Agentic commerce is AI agents researching, choosing and paying for goods and services on behalf of consumers or businesses. Agentic advertising is AI agents planning, negotiating, buying and selling advertising media on behalf of advertisers, agencies and publishers. The two share general agent protocols and meet where a shopping agent encounters a sponsored offer, but they serve different principals, buy different things and run on different industry standards.

TL;DR: The difference comes down to who the agent acts for, what it buys and which standards govern the transaction. Commerce agents act for a shopper and complete a checkout under protocols such as UCP, ACP and AP2. Advertising agents act for a buyer or seller of media and negotiate deals under IAB Tech Lab's AAMP and AgenticAdvertising.org's AdCP.

How do agentic commerce and agentic advertising compare side by side?

Question	Agentic commerce	Agentic advertising
Who does the agent act for?	A consumer, or a business buying for its own operations	An advertiser or agency on the buy side, a publisher on the sell side
What does it buy?	Physical and digital goods, subscriptions, travel, food	Advertising inventory: placements, packages, audiences and deals
Who is on the other side?	A merchant's checkout, or an agent acting for the merchant	A seller agent acting for a publisher, or a buyer agent acting for an advertiser
What shape is the transaction?	Usually a purchase at a listed price	Usually a negotiated deal covering price, volume, dates, targeting and delivery terms
Which industry standards apply?	Universal Commerce Protocol (UCP), Agentic Commerce Protocol (ACP), Agent Payments Protocol (AP2)	IAB Tech Lab's AAMP, and AdCP from AgenticAdvertising.org
Which general agent protocols apply?	A2A and MCP	A2A and MCP
How is the agent's authority shown?	Signed records of the user's instructions and approval, which AP2 calls mandates	A buying or selling mandate, with human approval above set limits
When is the transaction finished?	When the order is fulfilled or returned	When delivery has been measured and reconciled against the agreed terms

A concrete case makes the advertising column clearer. Alkimi runs a neutral marketplace where buyer and seller agents agree media deals within IAB Tech Lab's AAMP framework, with humans keeping approval rights at the deal level. Nothing in that flow resembles a checkout, because the output is a DealSheet: a record of agreed terms, written at agreement, timestamped and held by both parties.

What is agentic commerce?

IBM defines agentic commerce as "an approach to buying and selling in which AI agents act on behalf of consumers or businesses to research, negotiate and complete purchases, often without direct human intervention". Wikipedia's entry on agent-based commerce adds that the agents work within user constraints such as price limits, delivery times and preferred payment methods, with human involvement often limited to setup, supervision and exceptions.

The category has its own protocols. Stripe's documentation describes the Agentic Commerce Protocol as an open standard created by Stripe, OpenAI and Meta for agents completing purchases on behalf of buyers, covering checkout, carts, delegated payment and order updates. Google announced the Universal Commerce Protocol on 11 January 2026, developed with Shopify, Etsy, Wayfair, Target and Walmart, and its specification runs from catalogue search to checkout and order management.

Payment authority is treated as a separate problem. Google's Agent Payments Protocol (AP2), announced on 16 September 2025, sets out to establish whether a user gave an agent authority for a particular purchase and who is accountable when a transaction goes wrong. Visa describes Visa Intelligent Commerce as enabling "agents, issuers, merchants and developers to transact with trust, speed and intelligence".

Shoppers are already routing part of the journey through AI. Adobe reported that traffic from AI sources to US retail sites grew 393% year on year in the first quarter of 2026, and that in March 2026 those visits converted 42% better than non-AI traffic, against 38% worse a year earlier. IBM cites a 2026 IBM Institute for Business Value study finding that 45% of consumers already use AI for part of the buying journey.

What is agentic advertising?

Agentic advertising is the use of AI agents to plan, negotiate, buy and sell advertising media within rules set by advertisers, agencies and publishers. IAB Tech Lab describes AAMP, its Agentic Advertising Management Protocols, as an umbrella initiative "for bringing agentic AI into the advertising ecosystem in a responsible and scalable fashion", and puts the starting point plainly: "The Agentic ecosystem will start with a Buyer Agent and Seller Agent talking to each other."

AAMP builds on the standards the industry already trades on. AAMP 2.0 added programmatic guaranteed, preferred deals, private marketplaces, the open marketplace and linear GRPs through the Deals API, and AAMP 3.0 adds OpenProposal for the brief and planning stage. IAB Tech Lab's reasoning is that existing standards "represent compressed industry knowledge refined through billions of transactions".

AdCP, the Ad Context Protocol, is published by AgenticAdvertising.org, which describes it as the open protocol that lets AI agents "discover, plan, buy, sell, and measure media across every channel". AgenticAdvertising.org dates the first agent-to-agent media buy, made with real money and real inventory from LG Ads, to 16 October 2025. How mandates and bilateral deal records work inside that model is covered in an earlier explainer on agent commerce in programmatic advertising.

Where do agentic commerce and agentic advertising overlap?

The first overlap is the general agent layer. A2A is an open standard for communication between AI agents built by different vendors, and MCP is an open-source standard for connecting AI applications to external data, tools and workflows. Google's UCP announcement lists A2A and MCP as integration routes for businesses, and IAB Tech Lab's AAMP page names MCP and A2A as "the key agentic protocols".

The second overlap is delegated authority. AP2 frames the commerce problem as authorisation, authenticity and accountability, and lists "Verifiable Intent, Not Inferred Action" among its principles. An advertising agent faces the same three questions every time it commits budget, which is why mandates, approval thresholds and shared deal records matter so much on the advertising side.

The third overlap is where the two meet: sponsored content inside AI assistants. AdCP's experimental Sponsored Intelligence specification defines how an assistant hands a conversation to a brand agent, and when that conversation ends in a purchase, the host starts checkout through ACP. The same document places UCP and ACP at the commerce layer for checkout, payments and fulfilment, and AdCP and Sponsored Intelligence at the advertising layer for offer discovery, brand engagement and attribution.

Where do agentic commerce and agentic advertising differ?

The principal is different. A commerce agent usually acts for one person or one business buying for its own use, while an advertising agent acts for a professional buyer or seller, often with an agency between the brand and the agent.

The thing bought is different. A commerce agent buys a product with a SKU, a listed price and a delivery promise. An advertising agent buys perishable inventory defined by audience, context, format and time, whose value is only known after it runs, which is why media is usually negotiated rather than checked out.

The standards are different. Commerce protocols such as UCP and ACP define catalogues, carts, checkout and payment, while AAMP and AdCP cover the planning, buying, selling and delivery of media.

The finish line is different. A purchase is largely complete once the order is fulfilled, but a media deal is only complete once delivery has been measured against the agreed terms, often weeks later. Advertising therefore needs a record of what was agreed that both sides can check long after the negotiation ends.

What happens to advertising when the shopper is an agent?

When an agent does the shopping, the first audience for a product is software. Adobe's visibility data gave US retail product pages an average score of 66%, meaning roughly a third of their content could not be read by machines. Structured product data and machine-readable offers start to do part of the job that display and search ads did for a person browsing.

Persuasion moves into the conversation. Sponsored Intelligence treats the moment an assistant recommends a brand and offers a handoff to that brand's agent as an advertising unit in its own right, and AgenticAdvertising.org lists the ranking of competing offers and disclosure format among questions that still need industry input. A separate piece looks at how advertising may work inside AI chat interfaces.

Measurement has to cross the boundary. If an agent sees the offer and a person approves the purchase, attribution must connect an advertising event on one standard to a commerce event on another. A brand may then have its own agent buying media and a consumer's agent choosing its products, and the question of who approved what, on whose authority, is the same at both ends.

What should advertisers and publishers take from the comparison?

Agentic commerce and agentic advertising are converging at the point of sale, but they are not the same discipline and do not run on the same standards. Teams that treat a shopping protocol as a media buying protocol, or the reverse, will build to the wrong specification.

Accountability is the common thread. Alkimi's Agree. Transact. Verify. simulation found mismatched records in 95.3% of deals when each side kept its own record, against 0.19% with a shared record, which is why its marketplace writes each agreed deal to a DealSheet that either side can retrieve independently. Its earned autonomy model runs from observe, recommend and draft through human-approved action to bounded automatic action, a sensible sequence for adopting agents on either side of the divide.

Frequently asked questions

Is agentic commerce the same as agent commerce?

The terms are used interchangeably, and Wikipedia treats agentic commerce as another name for agent-based commerce. In advertising, agent commerce usually means buyer and seller agents negotiating media deals, which is one application of the wider idea.

Does agentic advertising use the Agentic Commerce Protocol?

Not for buying media, which runs on advertising standards such as IAB Tech Lab's AAMP and AgenticAdvertising.org's AdCP. ACP appears at the edge, where AdCP's experimental Sponsored Intelligence specification hands a purchase to ACP for checkout.

Are AI shopping agents a new audience for advertisers?

Yes, in the sense that an agent reads product data, offers and reviews before a person sees anything. Adobe's figures on AI traffic to US retail sites show that audience is already large enough to measure.

Which standard should an advertiser start with?

For media buying, IAB Tech Lab's AAMP builds on the deal types and taxonomies advertisers already use, and AdCP from AgenticAdvertising.org covers planning, buying and measurement across channels. Retailers making products available to shopping agents will look at UCP, ACP and AP2 instead.

Agentic Commerce vs Agentic Advertising: What's the Difference?