level roles now account for 59% of all SEO postings (meaning most hiring demand is at the senior strategic level, not being wound down).
AI and AEO/GEO appear as explicit skill requirements in a growing share of listings, and AI-related skills in SEO job descriptions increased 21% year-on-year.
| Company | Role | Salary range |
|---|---|---|
| Anthropic | SEO Lead | $255,000–$320,000 USD base |
| Stripe | AEO/GEO Marketing Manager | $143,400–$215,200 USD |
| Zapier | Sr. Product Manager, Organic Growth (SEO/GEO/AEO) | $170,700–$256,100 CAD + equity |
| Rocket Money | Senior SEO/GEO Manager | $145,000–$175,000 USD + bonus |
| Checkr | Senior SEO & AEO Manager | $146,000–$172,000 USD |
SEO is not shrinking. It is being re-tooled as organic growth and AI search visibility leadership, sitting at the intersection of content, engineering, and revenue.
What still works, and why strong sites are still growing
Despite the compression, certain sites are clearly gaining share. The question is what they have in common.
When a site is cited inside an AI Overview, research shows it can gain approximately 35% more organic clicks and 91% more paid clicks compared with not being cited (even in a SERP where overall CTR is down).
BrightEdge’s AIO research found that technical, in-depth content is over-represented in AI Overviews, with 48% more technical terms and 49% more complex queries triggering AI responses. Specialist, detailed, original content tends to get cited. Generic content tends to get replaced by a summary.
The 80/20 principle holds in organic search as clearly as anywhere else:
- Around 20% of pages drive approximately 80% of organic traffic
- Around 20% of keywords produce the majority of qualified visits
- A minority of inbound links carries most of the link equity
In 2026, that productive 20% tends to be topic clusters and pillar pages, high-intent service and product pages, and content that supports or builds branded demand. Branded queries are more resilient to CTR compression in AI-heavy SERPs because users are specifically looking for you.
Sites that shifted from publishing large volumes of thin posts to a smaller set of comprehensive, well-linked pillar pages reported traffic increases of around 70% over six months in several documented cases.
I keep seeing this pattern across projects in competitive niches. Even with organic results pushed further down the SERP (like everyone’s)… intent, structure, knowing what Google expects, understanding what users want, and consistency still win. When the fundamentals are right, growth does not disappear, it compounds. And it tends to compound in a way that becomes very clear when you zoom out past the monthly noise.
Organic traffic still matters more than the current numbers suggest. The goal has shifted: be the source Google references in its AI summaries, not the result it replaces with one.
What to focus on right now
If your traffic or impression data has dropped, the first step is separating the measurement changes from real performance shifts. If clicks and sessions are broadly stable while impressions are down, you are likely seeing the September 2025 and April 2026 data corrections working through the dashboards. If clicks are also down, the next question is whether that coincides with increased AI Overview coverage on your key queries, ranking shifts from the March 2026 updates, or both.
Check your Search Console coverage report. If you have a large number of pages sitting in “Crawled — currently not indexed,” that is a signal worth acting on before anything else. Google is already telling you it does not consider those pages worth keeping.
Practical priorities for limited time and resources:
- Identify the pages already driving the majority of your organic revenue or enquiries and invest in improving their content depth, internal linking, and page experience – this is the 20% that drives 80% of results
- Audit your index coverage: pages with “Crawled – currently not indexed” status need either improving to a standard Google will index, or removing and consolidating to stop them diluting overall site quality
- Prioritise queries where a visit is still necessary – transactional searches, localised searches, complex questions an AI Overview cannot fully answer in a single paragraph
- Focus link-building and digital PR on a short list of commercially critical pages rather than spreading effort thinly
- Stop comparing pre-September 2025 data directly with post-September 2025 data as if they are measuring the same thing … they are not.
One thing worth saying directly: the fastest SEO progress I see happens when businesses work with their SEO, not just assess the output. Access to the site, reasonable content approval turnaround, development resource, the ability to act on recommendations – all of it matters. When those things are in place, the compounding effect of consistent work becomes visible relatively quickly. When they are not, even a sound strategy stalls.
The right question is not only “what did SEO do this month?” but also “what did we do to help it grow?”
Pausing SEO activity in response to a decline almost always slows recovery. The sites that come through core update cycles in better shape are consistently the ones that kept building during the difficult months, and not because they ignored the data, but because they understood what it was and was not telling them.
SEO is not dying, but bad SEO always was.
If you want to talk through what your data is showing and where the real priorities are, get in touch.