Salesforce AI Commerce: Visibility Still Wins

Salesforce just made agentic commerce feel operational for enterprise retailers. The bigger lesson: AI shopping now needs two layers, connection and independent visibility.

TL;DR: Salesforce's July 6 Agentforce Commerce release connects owned storefront agents, ChatGPT, Google AI Mode, Gemini, search, order management, B2B buying, and POS into one commerce stack. That helps retailers operate agentic commerce, but it does not answer the market question: when a shopper asks an AI what to buy, are your products surfaced, described correctly, and chosen over competitors?

Salesforce says AI influenced 20% of global online sales during the 2025 holiday season, worth $262 billion, and retailers running their own shopper agents grew sales 59% faster than retailers still on the sidelines, according to its July 6 announcement: https://www.salesforce.com/news/stories/agentforce-commerce-announcement/. That is a signal that agentic commerce has moved from innovation deck to operating plan.

What did Salesforce announce for agentic commerce?

The release matters because it bundles discovery, owned storefront agents, B2B procurement, merchandising, order management, POS, and AI app syndication.

The most important pieces for retail teams are straightforward:

Shopper Agent is generally available for owned storefronts.
Agentic Commerce Search is generally available in July 2026, built from Salesforce's Cimulate acquisition.
ChatGPT integration is generally available in July 2026.
Google Search, including AI Mode, and Gemini app integrations are scheduled for summer 2026.
Storefront Next can launch a production-ready storefront in under 30 minutes and includes more than 40 prebuilt AI-coding integrations.

Those details come from Salesforce's full 12-highlight launch note: https://www.salesforce.com/news/stories/agentforce-commerce-announcement/. Salesforce is not just selling a chatbot. It is trying to make the commerce stack agent-ready, so an AI interaction can check inventory, apply customer context, route an order, and hand the shopper back to a brand-owned relationship.

Many AI shopping conversations focus on feeds, protocols, and whether a brand can be discovered in ChatGPT or Google AI Mode. Salesforce is saying that discovery is only the front door. The back room still has to work.

Why does this change the AI shopping conversation?

The conversation is moving from "can agents find products?" to "can commerce systems support what agents promise?" A recommendation is fragile if pricing, inventory, customer status, fulfillment, and product context do not line up behind it. That is why Salesforce's release is strategically important.

Salesforce's own framing is blunt: "Ensuring that your products show up in new LLM-driven channels is only half the battle." The other half, in its view, is connecting owned properties with external channels so a shopper from ChatGPT or Gemini remains one known customer. Nitin Mangtani, EVP and GM of Agentforce Commerce, put the owned-channel point this way: "The brands that win will have their Shopper Agent live on their own properties for the 2026 shopping season," in Salesforce's launch note: https://www.salesforce.com/ap/news/press-releases/2026/07/06/as-ai-agents-transform-commerce-salesforce-unleashes-its-biggest-agentforce-commerce-release-yet/.

That is the right direction. AI shopping does not end at the answer box. A retailer still needs product data, inventory, order status, customer history, loyalty logic, and service workflows to stay coherent. The mistake is assuming that a connected commerce platform automatically means the market's AI agents prefer your products.

Key stat: Salesforce says AI influenced 20% of global online sales during the 2025 holiday season, worth $262 billion, and that AI-referred traffic converts at eight times the rate of social traffic: https://www.salesforce.com/news/stories/agentforce-commerce-announcement/.

That conversion claim explains the urgency. Retailers need to understand what happens before the visit: which products were shown, which competitors were named, and whether the brand was cited, carded, or merely mentioned.

Where does visibility still break for retailers?

Visibility breaks in the gap between catalog syndication and answer selection. A product can be in the right feed, connected to the right platform, and still lose because the AI lacks the attributes, context, reviews, use-case language, or structured evidence needed to choose it.

This is where query fan-out becomes practical. A shopper rarely asks a clean keyword query anymore. They ask a specific, constraint-heavy question like "best trail shoes for wet city runs under $140 with wide sizes and no break-in period." AI systems break that into sub-questions about material, traction, sizing, reviews, availability, price, and fit.

If your product page only answers three of those sub-questions, the AI may skip you even if your feed is technically connected. The issue is not only feed submission. It is whether your product data answers the hidden questions the agent uses to compare options.

Layer	What it solves	What it does not solve
Commerce platform connection	Makes catalog, customer, order, and fulfillment data available to AI channels	Whether ChatGPT, Gemini, AI Mode, or Perplexity choose your product over a competitor
Product feed readiness	Gives AI systems machine-readable facts, price, availability, and variants	Whether the product has enough context for use-case and comparison queries
Independent AI visibility monitoring	Shows where products appear, how they rank, and who beats them	The commerce operations needed to fulfill the order

Retailers need all three. Salesforce is pushing hard on the first layer, with useful support for the second. The third remains a separate measurement problem. Paz.ai, an agentic commerce optimization platform, focuses on that third layer: monitoring how products appear across ChatGPT, Google AI Mode, and Perplexity, then helping teams identify the catalog gaps behind those rankings.

How should Salesforce Commerce Cloud teams respond?

Salesforce Commerce Cloud teams should treat this release as a readiness forcing function. If external AI channels are becoming part of the Commerce Cloud roadmap, teams need to know which products are agent-ready before those channels become a material traffic source.

Start with four questions:

Which parts of your catalog are ready for AI comparison?
Which products have complete attributes, use-case language, availability, price, image, and variant data?
Which competitors are already being recommended for your highest-value queries?
Which AI surfaces show product cards versus simple brand mentions?

A sunscreen query may turn on ingredients and skin type. A running shoe query may turn on gait, surface, fit, returns, and weather. AI agents compare products through contextual attributes, not just basic SKU fields.

Salesforce's release also raises a governance issue. If Shopper Agent, ChatGPT, Gemini, Google AI Mode, B2B Buyer Agent, and POS all read from overlapping systems, inconsistent data gets expensive quickly. One stale availability field can turn into a bad recommendation. One thin description can turn into a lost product card.

Teams should use this moment to audit the gap between commerce data and AI-readable product context. The fastest starting point is an AI readiness review at product and category level, followed by surface-by-surface monitoring.

What do Square and Visa signal about the same shift?

Square and Visa show that Salesforce is part of a broader infrastructure week, not an isolated platform release. Square brought U.S. food and beverage sellers into ChatGPT and Claude ordering, while Visa moved live agentic payments across European merchants and issuers.

Square says eligible U.S. food and beverage sellers with Square Online Ordering can be discovered in ChatGPT and Claude, with orders routed into existing Square Online Ordering, POS, and Kitchen Display System setups, with no added Square marketplace commission: https://squareup.com/us/en/press/claude-chatgpt-integrations. It also says Square helps more than 4.5 million sellers surface and transact across digital channels.

Visa announced live agentic commerce transactions across Europe with more than 30 issuers and participating merchants including lastminute.com, Frasers, Cleverbridge, and BrickDepot: https://www.visa.co.uk/about-visa/newsroom/press-releases.3457328.html. Visa's Trusted Agent Protocol and Agent Directory are designed to let merchants recognize verified AI agents, while Visa Payment Passkeys authenticate consumer-approved transactions.

The shared theme is simple: infrastructure is getting built faster than catalog teams are adapting. AI referral traffic measurement tells you what arrived. Visibility monitoring tells you what the AI chose before the click.

What to Do This Week

Map your agentic commerce surfaces. List every place your catalog may appear over the next 90 days: ChatGPT, Google AI Mode, Gemini, Perplexity, Salesforce-owned agents, Square-style local commerce surfaces, and retailer-specific agents.
Pick 25 high-intent shopping queries. Use natural language, not keywords. Include constraints a shopper would actually say: budget, use case, ingredients, size, compatibility, delivery, returns, and audience.
Run a product-card gap audit. For each query, record whether your brand appears as a product card, product mention, brand mention, or source citation. Compare that with your top three competitors. Use the same passage-level logic covered in Query Fan-Out for Commerce.
Enrich the weakest product records. Add attributes that answer comparison questions, not just merchandising copy. Prioritize the products with high margin, high search intent, and weak AI visibility.
Separate connection metrics from selection metrics. Track whether feeds are connected, but judge success by whether agents rank and recommend the product.

Frequently Asked Questions

Is Salesforce now solving agentic commerce for retailers?

Salesforce is solving a major operating layer for Commerce Cloud customers: owned agents, external AI channel integration, search, order management, B2B buying, and storefront tooling. It does not automatically prove that external AI systems will rank a brand's products ahead of competitors across every shopper query.

Does a ChatGPT or Gemini integration guarantee visibility?

No. Integration can make a catalog eligible, but AI visibility depends on product data quality, structured attributes, source authority, use-case coverage, availability, reviews, and competitive context. Eligibility is the starting line. Recommendation quality is the race.

What is the difference between AI traffic and AI visibility?

AI traffic measures visitors who arrive after an AI-assisted interaction. AI visibility measures what happened before the visit: whether your products appeared, where they ranked, how they were described, and whether a competitor was chosen instead. Both matter, but they answer different questions.

Why does query fan-out matter for commerce teams?

Query fan-out means an AI breaks one shopper question into multiple sub-questions before answering. For commerce, that means product pages must answer use-case, attribute, persona, price, availability, and comparison questions. Thin product data loses because it cannot satisfy enough of the hidden sub-queries.

Salesforce just raised the floor for agentic commerce operations. Retailers still have to win the shelf that AI agents create. The brands that treat connection as the beginning, then measure selection and fix product data gaps, will be the ones AI shopping systems can confidently recommend.