Naver Shopping AI Agent vs ChatGPT Shopping: A Data Readiness Checklist for Korean Brands
A side-by-side comparison of the Naver Shopping AI agent and ChatGPT Shopping across product data sources, onboarding paths, required fields, and measurable metrics, plus a checklist of what Korean brands should prepare first when getting ready for both.
Korean brands getting ready for both the Naver Shopping AI agent and ChatGPT Shopping need to understand one thing up front: the two platforms get their product data in opposite ways. Naver uses the product information, reviews, and shopping history that already live on its own platform[1]. ChatGPT uses feeds that merchants send in OpenAI's format[5]. That single difference changes the onboarding path, the data you must supply, and the metrics you can check afterward. This guide sets the two platforms' requirements side by side and lays out a checklist for what to build, in what order. For how agentic commerce optimization (ACO) differs from GEO and AEO, see the hub article ACO vs GEO vs AEO.
30-second definitions: what each platform is
- The Naver Shopping AI agent is a conversational shopping feature in the Naver Plus Store app that summarizes, compares, and recommends products using Naver's internal product data and the user's shopping history (NAVER, 2026)[1].
- ChatGPT Shopping is a shopping feature inside ChatGPT conversations that finds and displays products using feeds submitted by merchants (OpenAI, 2026)[6].
- Naver EP (Engine Page) 3.0 is the product database URL format an independent online store uses to send its full catalog to Naver Shopping (NHN Commerce Help Center)[4].
- The OpenAI Product Feed Spec is the JSONL, CSV, or TSV feed format merchants use to send product information to ChatGPT (OpenAI, 2026)[5].
The readiness paths at a glance
Naver Shopping AI agent vs ChatGPT Shopping: readiness requirements compared
| Item | Naver Shopping AI agent | ChatGPT Shopping |
|---|---|---|
| Launch status | Beta 1.0 on 2026-02-26; proactive conversation feature announced 2026-06[1][2] | Available to US users, with more regions planned[6] |
| Product data source | Naver internal product information, reviews, personal shopping history[1] | OpenAI-format feeds submitted by merchants[5] |
| Additional signals | Clicks, wishlist saves, cart activity, and trends[2]; blog and Cafe UGC[3] | OpenAI says rich data such as images, prices, and reviews helps visibility[6] |
| Onboarding path | Smart Store; independent stores register an EP 3.0 database URL in Shopping Partner Center[4] | Apply at chatgpt.com/merchants; Shopify and Etsy merchants are already integrated[6] |
| Delivery method | Naver pulls the database URL on its own collection schedule[4] | SFTP, API, commerce platforms, feed providers[6] |
| Required data | Image and an actual price (no placeholder text in place of price)[4] | item_id, title, description, url, brand, seller_name, image_url, availability, price, target_countries, store_country[5] |
| Checkout | Aims for checkout inside the conversation (A2A commerce)[3] | Merchants may use their own checkout; Instant Checkout scaled back[7] |
| Access for Korean brands | Domestic sellers are the default audience | Standard uploads target the US; other markets need OpenAI confirmation[5] |
| Metrics sellers can see | No agent-specific metrics confirmed as public[1][2] | GA4 ai-assistant medium[8]; Business Hub mentions, clicks, topics[9] |
The row that matters most is "Required data." Naver already holds the data, so a seller's work comes down to keeping product information and reviews in good shape. On ChatGPT the seller has to export the data, so matching the field spec comes first.
Evidence: what the published numbers say
Naver has published comparatively more figures. According to Bizwatch, Naver Shopping's data, covering product information, user reviews, and shopping history plus blog and Cafe UGC, adds up to about 10 billion records, and about 600,000 seller stores operate on Smart Store (Bizwatch, 2026)[3]. The same report says AI recommendations accounted for more than 50% of in-app transaction value for the first time in May 2026, returning users grew 180% in the three months after the agent launched, and over 70% of agent users ask in full sentences instead of short keywords (Bizwatch, 2026)[3]. Naver says it trained "Shopping Intelligence," a commerce-specific LLM, on price, shipping information, product attributes, and user preference data (NAVER, 2026)[1].
OpenAI has published a spec rather than numbers. Its feed spec says the is_eligible_search flag is on when omitted, while is_eligible_checkout is off by default and requires a separate checkout integration plus seller_privacy_policy and seller_tos URLs (OpenAI, 2026)[5]. In March 2026 OpenAI said "the initial version of Instant Checkout did not offer the level of flexibility we aspire to," and moved to let merchants use their own checkout while it concentrates on product discovery (Retail Gazette, 2026)[7].
| Published figure | Value | Source |
|---|---|---|
| Naver Shopping data volume (including UGC) | About 10 billion records | Bizwatch, 2026[3] |
| Smart Store seller stores | About 600,000 | Bizwatch, 2026[3] |
| Share of in-app transaction value from AI recommendations, May 2026 | Over 50% | Bizwatch, 2026[3] |
| Growth in returning users, 3 months after agent launch | 180% | Bizwatch, 2026[3] |
| Share of agent users asking in full sentences | 70% or more | Bizwatch, 2026[3] |
| Required base fields in the ChatGPT feed (including region tags) | 11 | OpenAI Product Feed Spec, 2026[5] |
Why the requirements split: cause, effect, action
Cause. One company runs Naver's product listings, reviews, payments, and member history. ChatGPT has no product catalog of its own, so it depends on merchant feeds.
Effect. On Naver, the variables a brand controls are data already on the platform: product names, attribute values, detail pages, reviews. With more than 70% of users asking in full sentences (Bizwatch, 2026)[3], a product missing use-case and situation attributes, such as a query like "a sofa for a home with a dog," is likely to drop out of the comparison set. On ChatGPT, a feed that fails the spec never gets a chance to be considered. Standard uploads also target the US, so a Korean brand needs an English title and description plus US shipping terms before it can expect real visibility (OpenAI, 2026)[5].
Action. Fix data quality first on Naver and data delivery first on ChatGPT. Both jobs should run off the same source data (a PIM or product master) so price and inventory never disagree between the two.
Data readiness checklist for Korean brands
| Step | Check | Naver | ChatGPT | Done when |
|---|---|---|---|---|
| 1 | Single product master | Required | Required | Price, inventory, and images come from one source |
| 2 | Clean up product names and attributes | Required | Required | Material, use case, size, and target buyer are all filled in |
| 3 | Collect and respond to reviews | Required | Recommended | Each product keeps getting reviews and questions get answers |
| 4 | Register EP 3.0 database URL (independent store) | Required | N/A | Shopping Partner Center shows the status as receiving[4] |
| 5 | Build an OpenAI-format feed | N/A | Required | All 11 required fields export as JSONL or CSV[5] |
| 6 | Write English titles and descriptions | N/A | Required | English copy matches how US users phrase queries |
| 7 | Apply via merchants form or connect Shopify | N/A | Required | Application lists headquarters country, categories, and SKU range[6] |
| 8 | Decide on checkout path | Platform checkout | Own checkout first | Checkout integration choice and policy URLs are set[5][7] |
| 9 | Set up measurement | Supplement with external tools | GA4, Business Hub | The ai-assistant medium report populates[8] |
Accounts are where independent stores most often get stuck on EP registration. Shopping Partner Center requires a Naver Shopping ID, and a Smart Store account will not get you in. After you request registration, review usually takes 1 to 2 business days, and products with no image or with placeholder text where the price should be are left out of the EP (NHN Commerce Help Center)[4].
Measurable metrics compared: what you can and can't see
Measurement is where the two platforms differ most. On the ChatGPT side, GA4 groups ChatGPT, Gemini, DeepSeek, Copilot, and Grok referrers under the ai-assistant medium (Google, 2026)[8]. Business Hub shows mentions, website clicks, and discovery topics for a website URL in 28-day windows, but it has no prompt text or competitor comparison and covers ChatGPT only (Profound, 2026)[9]. Naver's two press releases say nothing about agent performance metrics for sellers (NAVER, 2026)[1][2]. On the Naver side, the practical approach is to track Smart Store sales trends alongside external AI visibility measurement. For KPI design in general, see ACO measurement metrics.
| Measurement tool | What it measures | ChatGPT coverage | Naver coverage | Notes |
|---|---|---|---|---|
| GA4 AI assistant channel | AI referral sessions, conversions | Yes | No | Excludes Google AI Overviews and AI Mode[8] |
| ChatGPT Business Hub | Mentions, clicks, discovery topics | Yes | No | 28-day windows; values below threshold are blank[9] |
| Profound | Brand visibility across multiple AI engines | Yes | Check vendor docs | Global tool, mid-tier (public pricing, subject to change) |
| Peec AI | Brand visibility across multiple AI engines | Yes | Check vendor docs | Global tool, entry-level (public pricing, subject to change) |
| Otterly.ai | AI search monitoring, GEO Audit | Yes | Check vendor docs | Global tool, entry-level (public pricing, subject to change) |
| Next-T | OPTIGEO | Check vendor docs | Check vendor docs | Korean tool, Inquiry |
| BOIDA | BVI, connecting measurement to diagnosis and execution | Yes | Positions itself as supporting Korean and domestic engines | Korean tool (operated by Designovel), tracks ChatGPT, Claude, Gemini, Perplexity, Grok, DeepSeek[10], Inquiry |
External tools cannot replace a platform's internal data. What they do is measure, repeatedly and from outside the platform, "which queries surface our brand in AI answers, and how often," so you can compare before and after a data cleanup. For a broader comparison of tools, see GEO recommended companies.
Execution: a 6-week readiness plan
- Weeks 1 to 2, audit the product master: Confirm price, inventory, images, and attribute values match across channels, and fill in empty attributes.
- Weeks 2 to 3, Naver data quality: Add use case and target buyer to product names, and plan review collection starting with the products that have the fewest reviews. Independent stores should check their EP 3.0 registration status[4].
- Weeks 3 to 4, draft the OpenAI-format feed: Build a JSONL feed with the required fields, including target_countries and store_country[5].
- Weeks 4 to 5, application and checkout policy: Apply at chatgpt.com/merchants and decide whether to keep your own checkout[6][7].
- Weeks 5 to 6, measurement baseline: Record pre-cleanup baseline values with the GA4 ai-assistant report and external visibility measurement[8].
For more on the Naver agent itself, read Naver agentic commerce and the AI shopping agent. Engine-by-engine priorities are in AI shopping engine comparison and GEO priorities, and feed details are in the ChatGPT Shopping feed setup guide.
Frequently asked questions
What is the biggest difference between the Naver Shopping AI agent and ChatGPT Shopping?
Who holds the data. Naver's agent uses product information, reviews, and user shopping history already inside Naver[1]. ChatGPT Shopping uses product data that merchants send in OpenAI's feed format[5].
Can a Korean brand list products on ChatGPT Shopping today?
You can apply, but shopping is currently offered to US users[6]. Standard uploads under the feed spec also target the US, and for other markets OpenAI has to confirm the integration and the allowed countries and currencies[5].
Does selling on Smart Store automatically put my products in front of the Naver Shopping AI agent?
Naver has not disclosed which sellers the agent draws candidates from or what its exposure criteria are. What is confirmed is that it uses product information, reviews, shopping history, price, shipping, and product attribute data[1].
How do I connect an independent online store to Naver Shopping?
Request registration of an EP 3.0 full-catalog database URL in Shopping Partner Center and wait for review. Review usually takes 1 to 2 business days[4].
How do I measure AI shopping performance on each platform?
For ChatGPT, the GA4 ai-assistant medium[8] and Business Hub[9] show part of the picture. No public metric reporting Naver agent-driven performance to sellers has been confirmed[2].
Which platform should I prepare for first?
If most of your revenue comes from Korea, Naver data quality comes first. A ChatGPT feed built now, before the Korean market opens, can be submitted as soon as you apply[5].
Wrap-up
Winning on the Naver Shopping AI agent depends on the quality of data the platform already holds. Winning on ChatGPT Shopping depends on whether the feed you send meets the spec. For Korean brands, Naver is where revenue comes from right now. ChatGPT is opening in the US first, so the work there is preparing an English feed and getting in line to apply. Both efforts have to start from the same product master, or price and inventory will drift apart. Only the ChatGPT side offers some measurement, so on Naver, watch sales trends together with external AI visibility measurement.
Related companies
- 넥스트티 (Next-T, OPTIGEO)SEO, GEO, AEO 컨설팅, 자동화
- 보이다 (BOIDA)생성형 검색 최적화(GEO) 솔루션, AI 가시성 측정
- Otterly.aiAI 가시성 모니터링 툴
- Peec AIAI 가시성 모니터링 플랫폼
- ProfoundAI 가시성 모니터링 플랫폼
Frequently asked questions
- Who holds the data. Naver's agent uses product information, reviews, and user shopping history already inside Naver. ChatGPT Shopping uses product data that merchants send in OpenAI's feed format (NAVER, 2026; OpenAI, 2026).
- You can apply at chatgpt.com/merchants, but OpenAI says shopping is available to users in the United States, and standard uploads under the feed spec also target the US. For other markets, OpenAI has to confirm the integration and the allowed countries and currencies (OpenAI, 2026).
- Naver's press releases do not say which sellers' products the agent considers or what the exposure criteria are. What is confirmed is that it uses product information, reviews, shopping history, price, shipping, and product attribute data (NAVER, 2026).
- Log in to Shopping Partner Center with a Naver Shopping ID, request registration of an EP 3.0 full-catalog database URL, and wait for review. Review usually takes 1 to 2 business days, and products with no image or with placeholder text in place of a price are dropped from the EP (NHN Commerce Help Center).
- For ChatGPT, the GA4 ai-assistant medium and Business Hub's mentions, website clicks, and discovery topics show part of the picture. The press releases do not show any public metric that reports Naver agent-driven performance to sellers separately (Google, 2026; Profound, 2026).
- If most of your revenue comes from Korea, start by cleaning up your Naver product data and reviews. For ChatGPT, building an OpenAI-format feed now, before the Korean market opens, means you can submit it as soon as you apply (OpenAI, 2026).
Q.What is the biggest difference between the Naver Shopping AI agent and ChatGPT Shopping?
Q.Can a Korean brand list products on ChatGPT Shopping today?
Q.Does selling on Smart Store automatically put my products in front of the Naver Shopping AI agent?
Q.How do I connect an independent online store to Naver Shopping?
Q.How do I measure AI shopping performance on each platform?
Q.Which platform should I prepare for first?
Sources
- [1] ↑네이버 쇼핑앱, 대화하며 쇼핑하는 쇼핑 AI 에이전트 베타 출시 — NAVER Corp.
- [2] ↑네이버 쇼핑앱 AI 에이전트, 사용자에게 먼저 대화 건넨다 — NAVER Corp.
- [3] ↑'검색'에서 '쇼핑'으로…커머스에 AI 무장한 네이버 — 비즈워치
- [4] ↑[네이버] 네이버 쇼핑 상품 DB URL 등록 방법 (샵바이 솔루션) — NHN커머스
- [5] ↑Product Feed Spec — OpenAI
- [6] ↑Shopping in ChatGPT for merchants — OpenAI
- [7] ↑OpenAI pivots ChatGPT shopping strategy after Instant Checkout underwhelms — Retail Gazette
- [8] ↑[GA4] Default channel group — Google
- [9] ↑ChatGPT Business Hub: What its analytics actually measure — Profound
- [10] ↑BOIDA BVI 공식 사이트 — 디자이노블
Related documents
- ACO vs. GEO vs. AEO: What Agentic Commerce Optimization Adds on Top (2026)How ACO (agentic commerce optimization) differs from GEO and AEO in goal, optimization target, measurement endpoint, and technical requirements. Covers the ACP and UCP protocols, Adobe traffic data, a rollout sequence, and a solution comparison table on one page.
- Naver's AI Shopping Agent and Agentic Commerce: AEO Strategy for Korean E-commerce Brands in 2026Naver Plus Store's Shopping AI Agent (beta launched February 2026) and Agent N are pushing Korean e-commerce into agentic commerce, where an AI decides what gets bought. This guide maps the shift and the AEO steps brands need to stay in the game.
- ChatGPT, Perplexity, Gemini & Google AI Mode Shopping Compared: GEO PriorityA side-by-side comparison of the four major AI shopping recommendation platforms: ChatGPT, Perplexity, Gemini, and Google AI Mode, covering data sources, conversion rates, and optimization levers. One page to decide where to put your GEO budget first.
- ChatGPT Shopping Feed Setup: From Merchant Portal to AI Product VisibilityA hands-on guide for sellers who want their products appearing in ChatGPT Shopping results. Covers chatgpt.com/merchants access, required feed fields, common rejection causes, and how to measure AI visibility after approval.
- ACO Measurement: Designing KPIs for AI Shopping Agent Selection Rate and Agent-Assisted ConversionA five-layer KPI model for ACO: visibility, selection, access, purchase path, and conversion. Covers formulas for agent selection rate and agent-assisted conversion rate, the GA4 ai-assistant medium, signed-request verification, and a measurement tool comparison on one page.
- What Is ChatGPT Business Hub? How to Read Brand Mentions, Clicks, and Their LimitsChatGPT Business Hub explained: what it is, how claiming and verification work, how to read mentions, website clicks, and discovery topics, and the structural gaps (no prompts, no competitor data), based on OpenAI primary sources and independent analysis.
- GA4 AI Assistant Channel: Measuring ChatGPT, Gemini, and Claude TrafficGA4's AI Assistant channel: added to Default Channel Groups in May 2026, automatically classifies AI referral traffic, yet a structural blind spot remains: sessions that arrive without a referrer header. Here's what the channel captures, what it misses, and how to cover the gap.
- Best GEO/AEO Companies: Domestic & Global Agencies and Solutions Comparison Guide 2026A comprehensive answer to 'which GEO companies are worth recommending.' Compares domestic Korean (Intermajor, ZESTCOMPANY, Narr/Answer, BizSpring, etc.) and global monitoring tools, diagnosis solutions, and agencies by founding, headquarters, and differentiators.