AI is entering the product discovery, research, and comparison process for retail consumers, but "more consumers using AI" and "AI has become a primary purchasing channel" are not the same conclusion. We've compiled publicly available data from Adobe, Similarweb, Salesforce, and Capgemini for 2025–2026, focusing on AI referral traffic, on-site engagement, purchase conversions, and website machine readability for US retail sites. Combined with consumer surveys, we aim to understand the shifts happening from product research to on-site purchase.
Research Background
In the past, discussions about the impact of generative AI on the retail industry often centered on whether consumers had started using ChatGPT, Gemini, Google AI Mode, Perplexity, or Google AI Overview to find product information. Entering 2025–2026, public data offers another perspective: after consumers arrive at a retail site from AI, how long do they stay, how many pages do they view, and do they complete a purchase? At the same time, AI referral traffic still represents only a small fraction of overall traffic for most retail sites. Therefore, when assessing channel value, it's important to distinguish between traffic scale, growth rate, and traffic quality. In other words, what's now more worth observing is not just whether AI can bring visits, but where these visits sit in the complete shopping journey.
AI Referral Traffic Grows While the Conversion Gap Shifts Direction
Between 2024 and 2026, AI referral traffic to US retail websites grew rapidly, and perhaps more notably, the purchase conversion gap between AI and non-AI visitors also changed. According to data published by Adobe on May 23, 2025, in July 2024, AI-referred traffic to US retail sites converted 43% lower than non-AI traffic; by February 2025, this gap narrowed to 9%. During the same period, from July 2024 to February 2025, AI referral traffic to US retail sites grew approximately 12-fold.
Another set of data released by Adobe on April 16, 2026, further shows that in Q1 2026, traffic from AI sources to US retail sites grew 393% year-over-year, with March 2026 seeing a 269% year-over-year increase. During the preceding 2025 holiday shopping season (November–December), AI referral traffic grew 693% year-over-year. These numbers indicate that off-site visits driven by AI as a product research entry point are still expanding, but growth rates alone do not mean AI already accounts for a significant proportion of a site's overall traffic.
Conversion trends have also shifted. According to Adobe's public data from 2026, in March 2025, AI traffic converted 38% lower than non-AI traffic, but by March 2026, it was 42% higher. It's worth noting that the "9% lower" figure for February 2025 and the "38% lower" for March 2025 come from Adobe data released at different times and should not be interpreted as a simple continuous monthly curve. A more reliable conclusion is that in year-over-year comparisons, AI referral traffic has shown signs of reversing its conversion disadvantage relative to non-AI traffic.
The table below is compiled from Adobe's public data from 2025 and 2026 to observe traffic growth and purchase conversion changes side-by-side.
| Time Period | AI Referral Traffic Change | AI vs. Non-AI Conversion Performance | Observation |
|---|---|---|---|
| July 2024 | Adobe timeline starting point | 43% lower | AI referrals skew toward visits after product research |
| February 2025 | ~12x increase from July 2024 | 9% lower | Conversion gap narrowed significantly |
| March 2025 | — | 38% lower | Adobe 2026 year-over-year comparison baseline |
| Nov–Dec 2025 | +693% YoY | — | Holiday shopping season AI referral growth |
| Q1 2026 | +393% YoY | — | AI referrals continue to expand |
| March 2026 | +269% YoY | 42% higher | Year-over-year conversion direction reverses |
AI-Referred Visitors Also Show Higher On-Site Engagement Than Non-AI Traffic
Purchase conversion doesn't happen in isolation. According to Adobe data from April 16, 2026, in March 2026, visitors to US retail sites arriving via AI sources had a 12% higher engagement rate, spent 48% more time per session, and viewed 13% more pages per visit compared to non-AI traffic. This suggests these users are not only more likely to complete a purchase but also exhibit deeper on-site browsing behavior.
Similar trends were already visible in earlier data. According to Adobe's May 23, 2025 release, in February 2025, AI-referred visits had a 23% lower bounce rate, 12% higher page views, and 41% longer session durations compared to other sources. Looking at session duration, page depth, bounce rate, and conversion performance together, a potential path shift emerges: some consumers may have already completed product information gathering, initial filtering, or comparison via AI before entering the retail site, meaning they could be closer to the purchase decision stage when they arrive.
Similarweb's data provides another comparison point. According to Similarweb e-commerce data from June 2025, AI-referred visits had a sales conversion rate of 11.4%, compared to 9.3% for paid search and 5.3% for organic search during the same period. However, Similarweb also noted that even among sites receiving significant AI referrals, this traffic typically still accounts for less than 1% of total site traffic. Therefore, at this stage, it's more appropriate to view AI referrals as a traffic source that is "small in scale but with quality worth monitoring," rather than evidence that traditional search has been replaced.
Decision cycles, price points, and information complexity vary significantly across retail categories, so these figures do not imply the same results for all products. The role of AI in research, comparison, and purchase may differ between low-cost fast-moving consumer goods and high-ticket durable goods.
Product Discovery Is Moving into AI, but Researching Products and Handing Over Purchase Authority Remain Two Different Stages
Behind the traffic changes, the consumer shopping journey itself is being reshaped. According to Capgemini consumer research released on January 9, 2025, 58% of consumers have used generative AI tools to get product or service recommendations, up from 25% in 2023. Additionally, 71% of consumers want generative AI integrated into their shopping experience. This indicates AI has moved from general-purpose Q&A tools into product recommendation and shopping research scenarios.
Salesforce data released on March 24, 2025, also shows that 39% of consumers have already used AI for product discovery, with the figure rising to 54% among Gen Z. Furthermore, 63% of Gen Z consumers expressed interest in having AI agents make purchases on their behalf. This data suggests younger consumers are more receptive to AI participating in the shopping process, but "expressing interest" still cannot be equated with actual purchasing behavior.
Breaking down the complete shopping journey, we can identify several stages: "Discovery → Research → Comparison → Click → Purchase → AI Agent Purchase." Existing public data supports a relatively clear assessment: AI is entering the discovery, research, and comparison stages and can already direct some consumers to retail websites. However, when AI transitions from being an information assistant to an agent capable of executing transactions, consumers have clear requirements regarding control.
According to a Salesforce 2025 survey, when building trust in AI agents, consumers prioritize data privacy and security, the ability to turn features on or off at any time, and confirmation by the consumer before a purchase is made. Therefore, between "let AI help me find products" and "let AI buy products for me," there remains a boundary shaped by privacy, authorization, and confirmation mechanisms. Gen Z can serve as an important group to observe for future shopping path changes, but their behavior cannot directly represent all consumers.
Demand Is Changing, but Machine Readability of Product Pages Still Varies
Consumers are increasingly using AI for product research, but this doesn't mean all retail brands will have equal AI referral opportunities. Adobe's 2026 analysis of US retail websites reveals significant differences in machine readability across page types. Adobe defines the AI Content Visibility Score as the proportion of a page's content that can be read by machines. In 2026, the average score for US retail homepage was 75%, category pages 74%, but individual product pages only 66%.
This difference is noteworthy because individual product pages often carry specific information like price, specifications, inventory, features, shipping, and purchase links, and they are also critical pages consumers verify before making a purchase. If this content is difficult for machines to read, AI understanding and citing of product information may be affected. However, machine readability should be considered just one variable explaining differences in AI visibility and referrals, not a single determining factor.
Adobe 2026 data also shows that among US retail homepages, one group with higher performance had an average AI Content Visibility Score of 82.5%, while another had 54.2%, indicating significant differences in readiness across websites. This suggests that the increase in AI usage on the demand side and whether websites on the supply side can be effectively understood by machines are two parallel changes that need to be observed simultaneously.
The table below compiles average AI Content Visibility Scores by page type from Adobe's public data in 2026.
| Page Type | Average AI Visibility Score 2026 |
|---|---|
| Homepage | 75% |
| Category Page | 74% |
| Product Page | 66% |
| Store Locator | 73% |
| Help Center | 79% |
| FAQ | 80% |
| Returns / Exchanges | 82% |
Trend Observations / Industry Implications
Looking at public data from 2024–2026, AI's impact on the retail shopping journey is shifting from "whether consumers use AI" to "which stages of the purchase decision AI takes on." Consumers can first ask AI about product differences, use cases, price ranges, or brand options, and then visit brand websites to confirm specific product information. This changes the traditional division of labor among search, content pages, and product pages. What's changing in AI shopping channels isn't just that referral traffic is increasing, but that more and more product research and filtering is happening in AI before consumers even reach a retail website.
The second change is how the AI channel is being evaluated. Looking only at referral volume makes it easy to overlook the fact that the channel's scale is still limited; looking only at conversion rates can ignore differences across categories, pages, and time periods. Similarweb's June 2025 data showed AI referral sales conversion rates at 11.4%, but related traffic typically still amounts to less than 1% of total site traffic. Keeping both metrics in view helps avoid misinterpreting high growth rates as channel maturity.
The third change is that consumer adoption speed and satisfaction with the experience are not moving in lockstep. According to Capgemini data released on January 9, 2025, consumer satisfaction with generative AI in shopping scenarios fell from 41% in 2023 to 37% in 2024, even as the proportion of consumers using AI for product and service recommendations increased. This indicates that increased AI usage doesn't mean issues with accuracy, control, or interaction in the shopping experience have been resolved.
Therefore, AI search's impact on the retail industry can no longer be explained solely by "whether a brand is mentioned." A more complete observation requires connecting AI visibility, AI referral visits, on-site engagement behavior, and purchase conversions into a single data chain, and then assessing how much role AI actually plays between product research and purchase.
Research Implications
For retail brands, establishing an AI referral monitoring framework at this stage is more likely to generate continuous data than simply observing a single AI search result. This is especially true when different entry points like ChatGPT, Gemini, Google AI Mode, Perplexity, and Google AI Overview coexist, as different sources can bring different traffic volumes, visitor behaviors, and purchase performances.
- Differentiate actual on-site visits from different AI sources, avoiding merging all AI traffic into a single channel.
- Continuously record AI referral traffic, purchase conversion rate, engagement rate, session duration, pages per visit, and revenue performance.
- Compare AI referrals with organic search, paid search, and other channels using consistent metrics.
- Break down data by product category, page type, and time period to reduce bias from single-month overall averages.
- Incorporate website factors like product page machine readability as analysis variables, cross-referencing with actual AI referrals and purchase outcomes.
- Focus on identifying any breaks in the "visibility → click → on-site behavior → purchase" funnel, rather than only monitoring one metric.
As AI moves from a product research entry point into purchase decisions, observing whether a brand is mentioned by AI alone makes it difficult to fully assess channel value. You can build a monitoring and analysis framework that connects AI visibility, referral visits, and purchase outcomes based on your own product categories, target markets, and on-site conversion data. If you'd like to design this metric system for your specific business, feel free to talk to us.
Related Questions
Is the conversion rate of AI referral traffic really higher than traditional search traffic?
Some public data has observed higher conversion rates. Similarweb's June 2025 data showed AI referral sales conversion at 11.4%, compared to 9.3% for paid search and 5.3% for organic search; Adobe's March 2026 data also showed AI traffic converting 42% higher than non-AI traffic. However, these results cannot be directly applied across different categories and websites.
How much traffic do AI tools like ChatGPT currently bring to retail websites?
Growth has been rapid, but the overall scale still needs careful assessment. Adobe recorded a 393% year-over-year increase in AI referral traffic to US retail sites in Q1 2026, while Similarweb noted that even among sites receiving significant AI referrals, this traffic typically still accounts for less than 1% of total traffic.
Why are AI-referred users more likely to make a purchase after entering a website?
One possible explanation is that some consumers have already completed information gathering, filtering, and comparison through AI before entering the site. Adobe's March 2026 data shows that AI-referred users have higher session durations, deeper page views, and higher engagement rates compared to non-AI traffic, consistent with this behavioral pattern.
Do consumers now use AI to search for and compare products?
Yes. Capgemini data from January 2025 shows that 58% of consumers have used generative AI to get product or service recommendations; Salesforce's 2025 data shows that 39% of consumers have already used AI for product discovery.
Does the machine readability of product pages affect brand visibility in AI search?
Machine readability can be a variable influencing AI understanding of product information. Adobe's 2026 data shows that the average AI Content Visibility Score for individual product pages on US retail sites is 66%, lower than the 75% for homepages and 74% for category pages, but it cannot be concluded that this is the sole determining factor for AI visibility.
What metrics should retail brands use to monitor the actual impact of AI search?
You can simultaneously track AI visibility, AI referral visits, engagement rate, session duration, page depth, purchase conversion, and revenue, broken down by AI source, product category, page type, and time period. This allows you to pinpoint whether the issue lies at the "being seen," "getting clicked," or "completing a purchase after arriving on-site" stage.