Korea E-commerce AI Search Conversion Comparison: Coupang, Gmarket, and 11Street's 2026 Strategies

The conversion data says it plainly

In August 2026, 11Street became the first Korean e-commerce platform to publish AI search performance figures. During the July 2026 pilot, its first month of live operation, item conversion rate (ICVR) ran more than twice as high as conventional integrated search^[1]. The implication is concrete: when AI shifts from listing results to analyzing purchase intent and narrowing choices before the user has to scroll, the path from search to checkout shortens in a measurable way.

Korea's three largest e-commerce platforms, Coupang, Gmarket, and 11Street, all embed AI in search, but on different axes. Coupang bets on automated product comparison; Gmarket on hyper-personalization; 11Street on general-purpose intent analysis^[2]. Public conversion data to rank these approaches head to head is scarce for now, but the strategic divergence is sharpening as the second half of 2026 gets underway.

Core terms

E-commerce AI search departs from keyword matching by reading natural-language queries, purchase context, and individual browsing history to surface the products closest to the buyer's actual intent. The goal is less friction during the browse phase and a shorter path to checkout.

Item Conversion Rate (ICVR) is the share of users who searched for a product through a given channel and completed a purchase. Comparing ICVR across channels for the same product in the same period is a direct measure of search algorithm quality.

Agentic commerce is the mode in which an AI agent accepts conditions from the user, then searches, compares, and recommends products on their behalf. Rather than presenting a ranked list, the AI participates in the decision itself^[3].

---

---

Platform feature comparison

(Source: ZDNet Korea, 브릿지경제, 다음뉴스, August 2026)

How each platform approaches AI search

Coupang: trimming the spec overload

Coupang's AI search focus is on reducing information overload at the point of comparison. For product categories where specs drive the decision, refrigerators, air conditioners, beds, the platform converts text-based spec sheets into AI-generated infographics automatically. The feature is limited to Rocket Delivery listings^[2].

The practical value is clear: buyers digest decision-critical information faster, which compresses browse time. The current constraint is that the feature covers only specific categories and a single fulfillment tier. Coupang has not published conversion rate figures, and whether the scope expands will be the main signal to watch^[2].

Gmarket: personalization at depth: still in development

Gmarket has the broadest stated ambition of the three, though as of August 2026 the feature has not launched. The platform is building a hyper-personalized AI agent that goes beyond the product name to synthesize image, price, reviews, and user experience data, all to surface what a shopper wants even when they have not stated it directly^[2]. Reflecting individual consumption patterns rather than generic category signals is the core differentiator.

With no launch date confirmed, evaluating outcomes is premature. That said, hyper-personalization compounds with the amount of behavioral data available, which advantages platforms with deep first-party histories.

11Street: intent analysis, narrowed to five products

Of the three, 11Street has shared the most specific performance data. When a user searches "laptop, " the AI proposes concrete conditions, "easy to carry, " "optimized for gaming", before showing products, then filters down to the five best matches for the chosen condition. Price competitiveness, delivery speed, and reviews all factor into that shortlist^[1][3].

During the July 2026 pilot, ICVR exceeded twice the rate of conventional integrated search. 11Street attributes this to the AI reducing browse-phase friction before the user ever sees a product page^[1]. Applying the feature across all categories, rather than limiting it to a single segment, is another point of departure from Coupang's approach.

For brands: AI visibility becomes a core metric

As platform AI takes more of the search decision, brands selling on these platforms face a new measurement question: how often does your product appear in AI recommendations, and in what context? Unlike ranked keyword results, AI search is more sensitive to the completeness and contextual fit of product data. Missing from the AI's recommendation pool is a harder exclusion than dropping a few positions in a keyword ranking.

The table below compares tools that measure AI search visibility in the Korean market, covering not only platform-native AI search but also brand exposure in external AI engines like ChatGPT and Perplexity.

(Public information as of August 2026. Pricing: contact each vendor)

For a broader comparison of GEO and AI visibility tools, see Global GEO/AEO Solution Landscape.

Four steps for brands on AI-powered platforms

Step 1: Structure your product data
Clean up product names, categories, specs, prices, and reviews into a format platform AI can parse reliably. Matching each platform's data schema is the prerequisite for every downstream AI optimization.

Step 2: Write for conditional and conversational queries
Audit product names and descriptions against queries like "lightweight laptop with long battery life." Stating the use case and context explicitly outperforms keyword stacking when an AI is filtering for intent.

Step 3: Measure AI visibility
Use a tool such as BOIDA to track how frequently your products and brand appear in AI recommendation results, and in which contexts. Without measurement, there is no optimization direction.

Step 4: Differentiate by platform
On Coupang, prioritize spec data quality. On Gmarket, sharpen review completeness for personalization context. On 11Street, build out conditional keyword structures that align with intent-based filtering. Each platform's AI rewards different inputs.

For how AI search intersects with D2C conversion strategy, see AI Search and E-commerce D2C Conversion in Korea.

Where things stand

Korea's three largest e-commerce platforms are each building AI search in a different direction, and the competitive gap widens from the second half of 2026 onward. 11Street is the only one to have disclosed conversion figures, more than 2× vs. conventional search, which confirms that intent-first AI search measurably shortens the path to purchase^[1][3]. Coupang and Gmarket have not released equivalent data as of August 2026^[2].

For brands, the expansion of AI-driven search is both an opening and a risk. A product absent from the AI's recommendation pool faces a more complete form of invisibility than a low keyword ranking. Structuring product data and measuring AI visibility are the place to start, now, not after the market settles.

Related: E-commerce Product Recommendation GEO Strategy, AI Search Conversion vs. Organic ROI, Naver AI Tab Commerce CTR Update