Agentic Commerce in Q4 2026: How brands can get recommended by AI
Last Updated:
Key takeaways
AI shopping agents come in three kinds: answer engines (ChatGPT, Gemini), retailer shopping assistants (Alexa for Shopping, Walmart Sparky, TikTok) and personal AI agents (Meta Muse, Instinct, Dots).
Brand content is the one input a brand controls in all three: the open web for answer engines, the listing for retailer shopping assistants, and feeds and live pages for personal AI agents.
Paid is arriving in two of the three, through ChatGPT ads (now also bought via Amazon DSP) and Branded Conversations on Alexa for Shopping from January 2027. In every format, the brand pays for the placement and the agent still builds the answer from content.
ChatGPT visibility is a PR and comms job as much as an ecommerce one, because much of what it cites is earned media.
Meta Muse cites nothing and produces no answer to track, so scoring the pages it reads is the measurement available today.
Start Q4 by scoring priority pages against the 5Cs and recording a share of voice baseline before any ad spend goes live.
This holiday season, more of the shopping will be done by AI agents. Google’s VP of Retail, Courtney Rose, says 70% of shopping queries on Google Search focus on discovery, exploration or product research, and cites BCG research that shopping is the third most popular reason people use AI. On Amazon, more than 350 million shoppers used Alexa for Shopping over the past 12 months, CEO Andy Jassy said on the Q2 earnings call.
Our view is that the foundation for Q4 is the content these agents read. Paid options are arriving quickly, but in every format the brand pays for the placement and the agent still builds the answer from pages, listings and feeds a brand controls. Brands that lay that foundation now will find it far easier to adjust as autonomous buying grows.
Azoma is the Agentic Commerce Optimisation (ACO) platform that helps brands drive revenue through AI shopping agents like Alexa for Shopping, Walmart Sparky, ChatGPT and Gemini. We talk with enterprise CPG and retail brands regularly about their agentic strategy, and this piece reflects what they are planning next.
The three kinds of agents
Answer engines like ChatGPT and Gemini sit in the upper funnel and answer from the open web. Retailer shopping assistants like Alexa for Shopping and Walmart’s Sparky sit inside a retailer, read that retailer’s catalog and reviews, and close the sale there. Personal AI agents are the newest: they take a request, build a shortlist and can buy on the shopper’s behalf.
Meta launched Muse on September 8, and it passed 5 million US downloads in 22 days, according to Sensor Tower. Muse can open a browser, fill in forms and complete a purchase, and it asks for approval before it buys.
Two more personal AI agents followed within weeks. Instinct, an agent people reach by text, raised $1 billion at a $10 billion valuation on September 28 after launching invite-only in August. A day later, OpenAI launched Dots at DevDay, always-on agents that run on their own cloud computers with their own browsers. OpenAI hasn’t said how Dots handles shopping yet. If it does shop, OpenAI would have ads in ChatGPT and a personal AI agent of its own.
Retailers are choosing sides. Amazon began blocking Muse from browsing and buying on Amazon.com on September 20, because it wants agentic shopping on its site to run through Alexa for Shopping. For now, that keeps the largest retailer shopping assistant separate from personal AI agents.
We expect more retailers and commerce platforms to open up to personal AI agents over time, and that is already starting. On October 6, Sierra introduced the Personal Agent Protocol, an open standard that lets a company decide how a personal AI agent reaches it, whether through its website, APIs or its own agent. Meta, Instinct, Shopify, Stripe and Walmart are among its supporters.
The first version of the protocol is due later this month, so little changes for Q4. It does mean a brand’s content will likely reach agents by more routes than the page, which makes getting it right at the source more important.
The retailer group is getting wider, too. On October 5, TikTok announced a Shopping Assistant: a conversational AI agent, built into several parts of the app, that covers product details, shipping, sizing, availability and purchasing in one conversation. It came with Buy Direct, an in-app checkout that lets a shopper buy from a brand without leaving TikTok.
TikTok hasn’t said when either launches or what the assistant reads, but a social platform with its own shop now sits in the same group as Alexa for Shopping and Sparky. TikTok announced the assistant next to automated search ad campaigns and an update to GMV Max, its ad product for TikTok Shop sellers. If it follows Amazon’s path, the questions shoppers ask there could become an ad input as well.
Traffic is growing, too. Cisco’s latest survey data has AI network traffic on pace to double every six months, and its testing found agent-run tasks can generate up to 450% more total network traffic. More of that traffic is an agent reading a feed or a page on someone’s behalf. If agent traffic keeps taking share from human visits on the open web, the emphasis shifts to getting the content right on product feeds, retailer PDPs and brand-owned sites.
What every AI shopping agent reads
Each agent builds its recommendation from different material, and a brand controls a good part of it in every case. A single visibility score cannot carry across them. Visibility in one engine says little about the next, and a personal AI agent produces no answer to track at all, so what carries across all three is the content they read.
Agent | What it reads | What a brand can change |
|---|---|---|
Answer engines (ChatGPT, Gemini) | The open web, with sources named in answers | What the brand publishes, and who else writes about the product |
Retailer shopping assistants (Alexa for Shopping, Sparky, TikTok) | Retailer listings, reviews and outside sources | Attributes, answers to shopper questions, gaps against competitors |
Personal AI agents (Meta Muse, Instinct, Dots) | Product feeds and live merchant pages | brand.com, retailer pages and the feeds behind them |
Table 1. What each kind of AI shopping agent reads, October 2026. The personal AI agents row reflects how Muse works today.
Brand content is the one input a brand controls in all three. Agentic Commerce Optimisation (ACO) is the practice of making a brand’s products discoverable, understandable and recommendable by AI shopping agents across ChatGPT, Gemini, Amazon Rufus, Alexa for Shopping, Walmart Sparky and Perplexity.
We score that content with the 5Cs of Agentic Commerce: Completeness, Context, Citations, Correctness and Customer Acquisition. The framework applies to a brand site, a retailer listing or a feed, and each agent leans on different Cs.
How to get recommended by Alexa for Shopping
Alexa for Shopping, the assistant Amazon previously branded Rufus, recommends from the retailer listing first, so the listing does most of the work. A listing can be complete for a shopper who already wants the brand and still miss the plain category question, such as which product is best for a given use. Completeness and Context are where most listings have room, and share of voice by topic and product rank show which listings to start with.
What shoppers ask Alexa for Shopping
Sponsored prompt reporting is the best clue available to what shoppers in a category are asking AI. Brands running Sponsored Brands prompts now get a range of ad metrics on those prompts. Read them as a starting list of shopper questions, check each against the listing, and add the answers that are missing.
Branded Conversations and Conversational Briefs
Amazon announced Branded Conversations at unBoxed on September 29. When a shopper taps a Sponsored Brands prompt, the conversation with Alexa for Shopping will draw on brand-provided expertise as well as Amazon’s own signals. It launches in the US in January 2027, and Amazon says nearly 20% of shoppers who interact with a Sponsored Brands prompt continue the conversation about that brand.
Brands supply those inputs through Conversational Briefs, such as which products are bought together and free-text descriptions of features and use cases. They cannot pick which reviews the AI surfaces. For brands, this is the first direct input into the questions and answers they have spent most of 2026 trying to influence organically, and the Context work on a listing today is what goes into the brief in January.
The role of creators
Amazon is building more creator content into Alexa for Shopping, and we are starting to see it tied to media in page-one results. On August 27, YouTube added Amazon to its Shopping affiliate program, so eligible US creators can tag Amazon products in Shorts, long-form videos and livestreams.
How to get recommended by ChatGPT
ChatGPT and other answer engines answer from the open web with named sources, so Citations and Correctness carry the most weight. Share of voice by topic shows which questions a brand is missing from, and the citation list points to the outside sources that need attention. Much of that is earned media, which makes ChatGPT visibility a PR and comms job as much as an ecommerce one.
Paid is arriving fast. OpenAI says its ad business crossed a $1 billion annualized run rate after starting ad tests in February, expanded ChatGPT ads to more than 40 new countries, and said new ad formats, buying options and measurement capabilities are next. Since September 10, select US advertisers can also buy ChatGPT ads through Amazon DSP, so a brand can now run paid in both an answer engine and a retailer shopping assistant from one buying platform.
How to get recommended by Meta Muse
Meta Muse chooses and buys from what it reads on product feeds and live merchant pages, so Completeness and Correctness on brand.com and retailer pages matter most. Muse produces no answer to track and cites nothing, so the measurement available today is a score of the pages it reads. The work a brand does for retailer shopping assistants should carry over to Muse, Instinct and Dots.
Muse carries no ads, and Meta says conversations aren’t shared with its ad systems. Social strategy could become an indirect way in if Muse personalizes on what it knows about a shopper. When a request isn’t personalized, the brand likely has to earn the recommendation through the content, sales and sentiment on the pages the agent reads.
How to measure visibility and agentic growth across AI shopping agents
Tracking starts with a baseline for each kind of agent: share of voice by topic and citations in answer engines, share of voice, product rank and content gaps against competitors in retailer shopping assistants, and page and feed readiness for personal AI agents.
Nobody can fully measure agentic growth yet. The agentic commerce RFPs we have seen from top-50 CPGs ahead of Q4 are centered on revenue attribution: visibility tracking gave brands their first look at how they show up in LLMs, and ahead of 2027 they want growth outcomes they can point to. The questions brands should be asking are whether ads lift organic inside each agent, and whether a shopper carries over from a ChatGPT prompt to a lower-funnel moment on Amazon.
Customer Acquisition, the fifth C, is how we work on those questions. Every content action in Azoma’s Actions Bank is logged with a date and read against share of voice, citations, rank and revenue, with connectors to retailers and the brand’s product source of truth bringing live content and sales into one place. As ads take up more of these agents, paid data needs to sit next to that record, so a brand can see what it changed, what it spent and what moved.
Four practical steps to take before the holiday season
Score the pages: Run priority products through the 5Cs on brand.com and the top retailers. The gaps become the content plan, and the same fixes go out to every feed.
Baseline before the spend: Record share of voice by topic and citations in ChatGPT and Alexa for Shopping before any ChatGPT ad spend goes live, direct or through Amazon DSP. Without that baseline, there’s no way to tell if paid lifts organic.
Use the sponsored data: Read sponsored prompt reports as a list of what shoppers ask Alexa for Shopping, add the missing answers to the listing, and carry them into the Conversational Brief for January.
Change one thing at a time: Put content changes, share of voice, paid data and sales by category side by side each week, with a date on every change.
Across every agent, the brand pays for the placement and the agent still builds the answer from content.
Azoma tracks your visibility across AI shopping agents, scores your content against the 5Cs, and ties every change to revenue. Talk to our team to get ready before the holiday peak.
Article Author: Max Sinclair