Agentic Commerce Optimisation: A Guide for Marketplace Sellers
AI assistants now find, compare and sometimes buy products for shoppers. For marketplace sellers, preparing for them is mostly a data job: identifiers, attributes, and prices and stock that match reality.
Agentic commerce means AI assistants such as ChatGPT or Google's AI Mode find, compare and sometimes buy products for a shopper. For marketplace sellers, optimising for it is mostly a data job: correct identifiers, complete attributes, and prices and stock that match reality at every moment. GAPLI, a dropshipping platform, keeps catalogue prices and stock in sync for its sellers. Figures are as of 8 Oct 2026.
Key takeaways
- Three open protocols shape agentic shopping: OpenAI and Stripe's Agentic Commerce Protocol (Sep 2025), Google's Agent Payments Protocol (Sep 2025) and Google's Universal Commerce Protocol (Jan 2026).
- The first agent checkout features launched for US shoppers and merchants; in March 2026 OpenAI moved its focus from in-chat checkout to product discovery.
- Adobe measured a 693.4% rise in traffic from generative AI tools to US retail sites in the 2025 holiday season, while noting the user base is still modest.
- Agents and shopping feeds reject data that does not match the landing page: Google disapproves products whose price or availability differ from the site.
- What sellers control is product data quality: EAN, structured titles, complete attributes, delivery time, return terms and GPSR safety data.
What is agentic commerce and how is it different from marketplace SEO?
Agentic commerce is shopping in which an AI agent does part of the work a person used to do: searching, comparing offers and, where supported, completing the purchase. Marketplace SEO optimises a listing for a human scanning search results; optimising for agents means making the listing's facts machine-readable, consistent and current, because the agent decides what the shopper sees.
| Dimension | Marketplace SEO | Optimising for AI agents |
|---|---|---|
| First reader | A person and the marketplace search engine | A language model or shopping agent |
| What it uses | Title keywords, photos, ratings, sales history | Identifiers, attributes, price, stock, delivery and return data |
| Typical failure | Lower ranking | Offer not considered or rejected for inconsistent data |
| Where it happens | Inside one marketplace | Across marketplaces, stores and search surfaces |
The two overlap. A clear title with brand, model and key specifications helps a human and an agent. Persuasive adjectives help neither.
Which agentic commerce protocols exist in 2026 and who is behind them?
Three open standards matter for sellers: the Agentic Commerce Protocol (ACP) from OpenAI and Stripe, the Agent Payments Protocol (AP2) from Google, and the Universal Commerce Protocol (UCP), which Google co-developed with retailers and platforms. They define how agents read product data, build a cart and pay without exposing card details.
| Protocol | Announced | Backed by | What it does |
|---|---|---|---|
| Agentic Commerce Protocol (ACP) | 29 Sep 2025 | OpenAI and Stripe (open standard) | Connects merchants to AI agents; launched with Instant Checkout in ChatGPT for US users buying from Etsy sellers |
| Agent Payments Protocol (AP2) | 16 Sep 2025 | Google with more than 60 payments and technology organisations | A payment-agnostic framework for agent-initiated payments |
| Universal Commerce Protocol (UCP) | 11 Jan 2026 | Google, co-developed with Shopify, Etsy, Wayfair, Target and Walmart | Covers discovery, purchase and post-purchase; powers checkout on eligible listings in AI Mode and the Gemini app for US retailers |
The field is moving fast. On 24 March 2026 OpenAI said the first version of Instant Checkout did not give the flexibility it wanted, let merchants use their own checkout, and refocused ACP on product discovery in ChatGPT. For sellers the lesson is that discovery data matters before any checkout button does.
How much shopping traffic do AI assistants send today?
It is growing fast from a small base. In the 2025 holiday season (1 Nov–31 Dec), Adobe Analytics recorded a 693.4% year-on-year rise in traffic to US retail sites from generative AI tools, and Adobe itself noted that the base of users remains modest. There is no comparable public figure for Polish marketplaces.
Two practical conclusions follow. First, AI referrals are worth tracking now, because growth from a small base becomes visible quickly. Second, there is no reason to rebuild a catalogue only for agents: the data that agents need is the same data that marketplace search, comparison sites and buyers already reward.
Do AI agents shop on Allegro or Erli?
As of 8 Oct 2026, the agent checkout features announced by OpenAI and Google launched first for US shoppers and US merchants. Polish marketplaces are reached by AI assistants mainly the way search engines reach them — through public listing pages and the data on them. What you control is the quality of that data on every channel.
This matters most for sellers who list the same product on several channels. An assistant that sees the same EAN on Allegro, Erli and your own store with three different specifications, or with stock on one and none on another, has a reason to trust none of them. Consistency across channels is the cheapest improvement a multichannel seller can make. With GAPLI, every channel lists from the same catalogue, so specifications, price logic and stock come from one source — see how GAPLI works.
What product data do AI shopping agents need?
Agents need facts they can compare without guessing: a global identifier (EAN/GTIN), a structured title, complete category attributes, price including delivery, current availability, delivery time, return terms and safety information. Google's Merchant Center rules show the standard: price and availability in your data must match the landing page and checkout, or the product is disapproved.
| Data | Why an agent needs it | Good practice |
|---|---|---|
| EAN (GTIN) | Matches the same product across sellers and marketplaces | One valid EAN per variant; see the EAN glossary entry |
| Title | Quick identification of brand, model, variant | Brand + model + 2–3 key specs + colour or size; no promotional words |
| Category attributes | Filters such as size, power, compatibility | Fill every required and relevant optional field |
| Price and delivery cost | Total-cost comparison | Same logic on every channel; differences explained by delivery or service |
| Availability and handling time | Whether the offer can actually be bought | Updated automatically from supplier stock |
| Return and warranty terms | Risk comparison for the buyer | Stated in the offer, not only in a PDF |
| GPSR data | Required in the EU since 13 Dec 2024 | Manufacturer, EU responsible person, warnings |
| Specs in text, not images | Text is parsed reliably | Repeat key specs from images in the description |
Google is also adding new Merchant Center attributes for conversational discovery, such as answers to common product questions and compatible accessories or substitutes, according to its January 2026 announcement. If you run your own store, structured product data in a product feed is the way to supply them.
Why do stock and price accuracy matter more when an agent shops?
An agent compares many offers in seconds and favours the ones it can rely on. A listing that shows stock the supplier no longer has leads to a cancelled order, and a price that differs between feed and page gets the product disapproved. In dropshipping, supplier update speed decides how accurate your listings are.
GAPLI pulls prices and stock from wholesalers automatically and applies the changes to published listings. In the 24 hours before 8 Oct 2026 it updated 1,250,796 catalogue items, out of 980,891 products available for sale from 161 active wholesalers (GAPLI database).
| Channel | How GAPLI keeps the offer accurate |
|---|---|
| Allegro | Offer ended when stock runs out and reactivated when it returns; handling time synchronised; large wholesale price drops held for review before they reach Allegro |
| Erli | Stock and prices synchronised on a schedule; offers switched off when the product leaves the catalogue |
| WooCommerce | Stock and prices every few minutes |
| Shopify | Stock and prices every 10 minutes |
Holding suspicious price drops for review matters here too: an agent hunting for the lowest price will find a price error before any human does.
How should a dropshipping seller prepare a catalogue for AI agents?
Start with the data that every channel already uses, then add what agents specifically reward. In practice that is eight checks, most of which a platform can automate and a few of which remain your decision — for example which products to list and how to price them.
| Step | What to do | In GAPLI |
|---|---|---|
| 1. Identifiers | Keep only products with a valid EAN | Allegro offers are matched to product cards by EAN; products without one are not listed; Empik matches by EAN |
| 2. Categories | Map every product to the right marketplace category | Categories are mapped automatically on Erli and eMAG |
| 3. Attributes | Fill required and relevant optional fields | On TikTok Shop (beta), missing attributes are filled by AI with your own OpenRouter key |
| 4. Safety data | Add GPSR manufacturer data and warnings | Attached to Allegro offers automatically |
| 5. Stock and price | Keep them current on every channel | Automatic sync from wholesaler data |
| 6. Terms | Make delivery, returns and warranty explicit | Return, complaint and warranty terms are created on the Allegro account from global templates |
| 7. Consistency | Same specs and price logic everywhere | All channels list from one catalogue |
| 8. Own-store markup | Add Product structured data to product pages | Your store theme or plugin |
The full channel-by-channel scope is listed on GAPLI integrations. For a deeper look at product data work, see our guide to AI product data automation.
How do you measure whether AI assistants send you buyers?
Track referral traffic and orders from AI assistants in your own store analytics, and compare them month by month. Marketplaces do not usually show where a visitor came from before reaching your offer, so your own store is where agent traffic is easiest to see.
- Create a channel group for AI referrers, such as chatgpt.com, gemini.google.com, copilot.microsoft.com and perplexity.ai.
- Track conversions from that group separately, not only sessions.
- Test your products in assistants: ask for your category with your key specs and check whether your offer, price and stock appear correctly.
- Fix what they get wrong: a wrong price or specification usually points back to inconsistent data on one of your channels.
- Review quarterly, because protocols, eligible countries and assistant features change quickly.
If you want to automate reports or alerts around your catalogue, every GAPLI plan includes API access, and there is an n8n node, n8n-nodes-gapli. Opening an account and the Start plan are free — create an account.
Frequently asked questions
Do I need to integrate with ACP or UCP to be found by AI assistants?
Not as a marketplace seller. These protocols connect merchants' own systems to agents and, at launch, served US merchants. On marketplaces, assistants read the listing data you publish, so complete and consistent offers are what you can act on today.
Will AI agents replace marketplace search?
Nobody can say yet, and current numbers do not support a forecast. Adobe's holiday data shows fast growth of AI referrals from a modest base. A seller who keeps data clean benefits in both marketplace search and AI discovery, so the work is not wasted either way.
Should I rewrite my product descriptions with AI?
Only if a person checks the result. AI can structure supplier text and fill gaps, but an invented specification is worse than a missing one, because agents and buyers rely on it. Keep the supplier's original data as the source and use AI to format it, not to add facts.
Does GAPLI list products on ChatGPT or Google directly?
No. GAPLI lists products on its sales channels — Allegro, Erli, Aderlo, WooCommerce and Shopify live, with eMAG, Empik, TikTok Shop and eBay in beta. AI assistants can then find those listings like any other public page.
Sources
- GAPLI data (catalogue size, active wholesalers, items updated in 24 hours, channel status and sync behaviour): GAPLI database and GAPLI integration pages, as of 8 Oct 2026
- Stripe — Stripe powers Instant Checkout in ChatGPT and releases the Agentic Commerce Protocol, 29 Sep 2025
- Google Cloud — Announcing Agent Payments Protocol (AP2), 16 Sep 2025
- Google — New tech and tools for retailers in an agentic shopping era (Universal Commerce Protocol), 11 Jan 2026
- OpenAI — Powering product discovery in ChatGPT, 24 Mar 2026
- Adobe — 2025 holiday shopping season results
- Google Merchant Center Help — price attribute
- Google Merchant Center Help — availability attribute
- European Commission, Access2Markets — EU General Product Safety Regulation
Related Content
GAPLI Team
E-commerce automation & dropshipping insights.