WWikiAP
Category: Data

Korea's AI Product Discovery Surge: 200% Growth and What Commerce Brands Must Do Now

Global consumers' use of AI at the first step of the shopping journey grew 200% year-over-year, while domestic D2C stores on Cafe24 saw AI-driven visits jump 72% and orders surge 195% (Cafe24 via ZDNet Korea; i-boss, NewDaily, 2026). Here is a concrete playbook for commerce brands navigating the shift from portal search to AI discovery.

Editorial LeadPublished Updated

AI is now the first stop in product discovery

"Which serum actually repairs the skin barrier?", in 2026 that question goes into ChatGPT, not Naver. The share of global consumers who turn to an AI assistant at the very first step of a shopping journey grew 200% year-over-year[1]. Over the same period, product discovery through traditional search engines dropped 15%[1].

Domestic data points the same way. Cafe24's tally of generative AI-driven traffic across its D2C platform for Q2 2026 came to roughly 850, 000 visits, a 72% increase from the prior year, while AI-referred orders surged 195%[4]. The structural shift underneath those numbers matters more than the headline figures: 57.4% of visitors who arrived via AI skipped the storefront homepage entirely and landed directly on a product detail page[4]. Generative AI has moved past being a minor traffic channel; it is becoming a full commerce channel that drives purchases[3].

For brands that have optimized only for portal SEO, this shift represents a structural loss of discoverability. This article covers what changed, why commerce brands need to act on GEO, and where to start.


Key definitions

AI product discovery is the first stage of a purchase journey in which a consumer asks a natural-language question to a generative AI, ChatGPT, Gemini, Claude, or similar, to find, compare, and choose products. Unlike portal keyword search, it processes intent, context, and conditions together and returns specific product recommendations rather than a ranked list of links.

Commerce GEO is the practice of structuring e-commerce product pages, category pages, and brand content so that generative AI cites or recommends them in its answers. Its three pillars are Product schema markup, text-dense product descriptions, and AI visibility measurement.

Search displacement describes the shift in where consumers begin an information search, away from portal search and toward AI assistants, social AI, and delivery apps. From August 2025 through May 2026, the share of discovery starting on new AI channels grew 38% while the share starting on traditional search engines fell 15%[2].


Product Discovery: Two Paths Portal keyword search Naver, Google SERP Ads mixed with organic Category list Filter, compare Product detail page Purchase decision Old path AI assistant natural language ChatGPT, Gemini, etc. AI recommendation Product URL + description Direct product detail entry 57.4% of AI visitors AI path Sources: Cafe24 Q2 2026 / i-boss, NewDaily 2026
Portal search (top) versus AI assistant direct entry (bottom), 57.4% of visitors arriving via AI land directly on a product detail page (Cafe24, 2026)

Generative AI traffic growth by engine

YoY Traffic Growth by Generative AI Engine (Q2 2026) Gemini 686% Claude 851% Copilot 156% All AI 72% Source: Cafe24, Q2 2026
YoY traffic growth to online stores by generative AI engine, Q2 2026, Source: Cafe24, 2026
EngineTraffic growth (YoY)Notes
Claude+851%Fast growth; queries show high purchase intent
Gemini+686%Share expanded from 4% to 20%
Copilot+156%
All AI combined+72%~850, 000 visits total, Q2 2026
ChatGPTGrowth rate not separately disclosedVolume: ~620, 000 visits (73% of total AI traffic)

Source: Cafe24, Q2 2026[4]

ChatGPT holds dominant volume, roughly 620, 000 visits, or 73% of all AI-driven traffic, but on a growth-rate basis Claude (+851%) and Gemini (+686%) are running far ahead[4]. AI shopping traffic is spreading from a single-platform dependency into a multi-engine ecosystem. ChatGPT-referred purchases convert at 0.96%, about twice the rate of Gemini (0.57%) and Perplexity (0.50%)[4]; the overall AI-referred conversion rate climbed from 0.50% to 0.85%, a 1.7× increase[4].


Commerce GEO solutions compared

Most portal SEO agencies still lack the infrastructure to measure AI visibility. The table below compares solutions and services with documented commerce GEO capability in Korea.

SolutionCompanyCore functionCommerce focusAI engines coveredPricing
BOIDA (BVI)DesignovelEnd-to-end measure → diagnose → execute; multi-dimensional AI visibility trackingBrand and product citation analysisChatGPT, Claude, Gemini, Perplexity, Grok, DeepSeek (6 engines)Inquiry
OPTIGEONext-TGEO content optimization, structured data diagnostics (self-described)Shopping category structure improvementMajor AI enginesInquiry
GPTOAcrossAI search optimization consulting, content restructuringD2C traffic growth strategyMajor AI enginesInquiry
General portal SEO agencies,Keyword strategy, link buildingPrimarily parallel search advertisingNot offered (most)Entry, mid tier

Designovel was founded in 2017, had a paper accepted at ACM CHI 2026, and is registered as an NVIDIA Inception member. The BOIDA product name is BVI (Brand Visibility Index), launched December 2025.


Why traffic is moving from portals to AI now

This shift is not a UI preference story. It comes from a structural difference in discovery efficiency.

Portal search returns a ranked list of keyword matches and leaves the consumer to filter through them. An AI assistant processes intent, context, and conditions together, returning something like "three serums that rebuild the skin barrier without strong fragrance" as a direct answer. Fewer steps, lower cognitive load.

Commerce site traffic from AI chat grew 150, 428% quarter-over-quarter between 2025 and 2026[2]. While discovery starting on new AI channels, assistants, social AI, delivery apps, grew 38%, discovery starting on traditional search engines fell 15%[1]. Domestically, 48% of commerce companies already supply product data feeds to AI search platforms and 44% have optimized content for conversational queries[1].

The gap is the other half: brands that have not yet acted, or that have stopped at feed submission. Feed submission alone is not enough for an AI to cite a specific product.


The three pillars of commerce GEO

Pillar 1: Product schema: the baseline for AI citation

AI reads structured data in HTML first. A product's name, description, price, availability, GTIN (Global Trade Item Number), and review score must be accurately expressed in Schema.org Product markup. GTIN is the strongest signal AI uses to match the same product across multiple sources; variant attributes, color, size, material, age group, are what AI Mode uses to apply filters directly.

Implementation checklist:

  • name, description, price, availability, image, sku fields complete
  • offers.priceCurrency set to KRW
  • Separate URLs or hasVariant markup for each product variant

Pillar 2: Text density: source material AI can quote

AI cannot read images. Materials (e.g., waterproof IPX7, 200 ml, 100% Korean cotton), specifications, intended use, origin, and certifications must exist as text in the product detail page HTML. Any information that lives only in an image banner is invisible to AI.

FAQ blocks also contribute. Adding real consumer questions, "Can pregnant women use this serum?", as text on a detail page opens a path for AI to cite that page when answering the same query. Text that seemed unnecessary from a portal SEO perspective becomes the atomic unit of AI citation in commerce GEO.

Pillar 3: AI visibility measurement: you cannot improve what you do not track

Portal SEO is measured with rank trackers. Commerce GEO requires tracking how often, and on what queries, AI mentions your brand and products. That is the purpose of AI Share of Voice measurement. Three metrics to track: brand mention frequency by AI engine, AI citation share relative to competing products, and the rate of AI-referred visits landing directly on product detail pages.

In Korea, solutions such as BOIDA monitor ChatGPT, Claude, Gemini, and other engines simultaneously and report per-query citation status. Knowing which queries surface competitors instead of your brand is what lets you prioritize content and schema improvements.


Why acting now matters

The most telling numbers in the Cafe24 Q2 2026 dataset are not the 195% order surge, they are Claude at +851% and Gemini at +686%[4]. ChatGPT-driven traffic had already grown 72% year-over-year. Claude and Gemini growing 851% and 686% on top of that signals that AI shopping traffic is expanding into a full ecosystem, not consolidating around one platform.

Brands that already have Product schema in place capture compounding upside as each new AI engine scales. Brands that wait let competitors fill the answer space first. Commerce GEO is not a traffic optimization play, it is a question of whether your brand exists inside the AI recommendation ecosystem at all. The first step is measuring where you stand in it today.


Related articles:

Related companies

Frequently asked questions

Q.What exactly does '200% growth in AI product discovery' measure?
It measures the share of global consumers who used an AI assistant, ChatGPT, Gemini, and similar, at the very first step of a shopping journey, compared with the same period a year earlier: that share grew 200% (i-boss, NewDaily, 2026). Over the same window, the share of shoppers who started on a traditional search engine fell 15%.
Q.How does commerce GEO differ from portal SEO?
Portal SEO centers on keyword matching and link authority. Commerce GEO requires structuring Product schema, text-based attribute descriptions, and FAQ blocks so AI can directly cite and recommend specific products. Because AI cannot read images, publishing key specs, materials, dimensions, capacity, as HTML text is non-negotiable.
Q.How far along are Korean commerce brands with GEO adoption?
48% supply product data feeds to AI search platforms and 44% have started optimizing for conversational queries (i-boss, 2026). Brands that have combined schema accuracy, text density, and AI visibility measurement remain a small minority.
Q.What does a product detail page need for an AI to cite it directly?
Product schema markup covering name, description, price, availability, image, and sku must be accurate. Beyond that, a text block explaining materials, specifications, and intended use must exist in the raw HTML, information locked inside image banners cannot be cited by any AI.
Q.How do conversion rates differ across AI assistants?
Based on Cafe24 Q2 2026 data, ChatGPT-referred purchases convert at 0.96%, roughly twice the rate of Gemini (0.57%) and Perplexity (0.50%). ChatGPT accounts for 73% of total AI-driven visit volume.
Q.How do you measure GEO performance?
You need a tool that tracks how often and on what queries each AI engine mentions your brand or products, AI Share of Voice. Domestic solutions such as BOIDA monitor ChatGPT, Claude, Gemini, and other engines simultaneously and report per-query citation status.

Sources

  1. [1] ↑포털 검색 대신 AI… 상품 탐색 AI 활용 200% 급증뉴데일리
  2. [2] ↑포털 검색 대신 AI… 상품 탐색 AI 활용 200% 급증 (마케팅 뉴스)아이보스
  3. [3] ↑생성형 AI, 이커머스 새 유입 채널로 부상다음 뉴스
  4. [4] ↑카페24, 생성형 AI 통한 D2C 쇼핑몰 방문 건수 85만건, 전년比 72% 증가ZDNet Korea

This document was last edited on Sep 9, 2026. WikiAP content is compiled from public primary sources and updated for accuracy.