The “Great Decoupling”: when impressions grow but clicks collapse (explained by the man who coined the term)
Over the past few months, more and more business owners have been living the same nightmare: impressions keep rising while clicks collapse. Their pages still show up in search results, but users stop at the answer provided by Google’s artificial intelligence without ever visiting the site.
This phenomenon has a name: the “Great Decoupling”. The term was coined by Darwin Santos, Senior Technical SEO and AI Innovation Product Manager, an internationally recognized authority for Fortune 100 companies, top e-commerce sites and major online publishers. Beyond building tools that have become industry standards, such as GSC Guardian, Santos is the co-founder of AI Studio Lab, where he develops advanced solutions to integrate AI into business processes.
We reached out to him to understand, straight from the source, what is really happening behind the numbers, why the “Great Decoupling” is not simply a Google penalty, and how the rules of online competition are changing in the era of “agentic” search.
In his answers, Darwin did not stop at describing the problem: he explained how he sees it from the inside, which strategies can still work, and what to do today to avoid being crushed by the collapse in clicks.
An interview worth reading to the very end to understand how not to fall behind while the web changes before our eyes. Enjoy the read!
“If AI agents can’t find your brand, how can they recommend it?”
When you first talked about the “Great Decoupling”, were you simply describing an anomaly in the data, or had you already sensed that Google was rewriting the rules of the game, penalizing websites in favor of its AI-generated answers?
What I saw wasn’t just a spike, it was a genuine phase shift. The “Great Decoupling” is the moment when impressions keep climbing while clicks keep falling. It is not Google penalizing websites; it is a structural change in how people get their answers.
The explosion of ChatGPT caught the market by surprise, Google included — the company had had internal chat systems for years but had never integrated them into its core search. With ChatGPT, Google accelerated: first Bard (now Gemini), then SGE and AI Overviews, all the way to the current AI Mode. Search is turning from a list of links into an “agentic” experience where the answer comes first.
Is this collapse in clicks purely down to the efficiency of the algorithm, or has Big G consciously decided that websites have become “useless middlemen”?
It is mostly a matter of how users behave when faced with a new interface. Many have realized they can get enough information from a chatbot or directly in the SERP, so they stop there.
Links still matter: Google still needs the open web to ground its answers and continues to show its sources. But the mechanism is clear: if the answer on the screen satisfies the intent, fewer people click. Users favor speed, and the interface lets them stop sooner.
Some argue that the Great Decoupling is simply a death sentence for anyone living off generalist content. But are there really any “safe” industries, or is it only a matter of time before Google absorbs everything?
AI models cannot replace first-hand experience. They can synthesize, but little more than that for now. Today the most valuable clicks come from people who still need to take an action: requesting a quote, booking, buying, or verifying a complex, personal answer.
That is where the opportunity lies: offering genuine expertise, proprietary data and experience-based content. AI systems feed on public sources to learn and to cite; if you are not visible, you are left out. Fewer clicks, but with higher intent.
Today, if a business owner keeps measuring success with pre-AI Overview metrics, are they simply throwing their money away? Or are there still “old school” strategies that will make sense in 2025?
Classic strategies are still effective, but they should be complemented with new metrics. You need to maintain the technical health of the site, its information architecture, entity and brand building, feeds, local SEO and conversion optimization.
On top of that come new indicators: how often you appear in AI Overviews, your share of citations in AI Mode, mentions in LLMs, traffic coming from AI sources and the degree of “agent readiness” of your content — that is, whether it is machine-parsable and free of ambiguity.
You need to strengthen the areas where AIO has less of a presence, such as images, video and news, and make the most of the bottom-of-funnel pages that keep generating conversions.
Google claims AI Overviews are designed to improve the user experience. Yet content creators find themselves cannibalized by increasingly complete snippets, with nothing in return. Isn’t this a form of expropriation? The information produced by publishers ends up in AI products… without publishers seeing a cent! Isn’t it time to renegotiate the deal with the platform? Do you think Cloudflare and its “Pay Per Crawl” could help?
I have been advocating crawl- and access-based compensation since early 2023: if AI agents use a piece of content beyond the site itself, it must be possible to measure that and recognize its economic value. Compensating creators is essential.
Systems like “Pay Per Crawl” are a step forward compared to the current vacuum. They are not the definitive solution — practical enforcement and economic sustainability remain complex — but, together with direct licensing and clear opt-in/opt-out mechanisms, they offer a path toward fair compensation while preserving publishers’ ability to be discovered by AI agents.
In your articles you often talk about the rise of the Agentic Web. But in a context where AI doesn’t just suggest links, it fills in forms, writes emails and books services, what role do companies play in building their own brand? How do you create trust when the intermediary is no longer the user, but a language model?
AI agents should be treated like expert, impatient users. You need to be readable and verifiable by machines: publish canonical, structured information — such as prices, specifications, policies and locations — back up your claims with direct evidence and third-party citations, and expose endpoints agents can easily use, such as clean HTML, feeds and documented actions, so they can complete tasks accurately. A brand that cannot be found by agents cannot be recommended by agents.
Many business owners are increasingly frustrated by the decline in organic traffic, even on quality sites. If AI reduces the web to a middle layer to be exploited without giving anything back in terms of visibility, aren’t we heading toward a concentration of value in the hands of a few? What advice would you give to those who want to survive in the post-Search era?
Value is already concentrating in a handful of platforms. The most effective approach is to become an authoritative source, worth citing and worth parsing. That takes original, expert, easily quotable content, backed by strong brand signals — such as entity clarity, reviews, PR activity and active communities — that models and AIO can recognize.
You need to track the new visibility, for example your share of citations in AIO and your mentions in LLMs, and diversify demand through video, podcasts, email and communities, capturing the higher-intent clicks that still remain.
You write that “being found no longer means being visited”. But if Google no longer sends traffic and ChatGPT answers without citing its sources, who pays the bill for content production? The attention economy has collapsed: aren’t we living through the first true systemic crisis of online information?
It is a real stress test for the referral economy. The upside is that a new discovery channel exists, provided your content is present, parsable and attributable to agents. At the same time, new forms of compensation are emerging, such as licensing and paid crawling models.
Until things stabilize, ROI will come from fewer clicks with higher intent, as well as from downstream conversions influenced by agent-mediated discovery. That is what should be measured, not yesterday’s CTR.
Big G absorbs content to train its AI and gives back only clickless visibility. In essence, it feeds on value without redistributing it… How sustainable is this model, other than for Google itself?
Left unregulated, the risk is real: the “tragedy of the commons” could come true. Two countermeasures are taking shape: large-scale licensing deals for news platforms and communities, and infrastructure that charges for crawling instead of giving it away for free.
The solution is not a single measure but an integrated package combining products, policies and payments, alongside a re-optimization of brands to increase their visibility in the eyes of agents and their ability to generate conversions.
With AI Mode, Gemini and the expansion of AI Overviews, Big G has chosen to cannibalize its own referral system. But the real question is: how far are we from the point where the entire editorial web is reduced to free infrastructure for artificial intelligence? Who will be willing to produce content in a world without traffic?
Companies training AI have already strip-mined the public web more than once. Yet most traffic still concentrates on non-informational pages — transactional, local, or product and service pages — because the action still requires a click.
High-signal-value areas, such as video and news, continue to generate visits.
Content creation remains strategic when it earns agent recommendations, converts the remaining higher-intent traffic, and benefits from growing compensation through licensing and paid crawling models. Fatalism is a mistake: the way forward is re-optimization.
We are witnessing an almost frantic synchronization between OpenAI’s releases and Google’s immediate responses. How much is this race to innovate hurting the web ecosystem? Are we building better products for users… or an unstable, unmanageable system spinning out of control for the people who work in it?
Both. Users gain speed and capability: an agent can analyze hundreds of pages in seconds and complete a task.
That pace, however, introduces volatility for publishers and measurement systems. The responsible approach is to ship and correct: more accurate citations, more reliable summaries, clearer reporting and sustainable compensation models.
You wrote that “Microsoft made Google dance, but OpenAI was the DJ”. Who is really winning this game? And above all, what is the collateral damage for those who live off organic traffic and find themselves crushed under a war between AI titans?
Users are coming out ahead, able to gather information at unprecedented speed and scale. Platforms embracing agentic, answer-based models are preserving their market share.
The flip side is the weakening of the old referral economy, especially for informational content. Survival is not an epitaph: you need to produce content worth citing, make it readable by agents, measure visibility in the AI era, and capture the fewer but more targeted clicks that remain.
In the AI era, the winners are those who get cited and capture the clicks that really matter
The “Great Decoupling” Darwin Santos talks about is more than a statistical trend: it is the sign that the relationship between content, users and search engines is changing at its core. Google and AI agents no longer just point to sources; they often deliver the answer directly, shrinking the room for maneuver of those who produce information.
Competition is no longer played out only on SERP positions, but on the ability to be perceived — and treated — as indispensable sources by the algorithms. To get there, you need to speak the language of machines: canonical information, structured data, verifiable content and clean endpoints that agents can read and understand without ambiguity. A brand invisible to agents is, in effect, shut out of the recommendations that matter.
Organic traffic is not disappearing, but it is becoming rarer and more selective. Today ROI comes from fewer clicks with higher purchase intent, and from the conversions generated after users discover the brand through agents.
That is why measuring yesterday’s CTR no longer matters; what matters is assessing how well those few clicks translate into concrete actions.
Santos’s message is clear: the web is changing too fast to stand still.
Those who manage to adapt, building a brand that AI agents can recognize, will keep growing. Those who don’t risk being pushed to the margins of an ecosystem where visibility alone is no longer enough.
#avantitutta
