AI and Product Search: Q3 2025 Data

How AI Assistants Have Changed (Q3 2025)

While the first half of 2025 was dedicated to refining search and engagement, the third quarter (Q3) marked the beginning of the era of AI Agents. We are no longer just talking about chatbots that answer questions, but assistants capable of navigating the web and apps to complete complex tasks. In this quarter, the focus shifted from simple consultation to direct execution, reducing the distance between product discovery and the final purchase.
The main players introduced innovations that make e-commerce increasingly “invisible” and fluid:

• ChatGPT: Following the Q2 announcements, the partnership with Shopify has entered the operational phase. Q3 saw the rollout of native checkout, allowing a select base of users to purchase physical products using simple voice or text confirmation, with the AI autonomously managing the input of pre-saved shipping and payment data. Furthermore, the introduction of more advanced reasoning models (such as the o1 series) has enabled ChatGPT to handle extremely complex purchasing requests, such as planning an entire room’s furniture based on its dimensions.
• Gemini: Google has pushed the integration between visual intelligence and the YouTube Shopping ecosystem. In Q3, users began using Gemini to identify products within videos in real-time and add them to a universal Google cart. The “virtual try-on” feature expanded beyond clothing to the makeup and home furnishing sectors via augmented reality (AR) integrated directly into search.
• Claude (Anthropic): With the expansion of “Computer Use” capabilities, Claude surprised the market by allowing users to delegate the search for the “absolute best price” to the AI, which autonomously navigates between various comparison sites and official brand websites. Its technical analysis capacity has made it the preferred tool for high-budget purchases (professional electronics, technical sports equipment) where comparing technical specifications is decisive.
• Perplexity e SearchGPT: These AI search engines have transformed results into true “personalized shop windows.” In Q3, Perplexity introduced “Pro Shopping,” a feature that allows users to save their preferences (size, style, favorite brands) so that every product search is pre-filtered, eliminating the need to navigate dozens of filters on e-commerce sites.
• Mistral e DeepSeek: These have consolidated as “internal” engines for many independent retailers. Thanks to their efficiency, many e-commerce businesses have integrated these models to create proprietary virtual assistants with deep knowledge of the company catalog, offering customer service that is indistinguishable from a human’s.

These developments confirm that in Q3 2025, AI is no longer just a source of referral traffic but aims to create a sort of shopping operating system, where the assistant manages the entire funnel: from research and comparison to the final transaction.

The Impact on the E-commerce Sector (Q3 2025)

Aggregate data for the third quarter of 2025 indicates a consolidation of the AI channel, albeit with different dynamics compared to previous quarters. The volume of referral sessions from AI tools grew by 23.5% compared to Q2. This is a positive increase, but it highlights a slowdown in the growth rate compared to the first half of the year (+56.4% in Q1 and +46.7% in Q2).

This deceleration in the volume of external sessions can be attributed to a technological paradigm shift in the main platforms. With the spread of more advanced models, such as ChatGPT-5, there is a trend for LLMs to provide increasingly exhaustive answers within their own interface, sometimes reducing the frequency of outbound links to external sites.

This “zero-click” or direct response strategy seems to limit the overall volume of referral traffic, prompting the user to click only when there is a real intention to delve deeper or purchase.

Supporting this thesis, engagement and conversion metrics show more selective and targeted traffic:

• The average session duration dropped to 49 seconds (-2.3%), while the Bounce Rate improved significantly, settling at 29.9% (a -9.8% decrease). These data suggest that, despite spending less time on the site, the user lands on the page with a very clear idea of the product, having already received most of the preliminary information (prices, features, reviews) directly from the AI interface.
• The Add to Cart Rate (ATC) reached 5.7%, up from 4% in the previous quarter.

Main AI Referrals (Q3 2025)

The analysis of traffic sources in the third quarter of 2025 shows a landscape that is beginning to settle. ChatGPT maintains its position as the absolute leader, stabilizing at 86% of total sessions, the same share recorded in Q2. Despite the entry of new models and competitor updates, OpenAI‘s platform remains the primary reference point for users reaching e-commerce sites via AI.

However, significant changes are observed among other players:

• Perplexity saw its traffic share halved, dropping from 7% in Q2 to the current 3%
• Copilot and Gemini maintain marginal but steady shares at 2% and 1% respectively
• There is growth in the “Other AI referrals” category, which now represents 9% of sessions, indicating that a myriad of smaller tools and vertical assistants are starting to generate a non-negligible volume of traffic.

Q3 Revenue Analysis: Signals of Diversification

While in the second quarter the ability to generate revenue was almost exclusive to ChatGPT (94.5% of revenue), Q3 data shows a more articulated dynamic and a slight redistribution of value.

Although ChatGPT remains the primary source, its share of total revenue fell to 88.4%. This decline suggests that, despite maintaining the same traffic volume as the previous quarter, its absolute hegemony over monetization is beginning to be eroded by the efficiency of other tools.

The most relevant change concerns Perplexity: despite generating only 3% of sessions, its revenue share rose to 4.5%. This figure is particularly significant compared to Q2, where despite more traffic (7%), it generated less than 1% of revenue. This reversal indicates a marked improvement in the quality of traffic sent, with users much closer to the final conversion.

Other tools also show signs of vitality:

• The “Other AI referrals” category now contributes 5.7% of total revenue
• Gemini, despite having only 1% of traffic, accounts for 0.8% of revenue, showing a conversion capacity nearly equal to its session volume
• Copilot remains further behind, with a 0.6% revenue share against 2% of traffic.

Industries Most Impacted by AI Traffic (Q3 2025)

The evolution of commodity sectors during 2025 outlines a structural change in user search habits. While the beginning of the year was characterized by a polarization toward electronics and fashion, the third quarter consolidates a fragmentation of traffic toward more everyday and diversified purchases.

The most significant figure is the contraction of the Hard Goods (Consumer Electronics) sector, which continues its downward trend: from 34.6% in Q1 and 21.4% in Q2, it fell to 10.9% in Q3. This suggests that AI is no longer consulted primarily for technical and complex purchases but has become a support tool for a much wider range of products.

Conversely, the Retail & Shops sector has become the main beneficiary of referral traffic, reaching 20.2% (up from 14.5% in Q2). This increase indicates a maturation of the user who uses the AI assistant for general shopping searches and to locate points of sale, utilizing AI as a gateway to multi-channel retail.

Traffic distribution detail for Q3 2025:

Retail & Shops: 20,2% (growing)
Sports: 17,2% (steady growth)
Luxury: 17,0% (growing)
Fashion: 16,0% (stable/slight growth)
Hard Goods: 10,9% (declining)
Beauty: 9,3% (slight growth)
Home & Design: 7,4% (declining)
Food: 1,9% (newly monitored sector)
Kids: 0,1% (stable)

What This Data Tells Us: The Maturity of Q3 2025

The data from the third quarter of 2025 marks the definitive transition to maturity and the search for efficiency. The growth of referral sessions, while still increasing (+23.5%), shows a physiological slowdown that signals the normalization of AI as an acquisition channel. However, this quantitative slowdown is offset by a significant qualitative leap.

A sharply declining Bounce Rate (29.9%) and an Add-to-Cart Rate rising to 5.7% present an image of a user who no longer lands on an e-commerce site to “explore,” but to close. The slight dip in average session duration (49 seconds) is no longer a sign of lack of interest but the result of a pre-filtered experience: the user receives most of the information (prices, comparisons, technical sheets) directly from the AI interface and clicks the referral link only when ready to add the product to the cart.

The Q3 analysis also highlights that the slowdown in outbound clicks—potentially influenced by models like ChatGPT-5 that tend to resolve requests “in-app”—is the prelude to the era of agents. The challenge for e-commerce shifts: it is no longer just about “being linked,” but about being ready to receive orders from systems that are beginning to autonomously manage the checkout.

In conclusion, Q3 2025 delivers an ecosystem where AI acts as a powerful sieve: it sends less “distracted” traffic and more “decisional” traffic. For retailers, this means shorter but much more profitable sessions, provided they can integrate their offering into a conversational flow that increasingly aims for immediate transactions and less on traditional navigation.