# AI agents now shop your stores before your customers do.

> Agentic commerce is shopping in which an AI agent searches, compares and sometimes buys for a person. In 2026 most of it is discovery: ChatGPT, Google's AI Mode and Gemini, Copilot and Perplexity build a shortlist and send the shopper to a site or a store. For a retail chain, readiness means every store's products, prices, stock, hours and policies are published in a form agents can read and trust.

*Markdown version of [https://ocally.co/use-cases/agentic-commerce](https://ocally.co/use-cases/agentic-commerce), written for AI agents. All docs: [llms.txt](https://ocally.co/llms.txt). Deciding whether Ocally fits a business? Read [Is Ocally right for this business?](https://ocally.co/evaluate.md).*

Agentic commerce is moving the first visit from your aisle to an AI agent's shortlist. For a retail chain, the stores that win are the ones whose price, stock, hours and policies an agent can read and trust.

Ocally research (agentic commerce readiness), October 2026. 49 sources, cited inline as [n] and listed at the end.

## Key takeaways

- OpenAI is moving away from standalone Instant Checkout [1]. For now, agents recommend and people buy on your site or in your store.
- In SOCi's study of 350,000+ chain locations, ChatGPT recommended 1.2%. Google's local 3-pack showed 35.9% [2].
- The unit of readiness is the store-item pair: is it here, what does it cost here, can I have it today.
- In a landmark study of 370,000 store inventory records, 65% were wrong [3]. Agents turn that error into a public promise.
- Google's AI already phones nearby stores to confirm price and availability for shoppers [4]. Published stock answers that call before it rings.
- Skip custom agent checkout this year. Spend the money on store data you can stand behind, then measure what AI says.

## In this piece

1. Agentic commerce means software does the shopping legwork
2. AI companies, card networks and platforms are laying the rails
3. So far, agents recommend and people still buy
4. Agents pick the stores whose facts they can check
5. For a chain, every question ends in where and today
6. What changes for discovery, price, loyalty and the visit
7. Most chains are preparing for the wrong step
8. The risks are real, and most start as bad data
9. Where this goes in the next five years
10. The agentic commerce readiness checklist for this year
11. How Ocally helps

## 1. Agentic commerce means software does the shopping legwork

Agentic commerce is shopping in which an AI agent does part of the work a shopper used to do: searching, comparing, checking price and stock, and sometimes paying. The shopper states a need in plain words, like a cordless drill under $150 that a store nearby has today. The agent turns that into dozens of lookups, reads what comes back, drops what doesn't fit and returns a short answer. The shopper sees three or four options. Everything else might as well not exist.

> The agent finds, decides and acts. Chains win or lose on the first two.

The assistants doing this are already large. OpenAI says ChatGPT reaches a collective 1.2 billion weekly users [5]. Google's Shopping Graph holds more than 50 billion product listings, 2 billion of them refreshed every hour, and its AI can phone a store to ask what is on the shelf [6]. Microsoft's Copilot now has its own checkout [7]. What changed in the past year is that these assistants gained standard ways to read catalogs, prove who they are and pay.

It helps to split the agent's work into three jobs. It finds: it searches the web, product feeds and maps. It decides: it weighs price, ratings, distance and delivery against what the shopper asked for. It acts: it fills a cart, pays or books. Most of the news is about the third job, because that is where payment networks and AI companies are competing to sit in the middle of the transaction. A retail chain wins or loses on the first two, because that is where a store gets onto the list or never appears.

## 2. AI companies, card networks and platforms are laying the rails

The rails went down fast. In September 2025 OpenAI and Stripe launched Instant Checkout in ChatGPT on a new open standard, the Agentic Commerce Protocol [8]. The same month Google published the Agent Payments Protocol with more than 60 payments and technology partners [9]. In January 2026 Google announced the Universal Commerce Protocol, co-developed with Shopify, Etsy, Wayfair, Target and Walmart, to power checkout in AI Mode and the Gemini app [10]. Microsoft had launched Copilot Checkout three days earlier [7].

The card networks are building the trust layer. Visa's Trusted Agent Protocol, built with Cloudflare, helps a merchant tell a legitimate shopping agent from a malicious bot [11]. Visa and Mastercard both build on Web Bot Auth, which lets a site check an agent's signature and see whether it is browsing or paying [12]. The platforms are wiring their merchants in. Shopify products are discoverable in ChatGPT by default, and its Agentic Storefronts reach Copilot, AI Mode and Gemini [13]. By September 2026 Google said UCP enabled direct checkout for hundreds of thousands of brands and retailers [14].

Notice what none of this asks of a 40-store chain. Nobody needs to write a protocol. Every one of these rails asks the merchant for the same thing: product data that is complete, current and consistent with what the store actually has. Standards bodies are working the same problem from the other side. W3C and GS1, the standards body behind the barcode, held a workshop in September 2026 on e-commerce for humans and AI agents [15].

| When | What |
|---|---|
| Sep 2025 | OpenAI and Stripe launch Instant Checkout in ChatGPT and the Agentic Commerce Protocol. |
| Nov 2025 | Google's AI starts phoning nearby stores to check stock and price for shoppers. |
| Jan 2026 | Google announces the Universal Commerce Protocol with Shopify, Etsy, Wayfair, Target and Walmart. |
| Mar 2026 | OpenAI steps back from Instant Checkout. Purchases move to merchants' own sites and apps. |
| Mar 2026 | UCP adds loyalty linking, so member prices can follow shoppers into AI agents. |
| Sep 2026 | Google says UCP enables direct checkout for hundreds of thousands of brands and retailers. |

## 3. So far, agents recommend and people still buy

The first big test of checkout inside a chatbot went badly. In March 2026 OpenAI said the first version of Instant Checkout did not offer the flexibility it wanted, and moved its effort to product discovery [16]. Walmart said items sold inside the chatbot converted at a third of the rate of items that sent shoppers to its site, and Etsy said it saw little volume [17]. OpenAI's merchant page now says purchases are completed on merchant-owned websites or apps [1].

The traffic numbers tell a split story. In a study of 973 e-commerce sites, ChatGPT sent less than 0.2% of sessions, and those visits converted worse than search or affiliate traffic [18]. Adobe, measuring a later period, found AI traffic to US retail sites up 393% in the first quarter of 2026 and converting 42% better than other traffic by March [19]. Bain puts AI at up to a quarter of referral traffic for some retailers but under 1% of total traffic [20]. Target calls agentic shopping still small in total today [21].

Read together, the numbers describe a channel that is small, growing fast and getting better at sending ready buyers. They also undercount it. A shopper who asks ChatGPT where to find a part, then drives to the store or searches the chain by name, shows up in no AI report. The agent did the deciding and analytics credited something else. That is why readiness is best judged by what agents say about each store. Referral sessions catch only the shoppers who click.

## 4. Agents pick the stores whose facts they can check

OpenAI says ChatGPT's product results are selected independently and are not ads. It draws on structured data from merchants and third-party providers, and ranks merchants on availability, price, quality and whether they make or mainly sell the item [22]. Google says Merchant Center feeds and Business Profiles can help products and services show up in its AI answers, and that no special AI file or markup is needed [23]. Underneath both is the same input: structured product data, store facts and reviews.

> You cannot tune for one ranker. You can be the store whose facts agree everywhere.

How agents weigh that input is less stable than any search ranking. In an academic audit of AI shopping agents, models showed strong position biases and very different sensitivity to price, ratings and reviews, and a new model version could drastically reshuffle market shares [24]. A study reported in Harvard Business Review found that star ratings and price moved agent choices reliably, while scarcity messages, countdown timers and strike-through prices did not, and sometimes backfired [25].

Local answers are even less settled. Across 200,085 local searches, the businesses named overlapped only 20% to 33% from one run to the next [26]. In Yext's study, four AI engines picked the same top business in 4% of cases. The businesses all four agreed on collected far more new reviews each month and answered more of them, though Yext stresses that is an association, not proof [27].

Agents are also far pickier than a map. SOCi found ChatGPT recommended just 1.2% of more than 350,000 locations of multi-location brands, against 35.9% that appeared in Google's local 3-pack [2]. You cannot tune for one ranker that changes with every model release. What holds across all of them is a store whose facts are complete, current and identical everywhere an agent looks.

**Key figure:** 1.2%, of chain locations recommended by ChatGPT, against 35.9% in Google's local 3-pack [2].

## 5. For a chain, every question ends in where and today

E-commerce was 17.1% of US retail sales in the second quarter of 2026 [28]. The other 82.9% was bought some other way, mostly in stores. So for a chain, the most valuable thing an agent can say is a place and a time: in stock at the Riverside store, open until 8, ready for pickup in an hour. Two-thirds of US holiday shoppers say they will confirm online that an item is in stock before going to buy it [29].

Google is already acting on that need. Its AI calls nearby stores about stock, price and promotions, starting with toys, health and beauty, and electronics, then reports back to the shopper [6]. Google's help page lists confirming price and availability among the reasons it calls a business, and lets owners switch the calls off in Business Profile settings [4]. A store that publishes its stock answers the question before the phone rings. A store that doesn't gets the call, in the middle of whatever rush it is having.

That makes the unit of readiness the store-item pair. Take a chain with 40 stores and 8,000 products: that is 320,000 pairs, each with its own answer to three questions. Is it here? What does it cost here? Can I have it today? Google's local inventory feed is built for exactly this, with a store code and an availability value for each item at each store, an optional price and quantity, and how soon it can be picked up [30]. It must be updated at least once a day [31].

OpenAI's product feed has no field for a store at all [32], so whatever ChatGPT knows about one store's shelf has to come from the pages and listings it reads. The harder problem sits underneath. In a landmark study of nearly 370,000 inventory records across 37 stores, 65% were inaccurate [3]. Inside the store, a wrong count is a replenishment headache. Inside an agent's answer, it is a promise to a shopper who drives across town and finds an empty peg.

**Key figure:** 65%, of store inventory records were inaccurate in a study of 37 stores [3].

## 6. What changes for discovery, price, loyalty and the visit

Discovery narrows. A shopper used to open five tabs and walk past three shops; an agent hands over a short list. Your brand gets described in the third person, by reviews, listings and the agent's own summary. Yext found that 86% of the sources AI engines cite are ones brands already control, mostly websites and listings [33]. That is good news with a catch: the plain facts on those pages now carry your brand further than any campaign line does.

> An agent has no habit of driving to your store. It has instructions and linked accounts.

Price comparison becomes free and instant, and there is a useful precedent. When comparison websites made term life insurance easy to compare in the 1990s, prices fell 8% to 15%, and they did not fall for policies the sites did not cover [34]. Expect the same pressure on any product an agent can match by barcode across retailers. Margin moves to what is harder to compare: in stock nearby today, good advice, service, and ranges only you carry.

Loyalty changes shape. An agent has no habit of driving to your store. It has instructions and linked accounts. Google added identity linking to UCP in March 2026, so shoppers get the same member pricing or free shipping through an agent as when they are logged in to the retailer's site [35]. Google also reads loyalty program markup to show member benefits and prices with products in Search [36]. A program an agent can read and link becomes that shopper's default. One it cannot read gets skipped.

The store visit changes too. Fewer trips will be for browsing, and more will arrive with a specific item and a promise the agent made on your behalf. Pickup is already a big share of that traffic: in Salesforce's holiday 2025 data, nearly one in five orders was bought online for pickup in store [37]. When the promise is wrong, the store pays for it at the counter, and often again in a review.

## 7. Most chains are preparing for the wrong step

The first mistake is waiting for checkout. Leadership hears agentic commerce and books a payments project. Meanwhile the platforms are handing checkout to merchants through tools like Shopify's Agentic Storefronts and Google's UCP [13][10], and the biggest test of checkout inside a chatbot was scaled back [16]. The second mistake is treating it as a website SEO project. The website matters, but the agent's question is about a store, and most chains publish one national answer to it.

> Readiness is an org chart problem before it is a technology problem.

The third mistake is blocking agents by accident. Cloudflare began blocking AI crawlers by default for its customers in July 2025 [38]. OpenAI says sites that opt out of its search crawler will not be shown in ChatGPT search answers [39]. Visa lists bot detection that mistakenly blocks legitimate agent transactions as a merchant problem worth solving [11]. Someone at head office should know what your firewall says to each agent, and why.

The fourth mistake is unreadable facts. Return policies live in PDFs, holiday hours in a social post, and prices on a page that loads them by script. Adobe found 34% of retail product pages could not be properly read by AI systems [19]. The fifth is that nobody owns the answer. Merchandising owns the catalog, operations owns stock and hours, the digital team owns the site, the marketing team owns listings and legal owns the returns policy. The agent reads all of them in one pass and notices when they disagree.

| What a typical chain publishes | What an agent needs, store by store |
|---|---|
| One national product feed | Each item's availability at this store |
| Store hours on a locator page | This store's price, including promotions |
| Online prices, not store prices | Today's hours, including holidays |
| A return policy in a PDF | Pickup options and how soon |
| Stock that lives only in the POS | Policies written on the page it reads |

## 8. The risks are real, and most start as bad data

Agents get stores wrong. In Searchable's research across 72,000 questions about UK retailers, one in 16 AI answers was false, wrong postcodes turned up at a rate of one in 10, and smaller retailers were hit harder [40]. OpenAI warns that merchant price and shipping changes can take time to show up in ChatGPT [22]. Google can suspend a merchant's local listings when store prices on its pages keep disagreeing with its feed [41]. An agent quoting last week's price starts an argument at your register.

Agents can also be tricked. Brave showed that hidden text on a web page could hijack an agentic browser into leaking a user's data [42], and OpenAI has said prompt injection is unlikely ever to be fully solved [43]. Liability is unsettled. The Consumer Bankers Association warns that consumers may be liable for their agents' mistakes, and that merchants may face chargebacks even when the shopper authorized the agent [44].

Then there is dependency. A few platforms now set the rules for how your stores appear, and they are adding ads: OpenAI launched product carousel ads built from retailer feeds in August 2026 [45]. Blocking is a weaker defense than it looks. In August 2026 a federal appeals court found that when Perplexity's agent visited Amazon for a user, it was the user who accessed the site, and vacated an injunction against the agent [46]. A chain's real protection is data it owns and publishes in open formats, so no single assistant decides how its stores are described.

## 9. Where this goes in the next five years

Forecasts vary because they count different things. McKinsey projects up to $1 trillion of US retail revenue orchestrated by agents by 2030 [47]. Morgan Stanley puts agentic shoppers at $190 billion to $385 billion of US e-commerce by 2030, 10% to 20% of the market [48]. Bain says $300 billion to $500 billion, with spec-driven items like batteries moving first and apparel and travel later [49]. Treat the range as a direction to plan around.

Here is our view for a chain with stores, in order. Through 2027, discovery dominates. Agents shortlist and send people to sites and stores, and checkout inside chat stays small and arrives through platforms rather than custom builds. By 2028, nearby availability becomes a standard part of shopping answers, because it is the one thing a distant warehouse cannot offer. Chains that publish store-level stock get named. The rest get skipped, or get a robot on the phone.

Next, buying without a human in the loop starts with replenishment: pet food, filters, batteries and household basics, under spending limits backed by card network tokens. Paid placement grows inside answers, so the unpaid shortlist shrinks and data quality decides eligibility for both. Loyalty programs split into the linkable and the invisible. The chain that does best will be the one whose facts an agent can trust at 7am on a Saturday, when the shopper is still in the car.

## 10. The agentic commerce readiness checklist for this year

None of this needs a replatform or a new checkout. It needs unglamorous work, done store by store, with one person accountable for it. Here is the order we would do it in, starting this quarter, for a chain of 10 to 100 stores.

Two decisions sit above the list. Name one owner for store-level truth, with authority across the catalog, stock data, store pages and listings. Then set a stance on agents: which crawlers and buying agents you let in, whether store managers take Google's calls or you switch them off, and whether your loyalty program will link to outside agents.

And skip a few things this year. Don't build your own checkout endpoints; the platforms are doing it. Don't pay for tricks that promise to rank you inside ChatGPT, since the rankers change with every model. Don't publish exact stock counts your systems can't stand behind. Use the limited availability value for thin stock, and save the precision for the stores where your counts are reliable.

1. **Ask what AI says.** Put the 20 questions shoppers ask to ChatGPT, Google AI Mode and Perplexity, for every store. Save every answer and source. That is your baseline.
2. **Make store facts agree.** Hours, holiday hours, phone, services and pickup, identical across Google Business Profile, other listings and your store pages. Fix each one at its source.
3. **Publish each store's stock.** Every item's price and availability for each store, with product codes, in Google's local inventory feed, refreshed at least daily.
4. **Make policies readable.** Returns, pickup, shipping and member prices written on the page and in structured data that matches it. No PDFs for anything an agent should quote.
5. **Open the door, then measure.** Check that your firewall and robots.txt let in the agents you want. Tag AI referrals, re-ask the questions monthly and count wrong answers per store.

## 11. How Ocally helps

Ocally's part in agentic commerce is the data an agent reads before it decides: product feeds, pages AI can read, and checks of what AI says. It does not run checkout. Kai, Ocally's growth agent, works from one Knowledge Store per business, where every product, price, stock reading, store fact and policy keeps its source. He checks every product at every store against what shopping destinations require and names the exact gap. He builds one feed per store, checks it against the destination's published rules, and compares it with current price and stock every day.

Each week he asks ChatGPT, Perplexity and Google AI Overviews the questions shoppers ask about each store, and checks every answer against your facts. Anything that goes public waits for your yes. A growth engineer connects your catalog, POS and listings. A growth specialist checks Kai's work and helps your teams change how they run, store by store.

- **[A readiness check for every store](https://ocally.co/capabilities/product-feeds.md):** Every product at every store checked against what shopping destinations require, with the exact gap named: no photo, no price, no product code, no stock reading. Each gap goes to whoever can fix it.
- **[Store pages agents can read](https://ocally.co/capabilities/websites.md):** Store and product pages checked for structured data that matches what the page shows: products, prices, hours and policies. Fixes are prepared for your approval.
- **[What AI says about each store](https://ocally.co/capabilities/ai-answers.md):** The questions shoppers ask, put to ChatGPT, Perplexity and Google AI Overviews every week for each store. Every answer and its sources are saved and checked against your facts.
- **[Store facts that agree everywhere](https://ocally.co/capabilities/listings.md):** Hours, holiday hours, phone and services checked across your listings, with corrections prepared where they disagree, so an agent finds the same answer wherever it looks.
- **[Reviews that stay current](https://ocally.co/capabilities/reviews.md):** New reviews read for each store and replies drafted for your yes, so the ratings and recent feedback agents weigh stay fresh and answered.

## Questions

### What is agentic commerce?

Agentic commerce is shopping in which an AI agent searches, compares and sometimes buys for a person. The shopper describes a need, and the agent reads product feeds, store pages, listings and reviews to build a shortlist. In 2026 most agentic commerce still ends with a person buying on a retailer's site or in a store, so the data the agent reads decides who gets the sale.

### How does agentic commerce work?

An AI agent turns a request into many searches, reads structured product data and web pages, filters by the shopper's limits and returns a few options. Where checkout is supported, it fills a cart and pays with a tokenized card the shopper approved. Protocols such as OpenAI's Agentic Commerce Protocol and Google's Universal Commerce Protocol give agents a standard way to read catalogs and place orders.

### What is the Agentic Commerce Protocol?

The Agentic Commerce Protocol is an open standard co-developed by OpenAI and Stripe and launched in September 2025. It lets ChatGPT read a merchant's product feed and, where enabled, run checkout while the merchant stays the seller. Since March 2026 OpenAI has focused it on product discovery, and says purchases are completed on merchants' own websites or apps.

### What is the Universal Commerce Protocol?

The Universal Commerce Protocol is an open standard Google announced in January 2026, co-developed with Shopify, Etsy, Wayfair, Target and Walmart. It gives AI agents a common way to work with a merchant's catalog, cart, checkout and loyalty accounts. It powers checkout in Google's AI Mode and the Gemini app, and Google says it now covers hundreds of thousands of brands and retailers.

### How do AI shopping assistants choose which products and stores to recommend?

They weigh structured product data, price, availability, ratings and reviews against what the shopper asked for. OpenAI says ChatGPT ranks merchants on availability, price, quality and whether they make or mainly sell the item. Research shows agents respond strongly to star ratings and price, and that picks shift between models and between runs, so complete, consistent data beats any single tactic.

### What does agentic commerce mean for retailers with physical stores?

It moves the first visit from the aisle to an agent's answer. Most US retail spending still happens in stores, so the valuable answer is local: in stock nearby, at this price, open now. Chains that publish each store's stock, prices, hours and policies in a readable form get named. Where that data is missing, Google's AI already phones stores to ask.

### Do retailers need to build agent checkout?

Not this year, for most chains. OpenAI is moving away from standalone checkout inside ChatGPT, and platforms such as Shopify and Google's Merchant Center are bringing agent checkout to merchants through their own tools. The work that pays now is data: each store's products, prices, stock, hours and policies, complete and current, plus a regular check of what AI assistants say about each store.

### How can a retail chain get ready for agentic commerce?

Start by asking ChatGPT, Google AI Mode and Perplexity the questions shoppers ask about each store, and saving the answers. Then make store facts agree across listings and pages, publish each store's price and availability in Google's local inventory feed at least daily, put policies on the page in matching structured data, check that your firewall lets in the agents you want, and re-measure monthly.

## Sources

1. [OpenAI: Power product discovery in ChatGPT, merchant page (accessed October 2026)](https://chatgpt.com/merchants/)
2. [SOCi: In AI-driven discovery, few brands are chosen, most disappear (January 2026)](https://www.soci.ai/news/in-ai-driven-discovery-few-brands-are-chosen-most-disappear/)
3. [Management Science: DeHoratius and Raman, Inventory Record Inaccuracy: An Empirical Analysis (2008)](https://pubsonline.informs.org/doi/10.1287/mnsc.1070.0789)
4. [Google Business Profile Help: About automated calls and texts from Google to your business (accessed October 2026)](https://support.google.com/business/answer/7690269)
5. [OpenAI: DevDay 2026 recap (September 2026)](https://openai.com/index/devday-2026-recap/)
6. [Google: Google Shopping launches agentic checkout and more AI shopping tools (November 2025)](https://blog.google/products/shopping/agentic-checkout-holiday-ai-shopping/)
7. [Microsoft Advertising: Conversations that convert, Copilot Checkout and Brand Agents (January 2026)](https://about.ads.microsoft.com/en/blog/post/january-2026/conversations-that-convert-copilot-checkout-and-brand-agents)
8. [Stripe: Stripe powers Instant Checkout in ChatGPT and releases Agentic Commerce Protocol codeveloped with OpenAI (September 2025)](https://stripe.com/newsroom/news/stripe-openai-instant-checkout)
9. [Google Cloud: Announcing the Agent Payments Protocol (AP2) (September 2025)](https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol)
10. [Google: New tech and tools for retailers to succeed in an agentic shopping era (January 2026)](https://blog.google/products/ads-commerce/agentic-commerce-ai-tools-protocol-retailers-platforms/)
11. [Visa: Visa introduces Trusted Agent Protocol, a framework for AI commerce (October 2025)](https://usa.visa.com/about-visa/newsroom/press-releases.releaseId.21716.html)
12. [Cloudflare: Securing agentic commerce, helping AI agents transact with Visa and Mastercard (October 2025)](https://blog.cloudflare.com/secure-agentic-commerce/)
13. [Shopify: Millions of merchants can sell in AI chats (March 2026)](https://www.shopify.com/news/agentic-commerce-momentum)
14. [Google: Retailers, get ready for AI-driven holiday shopping (September 2026)](https://blog.google/products-and-platforms/products/shopping/google-shopping-updates-holiday-shopping/)
15. [W3C and GS1: Workshop on e-commerce for humans and AI agents (September 2026)](https://www.w3.org/2026/ecommerce-agents/)
16. [TechCrunch: OpenAI's plans to make ChatGPT more like Amazon aren't going so well (March 2026)](https://techcrunch.com/2026/03/24/openais-plans-to-make-chatgpt-more-like-amazon-arent-going-so-well/)
17. [Modern Retail: What went wrong with ChatGPT's Instant Checkout (March 2026)](https://www.modernretail.co/technology/what-went-wrong-with-chatgpts-instant-checkout/)
18. [Digiday: E-commerce sites see low sales from ChatGPT traffic, new study finds (October 2025)](https://digiday.com/marketing/e-commerce-sites-see-low-sales-from-chatgpt-traffic-new-study-finds/)
19. [TechCrunch: AI traffic to US retailers rose 393% in Q1, and it's boosting their revenue too (April 2026)](https://techcrunch.com/2026/04/16/ai-traffic-to-us-retailers-rose-393-in-q1-and-its-boosting-their-revenue-too/)
20. [Bain & Company: Agentic AI poised to disrupt retail, even with 50% of consumers cautious of fully autonomous purchases (November 2025)](https://www.bain.com/about/media-center/press-releases/20252/agentic-ai-poised-to-disrupt-retail-even-with-50-of-consumers-cautious-of-fully-autonomous-purchasesbain--company/)
21. [Target: Q2 2026 earnings call transcript (August 2026)](https://corporate.target.com/getmedia/d01cb805-63cb-4150-bae4-8d717f0d9ed3/Q2-2026-Target-Corp-Earnings-Call.pdf)
22. [OpenAI Help Center: Shopping with ChatGPT Search (updated October 2026)](https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search)
23. [Google Search Central: Google's guide to optimizing for generative AI features on Google Search (updated July 2026)](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)
24. [arXiv: Allouah, Besbes, Figueroa, Kanoria and Kumar, What Is Your AI Agent Buying? (version 3, December 2025)](https://arxiv.org/abs/2508.02630)
25. [Harvard Business Review: Sabbah and Acar, Research: Traditional marketing doesn't work on AI shopping agents (May 2026)](https://store.hbr.org/product/research-traditional-marketing-doesn-t-work-on-ai-shopping-agents/H096G7)
26. [BrightLocal: Study, AI local search data (September 2026)](https://www.brightlocal.com/research/local-ai-visibility-study/)
27. [Yext Research: Four AIs, one question, four different answers (September 2026)](https://www.yext.com/research/articles/four-ais-four-different-answers/)
28. [U.S. Census Bureau: Quarterly Retail E-Commerce Sales, 2nd Quarter 2026 (August 2026)](https://www.census.gov/retail/ecommerce.html)
29. [Google: 4 ways to use Google Shopping's AI tools for Black Friday, Cyber Monday (November 2025)](https://blog.google/products-and-platforms/products/shopping/black-friday-cyber-monday-shopping-tips-ai/)
30. [Google Merchant Center Help: Local inventory data specification (accessed October 2026)](https://support.google.com/merchants/answer/14819809)
31. [Google Merchant Center Help: Feed update frequency for local products (accessed October 2026)](https://support.google.com/merchants/answer/7371671)
32. [OpenAI Developers: Products, the Agentic Commerce product feed specification (accessed October 2026)](https://developers.openai.com/commerce/specs/file-upload/products)
33. [Yext: 86% of AI citations come from brand-managed sources (October 2025)](https://yext.com/about/news-media/ai-citations-release)
34. [NBER: Brown and Goolsbee, Does the Internet Make Markets More Competitive? (2000, published in the Journal of Political Economy 2002)](https://www.nber.org/papers/w7996)
35. [Google: Universal Commerce Protocol updates improve AI shopping for retailers (March 2026)](https://blog.google/products-and-platforms/products/shopping/ucp-updates/)
36. [Google Search Central: Loyalty program structured data, MemberProgram (updated September 2026)](https://developers.google.com/search/docs/appearance/structured-data/loyalty-program)
37. [Salesforce: Salesforce reveals 2025 holiday shopping data (January 2026)](https://www.salesforce.com/news/stories/2025-holiday-shopping-data/)
38. [MIT Technology Review: Cloudflare will now block AI bots from crawling its clients' websites by default (July 2025)](https://www.technologyreview.com/2025/07/01/1119498/cloudflare-will-now-by-default-block-ai-bots-from-crawling-its-clients-websites/)
39. [OpenAI: Overview of OpenAI crawlers (accessed October 2026)](https://developers.openai.com/api/docs/bots)
40. [Retail Tech Innovation Hub: Many UK high street retailers are suffering from AI chatbot inaccuracies, Searchable research finds (July 2026)](https://retailtechinnovationhub.com/home/2026/7/15/many-uk-high-street-retailers-are-suffering-from-ai-chatbot-inaccuracies-searchable-research-finds)
41. [Google Merchant Center Help: How to fix inconsistent prices for in-store product pages (accessed October 2026)](https://support.google.com/merchants/answer/15070419)
42. [Brave: Agentic browser security, indirect prompt injection in Perplexity Comet (August 2025)](https://brave.com/blog/comet-prompt-injection/)
43. [TechCrunch: OpenAI says AI browsers may always be vulnerable to prompt injection attacks (December 2025)](https://techcrunch.com/2025/12/22/openai-says-ai-browsers-may-always-be-vulnerable-to-prompt-injection-attacks/)
44. [Consumer Bankers Association: Agentic AI Payments, navigating consumer protection, innovation and regulatory frameworks (January 2026)](https://consumerbankers.com/wp-content/uploads/2026/01/CBA-Agentic-Symposium-White-Paper-2026-01v2.pdf)
45. [Modern Retail: OpenAI brings product carousels to ChatGPT ads (August 2026)](https://www.modernretail.co/marketing/openai-brings-product-carousels-to-chatgpt-ads/)
46. [Cooley: Ninth Circuit rules on AI agent access to third-party websites under the CFAA (August 2026)](https://www.cooley.com/news/insight/2026/2026-08-06-ninth-circuit-rules-on-ai-agent-access-to-third-party-websites-under-cfaa)
47. [Digital Commerce 360: McKinsey, up to $5 trillion in agentic commerce sales by 2030 (October 2025)](https://www.digitalcommerce360.com/2025/10/20/mckinsey-forecast-5-trillion-agentic-commerce-sales-2030/)
48. [Morgan Stanley: Agentic commerce impact could reach $385 billion by 2030 (December 2025)](https://www.morganstanley.com/insights/articles/agentic-commerce-market-impact-outlook)
49. [Bain & Company: 2030 forecast, how agentic AI will reshape US retail (December 2025)](https://www.bain.com/insights/2030-forecast-how-agentic-ai-will-reshape-us-retail-snap-chart/)

**Find out which of your stores agents can read.** In a free consultation, we'll walk through what Kai would check first in each store, and what your team would need to do. [Book a free consultation](https://ocally.co/consultation.md).

## Next steps

- [Is Ocally right for this business?](https://ocally.co/evaluate.md): who it fits and who it does not, the cost, the risks, and how to start.
- [Pricing](https://ocally.co/pricing.md): customized, with a minimum of $500 per location per month.
- [Book a free consultation](https://ocally.co/consultation.md): 30 minutes on one store. Calendar: https://cal.com/ocally/30min.
- [See if we're a fit](https://ocally.co/fit.md): a two-minute check; an agent can submit it for its human, with consent.
- [Every doc](https://ocally.co/llms.txt): the index.

Questions a doc doesn't answer: onboard@ocally.co.