AI and product discovery: Q2 2025 data

How AI assistants evolved (Q2 2025)

AI-powered assistants continue to evolve rapidly, enhancing their initial capabilities as simple interactive chatbots. In the first quarter of 2025, the focus was primarily on real-time web search integration and opening new purchasing channels. For example, ChatGPT had refined its ability to suggest products, while Claude had just introduced its web search function. Perplexity also maintained its strong integration with e-commerce platforms, while Gemini focused on AI Overviews in SERPs. Emerging models like Mistral and DeepSeek began to make their mark in the context of product search, albeit with still limited volumes.

The second quarter of 2025 marked an acceleration of these trends, bringing concrete innovations that radically transform the online shopping experience and user interaction. The main players have strengthened their platforms with targeted features:

• ChatGPT: The strategic partnership with Shopify is laying the groundwork to revolutionize e-commerce by working on the introduction of an integrated checkout. This upcoming feature is set to usher in the era of conversational commerce, as it will allow users to purchase products directly within the chat, without ever leaving the conversation flow.
• Gemini: It has evolved into a more proactive and multimodal shopping assistant, deeply integrating into the Google ecosystem. It has introduced features like a virtual try-on for clothing, which allows users to upload a photo to see how an item would look, and “smart shopping” that automates purchases when a desired price is reached.
• Claude: It has made web search accessible to all users, not just subscribers, making its responses more accurate and up-to-date. Furthermore, it has refined its analytical capabilities, with the new Sonnet 3.7 model capable of providing more detailed and reasoned answers, as well as improved mathematical analysis.
• Mistral and DeepSeek: Both have continued to improve their models and interfaces, focusing on efficiency and specific solutions for developers, proving they can compete on code and reasoning benchmarks.
• Copilot: It has solidified its position as a productivity assistant, integrating even more deeply with the Microsoft Office suite.

These developments highlight a clear shift from simple search to direct purchase facilitation and an increasingly personalized and multimodal user experience.

Ecommerce impact (Q2 2025)

The analysis of aggregate traffic for the e-commerce sector continues to highlight the increasingly established role of AI tools (like ChatGPT, Perplexity, Gemini) as referral sources. After a 56.4% growth recorded in the first quarter of 2025 compared to the end of 2024, the trend stabilized in the second quarter, with a further 46.7% increase in sessions compared to Q1.

While the data from the beginning of the year showed very positive engagement signals, such as a 12.9% increase in average session duration, the second quarter offers a more mature and, in some ways, more interesting view. Despite a slight decrease in average duration (-7% compared to Q1), purchase intent has improved.

The Add to Cart Rate (ATC), in fact, continued its growth: after reaching 3.3% in Q1 (with a +101.2% increase also influenced by winter sales), it reached an excellent 4% in Q2. This not only represents a 41% improvement over the previous quarter but also suggests that traffic from AI is becoming increasingly qualified and prone to conversion, even in the absence of sales or other seasonal promotions.

Top AI referral sources (Q2 2025)

Analyzing traffic sources, the landscape updated to Q2 2025 sees ChatGPT confirming its dominant position, albeit with a slightly lower market share, dropping from 89.3% of sessions in Q1 to the current 86%. This slight decline has been absorbed by other AI tools such as Copilot, which is emerging as a new key player, capturing 2% of referral traffic and surpassing Gemini, whose share has dropped to 1%.

The most interesting dynamic, however, emerges from the revenue analysis. Despite the drop in traffic volume, ChatGPT’s dominance as a revenue source has strengthened, increasing from 93.75% of the total in Q1 to about 94.5% in Q2. This figure suggests a high quality and conversion propensity of users coming from this platform.

In contrast, other tools show a lower performance relative to their traffic weight. Perplexity, despite generating 4% of sessions, contributes less than 1% of the revenue, a significant drop from 4.16% in the first quarter. Copilot and Gemini also follow at a great distance, underscoring that, at the moment, the ability to monetize AI traffic is almost entirely concentrated in the hands of ChatGPT.

Industries most affected by AI traffic (Q2 2025)

The analysis of industry sectors impacted by AI traffic has seen a shift over the course of 2025. While the first quarter was clearly dominated by Hard Goods (34.6%) and Fashion (26.6%), which together accounted for over 60% of sessions, Q2 reveals a much more diversified and balanced landscape.

These two sectors, while still relevant, saw their traffic share more than halved, dropping to 21.4% and 14.8%, respectively. This contraction does not indicate an absolute drop in interest, but rather an expansion of users’ search areas.

The real news of the second quarter is, in fact, the growth of other categories. The Luxury and Retail & Shops sectors grew from about 6% each to a much more significant 14.5%, nearly matching the share of Fashion. The Beauty sector also saw a surge, tripling its incidence (from 2.96% to 8.4%). The Home & Design sector also consolidated its position, rising to 10.6%, while the Sports category proved to be the most stable, remaining around 15% in both quarters.

This rebalancing suggests that the user who relies on AI for purchasing advice is maturing, expanding their searches from “traditional” sectors like electronics to a much broader range of interests, including luxury, beauty, and general retail.

• Hard Goods (Consumer Electronics): 21.40% of sessions
• Fashion: 14.8% of sessions
• Sports: 14.9% of sessions
• Home & Design: 10.6% of sessions
• Luxury: 14.5% of sessions
• Retail & Shops: 14.5% of sessions
• Beauty: 8.4% of sessions
• Kids: 0.2% of sessions

What the Q2 2025 data tells us

The data from the second quarter of 2025 marks a shift from a phase of adoption to one of maturation in the use of AI for e-commerce. The most significant change is no longer in the volume of traffic, which is stabilizing while still growing, but in its quality.

The decrease in average session duration, combined with a decisive increase in the Add-to-Cart Rate, indicates that the user is now more aware and targeted: they use AI for specific searches with a clear purchase intent. This translates into traffic with higher conversion potential.

We are also witnessing a dual market movement: on one hand, user interests are diversifying into new sectors like Luxury and Retail; on the other, the ability to monetize this traffic is becoming even more sharply concentrated on ChatGPT. For the future, the analysis will increasingly shift from “how much traffic” to “what kind of traffic,” monitoring the ability of new AI players to close the conversion gap and the impact of these dynamics on the entire sales funnel.