Google, AI engines, and web traffic: what data really say
AI Overview, AI Mode, ChatGPT, Perplexity: over the last year, traffic data on our websites has begun to change. Fewer clicks, more direct responses, new models of access to information. But what is the real scope of this change? The numbers paint a more complex picture than the alarmist predictions: Google is growing, AI tools are advancing but remain marginal, and the value of organic traffic is shifting from quantity to quality. Here’s what’s really happening, with numbers to back it up!
Web traffic: Google Search continues to grow, AI engines follow
Google remains at the center of online search. Despite growing attention to AI-based tools such as ChatGPT and Perplexity, the Mountain View engine continues to see growing volumes. According to data collected by SparkToro in collaboration with Datos, in 2024 Google surpassed 5 trillion annual searches, marking a 21.64% increase over the previous year. This is a significant result for a platform considered by many to be in decline—even though, for the first time in years, Google’s market share has fallen below 90% in recent months (but we are talking about decimals: it currently stands at 89.54%).
This trend contradicts the idea that the advent of AI-based solutions is eroding Google’s user base. On the contrary, traditional search remains central to the everyday digital experience, even with the arrival of tools such as ChatGPT, Perplexity, Claude, and Copilot. ChatGPT, currently the most widely used AI engine, handles around 37.5 million requests per day: a figure that, compared to Google’s 14 billion daily queries, highlights a 373-fold gap. Even when other competitors are added, the total share of AI engines does not reach 2%. Furthermore, according to analysis by Semrush, ChatGPT sessions tend to have an average of eight messages, confirming that its use is more conversational than comparable to a classic search engine.
While media attention seems to favor novelty, the numbers put the phenomenon into perspective: the public currently continues to prefer familiar and established search models. And Google’s move to integrate AI innovations directly into the search experience seems to be paying off, as it maintains usage volumes that no other engine, traditional or AI-based, is currently able to match.
The real impact of AI Overview
With regard to the AI features of Google Search, and pending actual data from reporting tools, there are already several studies that have analyzed the effects of AI Overview on site visits—and they all agree on a reduction in organic traffic.
These AI-generated responses have already changed the way users interact with search results, going even further than the difficulties that emerged when featured snippets were introduced, because they are prominently placed in SERPs and further reduce the visibility of traditional organic results.
The numbers leave no room for doubt: according to an analysis of 700,000 keywords conducted by Amsive, the average CTR for queries that trigger an AI Overview has decreased by 15.49%.
The effect is even more pronounced on non-branded searches, where the reduction in clicks reaches -19.98%. In contrast, searches related to a specific brand show significant resistance: for these queries, the presence of AI Overviews is associated with an average increase in CTR of 18.68%. No less significant is the dynamic of overlap between AI Overview and featured snippets: according to the same Amsive data, when both elements appear simultaneously, competition for space “above the fold” intensifies, leading to a further contraction in CTR, which falls by an average of 37.04%.
This dynamic is not uniform across different sectors. Amsive highlights how sectors such as Healthcare and Education are among the most penalized in terms of CTR decline, while areas such as Finance, characterized by strong and recognizable brands, show greater resilience. The ability to maintain organic visibility therefore seems to be linked to brand strength and query structure.
The overall effect of these transformations has been described by Barry Schwartz of Seroundtable as the “Great Decoupling”: the growth of impressions in AI-powered SERPs no longer translates into proportional growth in clicks. The user’s attention is captured on the page itself, without the need to delve further into external sites.
AI traffic moves mainly from desktop, but mobile remains strategic
The adoption of AI-based tools is not uniform across different devices. According to the latest report from BrightEdge, over 90% of referral traffic generated by platforms such as ChatGPT, Perplexity, Bing AI, and Google Gemini comes from desktop. In detail, ChatGPT receives 94% of its traffic from desktop devices, Perplexity reaches 96.5%, while Bing stands at 94%. Even Gemini, Google’s AI assistant, follows the same pattern with a desktop share of over 90%.
This contrasts sharply with the behavior observed for traditional Google Search, which maintains a mobile prevalence: 53% of traffic comes from mobile devices, compared to 44% from desktops. This confirms that, despite the advancement of AI technologies, mobile search remains a staple for most everyday searches.
One possible explanation for this polarization is linked to the context of use. AI tools are mainly used in professional environments and from fixed locations, where longer and more complex interaction with generative models is more natural than the quick consultation typical of mobile devices.
Another key factor concerns Apple. According to BrightEdge, approximately 58% of mobile traffic to US and European brands comes from iPhone devices. Since Safari is the default browser on these devices, any changes to the default search engine could have a significant impact on the balance of market share. At present, Google maintains a dominant position, but Apple’s control over the mobile gateway remains something to watch closely in the coming months.
Let’s look at the data: the case of publishing in the US and SEOZoom
This technical snapshot is accompanied by more structural data, which puts its finger on the sore spot of a sector that is apparently paying the highest price for this situation, namely publishing sites, which had already been hit by the dynamics of “SEO journalism.”
According to an analysis by Similarweb published by Axios, between May 2024 and February 2025, the leading US news sites lost an average of 15% of their traffic from Google. In absolute terms, each publication recorded a drop of around 800,000 monthly visits, falling from 5.3 to 4.5 million.
During the same period, referrals from AI chatbots (ChatGPT, Gemini, and others) grew by 2,100%, but these numbers are still incomparable. In February 2025, AI engines generated 6 million total visits to the top 500 US sites, compared to 2.3 billion from traditional search engines: a ratio of 1 to 400.
Another concrete example comes from the analysis of traffic on our website seozoom.it, which shows that the impact of AI-based tools is still marginal compared to traditional search.
According to data from Google Analytics for May 2025, the site recorded a total of 135,624 sessions.
Of these, direct traffic accounts for approximately 44.43%, while organic traffic from Google Search accounts for 30.94%.
The most interesting figure concerns traffic generated by AI-based engines:
- chatgpt.com/referral brought in 491 sessions, equal to 0.36% of total traffic.
- perplexity.ai/referral generated 176 sessions, equal to 0.13%.
- There is also a minimal amount of “not set” traffic associated with ChatGPT, with 33 sessions (0.02%).
The sum of traffic from generative AI tools therefore represents less than 1% of total sessions.
These numbers offer a concrete representation of the current situation: despite the public debate on the rise of AI engines, traffic from tools such as ChatGPT and Perplexity remains a marginal component compared to traditional Google search. Even on a highly specialized site capable of attracting a highly profiled audience, organic search continues to be one of the main sources of traffic.
Focus on e-commerce and retail sites: the new role of AI engines
The weight of AI engines seems to be changing if we look at trends in the e-commerce sector, where somewhat different signals are emerging. According to an analysis by Forrester Research, traffic to US retail sites has already fallen by between 15% and 50% year-on-year, partly due to the spread of AI search tools such as ChatGPT, and the rise of AI Mode “will further reduce organic traffic to merchant sites,” according to Nikhil Lai, principal analyst at Forrester.
At the same time, data from Adobe Analytics for the first quarter of 2024 shows that:
- 16% of traffic leading to a conversion comes from organic search
- 28% comes from paid search
- This is followed by direct visits (20%) and affiliates (17%).
These figures clearly indicate that, for online merchants, search-generated traffic—particularly organic traffic—remains a key asset for conversion, and that any change in user behavior on search engines can have immediate economic effects.
Added to this is a change in search habits: according to a February 2025 Adobe survey, 39% of US consumers use generative artificial intelligence tools to shop online. The most frequent activities include:
- Product search (55%)
- Receiving recommendations (47%)
- Searching for deals and promotions (43%)
AI traffic to retail sites—while still modest compared to other channels—grew by 1,200% between July 2024 and February 2025, doubling every two months.
However, users coming from AI sources browse more but convert less: +12% page views per visit, -9% likelihood of conversion compared to other channels. This suggests that, for now, the visibility gained from AI referrals does not compensate for the traffic and conversion losses suffered on Google Search.
Fewer visits does not (always) mean fewer opportunities
We have been talking about the phenomenon of zero-click searches in Google for years now — searches that do not generate any clicks to external sites — but now the situation seems to have worsened, from the perspective of those who have a digital project and are helplessly watching their website traffic decline.
This is not a new problem, and it is not solely dependent on AI: recent data from SparkToro for 2024 already indicated that only 36% of Google searches in the US and Europe generated a click to the open web. This means that around 64% of searches no longer drive direct traffic to external sites, split between clicks on Google’s own properties, such as YouTube and Maps, and searches that end without any action.
There is now a new level of difficulty for those working on online visibility: approximately 27.2% of queries end without any user action. In other words, almost one in three people no longer clicks—either on the open web or on Google properties—because they find an immediate and satisfactory answer in the SERP and do not feel the need to explore further or navigate elsewhere.
This evolution has often fueled the perception of a crisis for websites and traditional SEO. However, Rand Fishkin, founder of SparkToro, urges us to read these figures carefully: “Fewer clicks does not mean fewer opportunities,” because traffic may decrease in absolute terms, but not necessarily the value of interactions or conversions—as we also discussed in our first meeting with SEOZoom Caffè.
According to Amsive‘s analysis, the change mainly affects the conversion funnel: users arrive at websites at more advanced stages of the decision-making process, reducing scattered traffic and increasing the intentionality of visits. Therefore, metrics such as conversion rate and average value per session become more reliable indicators of the effectiveness of online presence, surpassing simple raw traffic counts.
Google’s position: more qualified clicks and more informed interactions
It is interesting to note that this approach is similar to the “narrative” being put forward by Google’s public voices, which in recent weeks have alternated between ultra-futuristic descriptions of the benefits of AI—as in the videos shown at Google I/O 2025—and “defensive” and cautious statements about the transformation taking place in online search and accusations of worsening the zero-click effect.
In summary, the company argues that while the overall number of clicks to the open web is decreasing, the quality of interactions is improving.
Sundar Pichai himself has reiterated on several occasions that “AI Mode and AI Overview are part of a broader transformation: we want search to evolve to be more predictive, more contextual, and less dependent on click-based models.” According to Alphabet’s CEO, this work serves to “improve user satisfaction”: with the new features, people ask “longer and more complex questions and explore a wider range of websites than in the past.”
This line was also taken up by Elizabeth Reid, VP and Head of Search at Google, who in an interview with Indian Express used the same words as her boss: “We’re seeing more comprehensive searches. People are asking longer, more complex questions, and often clicking on higher-quality web pages, spending more time on those sites.“ This is a sign of a common ”attempt” to shift the focus from the quantity of clicks to the quality of queries and deeper engagement.
Reid added that “with AI Overview, users are learning to ask questions they didn’t feel confident asking before” and that, in the future, “the traditional search bar will become less prominent as voice and visual search grows” – echoing the message “let Google do the Googling for you” (let Google Google for you), a clear message about the concise and central role that its artificial intelligence wants to play in filtering and returning information, further abandoning historical modes of interaction.
Some insights also come from the business model side, as Reid pointed out that “AI Mode is experimental and we aim to test the integration of ads without compromising the user experience,” confirming that Google is proceeding cautiously with the introduction of new forms of AI-based monetization.
The topic was also central to Google Marketing Live 2025, where Jenny Cheng, Vice President of Google Merchant Shopping, said that when users “decide to click away, it’s a more qualified click… What we hope to see over time, and we don’t have any data to share on that, is more time spent on the site, which is what we see organically in a much more qualified visitor to the website.”
Cheng explained that tools such as AI Overview and AI Mode are designed to “offer consumers new ways to discover information and get answers to their most important questions,” promoting more natural interaction with the search engine, with more conversational queries and follow-up questions that refine the search.
However, as mentioned, Google has not yet provided independent or verifiable data to support the alleged higher quality of interactions and traffic generated by the new search modes. And in an article published by Search Engine Journal, less optimistic views emerge from former Google executives: “Driving traffic to publishers’ sites is a kind of necessary evil,” is the summary, reinforcing the idea that Google is shifting its focus toward more internal use within its own ecosystems, reducing its dependence on external traffic.
Why Google is betting on AI Mode: to strengthen its weakened dominance
Google’s insistence on AI Overviews and the introduction of AI Mode is obviously not accidental. According to official statements from the company, this is one of the most important launches in online search in the last ten years. And the first internal results seem to support this choice.
As reported by McMarketeronline, between September 2024 and April 2025, Google saw an increase of more than 10% in the use of its search engine for queries that activate AI Overview, both in the US and India. In other words, AI features are not only changing SERPs: they are increasing overall Google usage.
But the adoption of artificial intelligence in Search also responds to a competitive need. Data from Forrester Research’s Consumer Spending Pulse Survey, conducted in February 2025, shows that users are already exploring alternatives:
- 80% used Google in the last month for informational searches
- 41% used Amazon
- 32% used Instagram
- 31% used ChatGPT
- 25% used TikTok
- 4% have tried Perplexity.ai
In this context, AI Mode represents a strategic response: a conversational and multimodal search interface capable of handling complex questions, preferences, and constraints (time, price, location) and returning results already organized in the form of “expert” and cited content. According to Google, the goal is to prevent search intent from dispersing to other ecosystems and maintain Google’s centrality as the preferred point of access to online information — even in an era where the interface is no longer the classic search bar, but an AI assistant widespread in every context.
Concrete strategies for addressing the new competitive landscape
So, what can we do in this transition phase to maintain (or gain) visibility and acquire users? It is clear that a review of classic strategies is needed: changes in how information is accessed, the growth of zero-click searches, and the increasing presence of AI Overviews require rapid and effective adaptation. Responding to this scenario requires targeted action on multiple levels.
- Strengthening brand identity
The data shows that branded queries are less affected by AI Overviews, recording growing click rates. Building a recognizable and authoritative brand becomes both a defensive and offensive strategy. In a context where organic visibility is shrinking, the brand represents an anchor of recognition and trust, capable of influencing user choices even before they click.
- Optimize text for featured snippets
Even with AI Overview, featured snippets remain one of the few elements capable of maintaining a good ability to attract clicks. Structuring content clearly, answering frequently asked questions, and using suitable formats (lists, tables, concise paragraphs) increases the chances of being selected. This type of optimization still offers visibility on the first page and can reduce the negative impact of the new SERP organization; it also provides the type of information that AI engines like, so we can get a double benefit.
- Diversify your channels
Access to information is increasingly happening through alternative channels. Specialized forums, social platforms, AI-based environments: being present where users look for opinions, reviews, or additional information is becoming crucial. It is no longer just a matter of attracting traffic to the site, but of monitoring all touchpoints where purchasing or interaction decisions are made. It is no coincidence that, at SEOZoom day 2025, we embraced the idea of reformulating the acronym SEO to Search Everywhere Optimization, aware that multichannel is increasingly the keyword we need to do SEO today.
- Monitor the evolution of mobile search
Mobile remains the dominant channel for everyday search. With Safari controlling a significant share of traffic through iPhone devices and the potential for changes in the default search engine, it is essential to keep a close eye on this segment. The strategy must include a mobile-first approach, with sites optimized for speed, user experience, and compatibility with the best-performing rich media formats on mobile.
- Measure new KPIs focused on quality
The quantity of traffic is losing importance in favor of the quality of interactions. Metrics such as conversion rate, session duration, and average number of page views are becoming more reliable indicators of the effectiveness of visibility strategies. In an environment where clicks are fewer but more qualified, it becomes a priority to measure the real impact of visits and the value generated for the user.
- Focus on E-E-A-T to increase citability
In the new scenario, getting traffic is not just about ranking well, but about being selected as a reliable source by an AI engine. For this reason, E-E-A-T signals become crucial: experience, expertise, authority, and trustworthiness are no longer just guidelines, but real parameters used by AI to choose whether—and how—to cite a site in generated content. Google has stated that generative systems use E-E-A-T to select sources to include in AI Overview responses. Strengthening domain credibility, signing content with verifiable authors, and offering clear sources and structured content are becoming prerequisites for accessing new visibility. As several recent cases show, AI responses increasingly cite recognized sources, well-referenced technical content, and articles that offer real informational value. Focusing solely on keywords and structure is not enough: it is the overall reliability of the site and content that determines the new form of ranking — citability.
A new balance in online search
The spread of AI Overview, the increase in zero-click searches, and the growth in the use of AI-based tools are profoundly reshaping the dynamics of web traffic. User interaction with search engines no longer follows a single linear path but is fragmented into micro-moments of information, where answers often arrive without the need to visit a website.
At the same time, Google continues to be the hub of the information system. Data confirms steady growth in search volume — over 5 trillion in 2024 — and continued dominance over competitors, including those with generative technologies. The company says it aims to strike a balance between innovation and continuity: new response models, but maintaining the ability to generate qualified traffic for publishers. However, this balance has not yet been confirmed by independent data.
In the meantime, the signals gathered from reports — and from direct experience, such as that of our site — show a system that is changing but not collapsing. Traffic from AI tools is still marginal, mobile continues to dominate traditional search, and the role of the brand and content quality are becoming increasingly important for gaining visibility, even in a more crowded and disintermediated environment.
New operational priorities are therefore emerging: building “citeable” content for AI, optimizing not only for ranking but also for engagement, monitoring alternative channels, and critically reading new KPIs. Visibility on the web no longer depends solely on ranking, but on the ability to intercept intent in all its forms, including that conveyed by artificial intelligence.
