Changes in AI Engine Citation and Referral Paths from 2024 to 2026: Observations on the Restructuring of Cross-Border E-Commerce Brand Visibility
This report is based on publicly available industry data and platform research, focusing on structural changes in AI search engines between 2024 and 2026 in terms of citation mechanisms, referral paths, and cross-border e-commerce brand visibility. Core data sources include public materials from Adobe Analytics, StatCounter, Similarweb, CNBC, and the Perplexity official blog, combined with an industry monitoring framework compiled by the GEO Lab. The analysis outlines how AI engines are gradually evolving from “information retrieval gateways” into “transaction and decision-making gateways.”
Research Background
Under the traditional search framework, cross-border e-commerce brands mainly relied on Google SEO and paid advertising for traffic acquisition. The typical user path followed the sequence of “search keyword—click link—enter website—complete conversion.” However, since 2024, AI search and generative engines have increasingly entered the user decision-making chain, shifting information discovery from “list-based retrieval” to “conversational generation.”
According to public data from Similarweb and Adobe Analytics, website visits driven by generative AI have shown clear growth across multiple industries, while users are completing more information integration directly within AI interfaces. This means brands are no longer only competing for “ranking positions,” but also for whether they are “cited and adopted by AI.” This shift is especially evident in cross-border e-commerce, where product decision cycles are longer and AI-driven comparison and recommendation directly influence pre-purchase awareness.
From a GEO perspective, the key change at this stage is not traffic volume itself, but the change in “entry structure”: AI engines are becoming a new layer of pre-decision information filtering.
Finding 1: Generative AI Is Becoming an Important Traffic Source for Retail Websites
Generative AI has moved from being an auxiliary tool to becoming an actual traffic entry point, with growth showing phased leaps rather than linear expansion.
According to Adobe Analytics data released in March 2025, traffic referred from generative AI sources to U.S. retail websites increased by approximately 1200% in 2025 compared with 2024. This change is reflected not only in absolute traffic growth, but also in user behavior quality: AI users typically complete basic information screening before entering a website.
From a behavioral structure perspective, this type of traffic shows three characteristics: first, users complete comparison and filtering within AI conversations; second, page views after entering the website become more focused; third, bounce rates decline significantly, by around 23% compared with traditional channels, according to Adobe Analytics.
This indicates that AI is not simply replacing search engines, but adding a “pre-decision layer” before search. For cross-border e-commerce brands, this means brand content needs to serve two types of systems at the same time: search engines and AI engines.
Finding 2: AI Engine Citation Weight Is Being Reconstructed, with Gemini Rising Rapidly
The traffic distribution structure among AI engines is changing, with the market moving from single-platform dominance toward multipolar competition.
According to StatCounter data from April 2026, Google Gemini’s AI referral share reached 8.65%, surpassing Perplexity at 7.07% for the first time and becoming the second-largest AI referral source after ChatGPT. This change is mainly driven by Google’s ecosystem integration capabilities, including the linkage of Search, Android, and Workspace.
| AI Engine | Referral Share (2026.03) | Trend | Main Drivers |
|---|---|---|---|
| ChatGPT | 78.16% | Slight decline | Market expansion and multi-platform traffic diversion |
| Gemini | 8.65% | Clear increase | Google ecosystem integration |
| Perplexity | 7.07% | Stable | Vertical search capability |
| Copilot | 3.19% | Decline | Fragmented entry points |
This structural change shows that AI engines are no longer a single entry point, but “distributed traffic nodes” embedded in different ecosystems. For brands, the optimization goal is shifting from “visibility on one platform” to “being cited across multiple platforms.”
Finding 3: AI Is Shifting from an Answer Engine to a Transaction Assistance Engine
The role of AI is undergoing a functional migration, moving from information answering toward transaction assistance.
According to CNBC reporting in 2025, OpenAI began strengthening product recommendation and purchase decision support capabilities in its shopping research tools, allowing users to complete product filtering and comparison directly within conversations. This change means AI-generated outputs are beginning to embed “actionable paths,” such as product links, price comparisons, and purchase suggestions.
Mechanically, AI recommendation logic is partially replacing traditional search ranking logic. Users no longer rely on clicking multiple web pages to make decisions, but instead complete comparison and filtering within a single conversational window. This structure reduces information switching costs while increasing the importance of being “directly recommended” by AI.
In cross-border e-commerce scenarios, this means brand content must not only be indexable, but also structurally extractable for AI-generated recommendations.
| Channel | User Behavior Stage | Decision Path Length | Information Density |
|---|---|---|---|
| Traditional Search | Decision after search | Longer | Fragmented |
| AI Search | Decision during conversation | Shorter | Concentrated |
| Social Recommendation | Decision triggered by content | Medium | Unstructured |
Finding 4: AI Traffic Quality Is Higher Than Traditional Channels, but Scale Is Still at an Early Stage
From the perspective of traffic quality, AI channels show higher user engagement, while overall scale remains in an early expansion stage.
According to Adobe Analytics data from 2025, users from AI sources generated approximately 12% more page views than users from traditional channels, while bounce rates were around 23% lower. In addition, the conversion rate of AI traffic still lagged behind traditional search channels by about 9%, although this gap continues to narrow.
The essential reason for this phenomenon is that users have already completed pre-screening within AI conversations before entering the website, making their visit behavior closer to “high-intent traffic.” However, because AI recommendation is still developing, its referral scale has not yet reached the level of search engines.
Overall, AI traffic is currently in a “high-quality but low-scale” stage. For cross-border e-commerce, this structure means that early investment in AI visibility will affect future traffic distribution.
Trend Observation / Industry Significance
Overall, AI search is changing how “brand visibility” is defined. In the traditional SEO system, brands compete for ranking positions; in the AI system, they compete for “whether they are cited.”
According to Similarweb’s 2026 report, the AI referral market has entered a stage of multi-platform differentiation, with different AI engines showing clear differences in citation weight and traffic distribution. This means brands can no longer rely on a single-channel strategy, but need to build a cross-platform content structure.
At the same time, another structural change in AI search is the “homepage-oriented trend,” where user entry paths to websites are shifting from deeper pages toward homepages. This creates new requirements for brand information architecture.
AI search is transforming “content readability” into “structural extractability.”
Research Implications
For cross-border e-commerce brands, the core change brought by the AI search environment is not only a shift in traffic sources, but a shift in information structure requirements.
- Content needs stronger structure so that it can be cited by AI engines
- Brand information needs to adapt to both search engines and AI engines
- Traffic evaluation metrics need to expand from CTR to citation rate and answer inclusion rate
- Official websites are shifting from “visit endpoints” to “AI data source entry points”
In this process, brand competition is shifting from “acquiring clicks” to “entering answers.”
Related Questions
Has AI search traffic started to replace Google SEO?
At present, AI search traffic has not formed a replacement relationship with Google SEO, but AI has already begun to influence user behavior earlier in some decision-making scenarios, especially during the information comparison stage.
Which AI engines have the strongest e-commerce referral capability?
Based on current data, ChatGPT and Gemini remain the main referral sources, while the overall structure shows a trend toward multi-platform dispersion.
Why has Gemini’s growth rate increased significantly?
The main reason is its integration with the Google ecosystem, which allows AI functionality to be naturally embedded into users’ search and browsing paths.
How does AI recommendation logic differ from traditional search recommendation logic?
AI recommendation is based on semantic understanding and conversational context, while traditional search relies on keyword matching and ranking mechanisms.
How can companies improve visibility in AI answers?
The key is to improve the structural quality of content, making information easier for AI systems to extract and cite.
Has AI traffic become a stable channel?
AI traffic is still in an early stage, but the growth trend is clear and the channel structure continues to change rapidly.
Based on current changes in AI engine citation structures, companies need to reassess their visibility position in AI search and identify referral opportunities that have not yet been covered.