Your Next Online Purchase May Be ‘Negotiated’ by a Bot — The New Standard Retailers Quietly Built for AI Agents in 2026
In-assistant checkout turned shopping agents from “helpful” into transactional. The real shift isn’t smarter AI—it’s the new trust rails letting merchants safely accept agent-driven purchases.
By TheMurrow Editorial
March 22, 2026
Key Points
- 1Watch checkout move into assistants: agent-led flows now handle discover, confirm, authenticate, and pay without sending you to retailer pages.
- 2Understand the real breakthrough: trust rails—identity, authorization, accountability—let merchants accept consumer-approved agents without treating them like hostile bots.
- 3Protect yourself with boundaries: set constraints, require confirmation for payments and subscriptions, and keep itemized summaries for disputes and refunds.
Your next big purchase may not be “made online” so much as delegated.
Picture the moment: you tell your assistant what you need, set a few rules—deliver by Friday, under $300, prefer refurbished if the warranty is solid—and go back to your day. While you’re in meetings, a software agent compares sellers, checks return policies, applies loyalty perks, and completes checkout. You get a clean summary and a final confirm prompt. The receipt lands before lunch.
That scenario stopped sounding like science fiction in late 2025. Not because consumers suddenly embraced “shopping bots,” but because the plumbing changed: major AI platforms and payment networks began enabling in-assistant checkout, reducing the friction that kept agents stuck in recommendation mode. Axios described the shift plainly: major platforms moved from “search + click” toward agent-led purchase flows—discover, confirm, authenticate, pay—without being bounced to a conventional retailer site. (Axios, Sept. 29, 2025)
The quieter story sits underneath: a new layer of protocols and trust frameworks that let merchants distinguish a legitimate, consumer-authorized agent from the bot traffic retailers have spent two decades trying to block.
The breakthrough wasn’t a smarter bot. It was giving bots a way to pay—safely, with permission.
— — TheMurrow
The new meaning of “negotiated by a bot” (and why it suddenly works)
In 2026, the phrase “negotiated by a bot” misleads as often as it clarifies. Most consumer commerce does not involve an AI agent bartering in a bazaar. The “negotiation” is usually more subtle—and more scalable.
Negotiation as automated deal-seeking
For many purchases, agent negotiation looks like structured optimization:
- Comparing prices across sellers
- Weighing shipping speed against cost
- Filtering by return windows and warranties
- Applying loyalty perks or member pricing where available
A capable agent can do that in seconds, repeatedly, without fatigue. The consumer experience becomes less like browsing and more like setting a policy: what you value, what you’ll compromise on, and what you won’t.
- Comparing prices across sellers
- Weighing shipping speed against cost
- Filtering by return windows and warranties
- Applying loyalty perks or member pricing where available
A capable agent can do that in seconds, repeatedly, without fatigue. The consumer experience becomes less like browsing and more like setting a policy: what you value, what you’ll compromise on, and what you won’t.
Negotiation as rule-based tradeoffs
The more important shift is that users can specify constraints:
- “I’ll accept open-box if it’s at least 15% cheaper.”
- “Only buy if delivery is before Friday.”
- “I want the most reliable seller, not the lowest price.”
That structure turns “shopping” into something closer to a procurement workflow. Consumers already think this way; they simply haven’t had tools that reliably execute those preferences across multiple merchants.
- “I’ll accept open-box if it’s at least 15% cheaper.”
- “Only buy if delivery is before Friday.”
- “I want the most reliable seller, not the lowest price.”
That structure turns “shopping” into something closer to a procurement workflow. Consumers already think this way; they simply haven’t had tools that reliably execute those preferences across multiple merchants.
The missing ingredient: transaction capability
Before late 2025, assistants could recommend products, but they couldn’t consistently close the loop. Checkout meant leaving the assistant, re-entering payment details, passing fraud checks, and navigating a merchant site. That friction wasn’t cosmetic—it was structural.
Axios reported that major platforms began enabling in-chat checkout, allowing the flow to remain inside the assistant from discovery to payment. (Axios, Sept. 29, 2025)
Axios reported that major platforms began enabling in-chat checkout, allowing the flow to remain inside the assistant from discovery to payment. (Axios, Sept. 29, 2025)
Agents don’t need charm to negotiate. They need rails—identity, authorization, and a checkout that merchants won’t block.
— — TheMurrow
The real story is the rails: protocols built for agent commerce
The most consequential development isn’t a single killer assistant. It’s the emergence of interoperable standards—technical languages and trust mechanisms that let agents transact across merchants without bespoke one-off integrations.
Three themes show up across the major announcements:
1. Identity: the agent can prove who (or what) it is
2. Authorization: the agent can prove the consumer permitted a specific action
3. Accountability: the system can log what happened and resolve disputes
These are not abstract concepts. They are the practical prerequisites for letting “bots” into the parts of commerce that used to be guarded by CAPTCHAs, fraud tooling, and bot mitigation systems.
Visa framed the core challenge as a trust problem—distinguishing legitimate, consumer-authorized agents from malicious automation. (Visa press release, Oct. 14, 2025)
Stripe and OpenAI framed it as a standardization problem: merchants need a consistent protocol, not a new integration for every assistant. (Stripe newsroom, Sept. 29, 2025)
Mastercard emphasized tokenization and acceptance specifications so merchants can identify trusted agent traffic. (Mastercard press release, Apr. 29, 2025; Mastercard commentary, 2025–2026)
Even if these frameworks compete, they share an implicit concession: agentic commerce is coming, and it will only work if merchants can trust it.
Three themes show up across the major announcements:
1. Identity: the agent can prove who (or what) it is
2. Authorization: the agent can prove the consumer permitted a specific action
3. Accountability: the system can log what happened and resolve disputes
These are not abstract concepts. They are the practical prerequisites for letting “bots” into the parts of commerce that used to be guarded by CAPTCHAs, fraud tooling, and bot mitigation systems.
Visa framed the core challenge as a trust problem—distinguishing legitimate, consumer-authorized agents from malicious automation. (Visa press release, Oct. 14, 2025)
Stripe and OpenAI framed it as a standardization problem: merchants need a consistent protocol, not a new integration for every assistant. (Stripe newsroom, Sept. 29, 2025)
Mastercard emphasized tokenization and acceptance specifications so merchants can identify trusted agent traffic. (Mastercard press release, Apr. 29, 2025; Mastercard commentary, 2025–2026)
Even if these frameworks compete, they share an implicit concession: agentic commerce is coming, and it will only work if merchants can trust it.
Trust primitives every agent checkout needs
Identity (who/what the agent is), authorization (what you permitted), and accountability (logs + dispute resolution) are the non-negotiables behind “bot shopping.”
OpenAI + Stripe’s ACP and “Instant Checkout”: the assistant becomes a cashier
On Sept. 29, 2025, Stripe announced it powers Instant Checkout in ChatGPT and released the Agentic Commerce Protocol (ACP), described as an open standard co-developed with OpenAI for programmatic commerce flows between buyers, AI agents, and businesses. (Stripe newsroom, Sept. 29, 2025)
That date matters. It marks a shift from “AI that helps you shop” to “AI that can complete a purchase,” at least within supported merchant environments.
That date matters. It marks a shift from “AI that helps you shop” to “AI that can complete a purchase,” at least within supported merchant environments.
What ACP is trying to standardize
Stripe positioned ACP as a shared technical language so merchants don’t have to build custom flows for every AI platform. Standardization is not glamorous, but it’s how payment ecosystems scale. A protocol makes routine what used to be exceptional.
Axios and AP reporting described Etsy as an initial supported surface at launch, with broader merchant support positioned as “coming soon.” (Axios, Sept. 29, 2025)
Later, OpenAI highlighted Instacart as a deeper integration: a grocery and checkout experience inside ChatGPT, explicitly stating Instacart’s checkout is powered by ACP. (OpenAI partnership announcement)
Salesforce also announced collaboration around ACP for commerce integrations on Oct. 14, 2025. (Salesforce press release, Oct. 14, 2025)
Those partnerships are not just logos on a slide. They indicate where agent checkout is most immediately valuable:
- Marketplaces with complex inventory and seller variation
- Groceries and replenishment shopping where speed matters
- Enterprise commerce stacks where integration can spread fast
Axios and AP reporting described Etsy as an initial supported surface at launch, with broader merchant support positioned as “coming soon.” (Axios, Sept. 29, 2025)
Later, OpenAI highlighted Instacart as a deeper integration: a grocery and checkout experience inside ChatGPT, explicitly stating Instacart’s checkout is powered by ACP. (OpenAI partnership announcement)
Salesforce also announced collaboration around ACP for commerce integrations on Oct. 14, 2025. (Salesforce press release, Oct. 14, 2025)
Those partnerships are not just logos on a slide. They indicate where agent checkout is most immediately valuable:
- Marketplaces with complex inventory and seller variation
- Groceries and replenishment shopping where speed matters
- Enterprise commerce stacks where integration can spread fast
Practical implication for readers
If a protocol like ACP gains traction, consumers will increasingly “shop” by describing intent and constraints rather than navigating product pages. The assistant becomes a cashier—one that can search, compare, and execute within approved boundaries.
Checkout is the choke point. Once it moves into the assistant, everything upstream—search, ads, product pages—has to adapt.
— — TheMurrow
Visa’s Trusted Agent Protocol: keeping bots out, letting trusted agents in
On Oct. 14, 2025, Visa announced Trusted Agent Protocol, describing it as an ecosystem-led, open framework built on existing web infrastructure to enable safe agent-driven checkout. The explicit goal: help merchants distinguish malicious bots from legitimate AI agents acting on behalf of consumers. (Visa press release, Oct. 14, 2025)
Visa also markets Visa Intelligent Commerce as a service layer for secure agentic commerce across the Visa network, emphasizing credentials, tokenization, and trust. (Visa Intelligent Commerce overview)
Visa also markets Visa Intelligent Commerce as a service layer for secure agentic commerce across the Visa network, emphasizing credentials, tokenization, and trust. (Visa Intelligent Commerce overview)
Why Visa’s framing matters
Merchants have spent years fighting bots that:
- Scrape pricing and inventory
- Hoard limited supply (think ticketing and launches)
- Run fraud attempts at scale
If “agents” look like bots, they’ll get blocked like bots. Visa’s approach treats legitimacy as something to be proven rather than assumed.
Visa’s language—authenticity, authorization, accountability—signals that the company expects agent traffic to become normal and wants the network to handle it without increasing fraud.
- Scrape pricing and inventory
- Hoard limited supply (think ticketing and launches)
- Run fraud attempts at scale
If “agents” look like bots, they’ll get blocked like bots. Visa’s approach treats legitimacy as something to be proven rather than assumed.
Visa’s language—authenticity, authorization, accountability—signals that the company expects agent traffic to become normal and wants the network to handle it without increasing fraud.
The consumer angle: permission and recourse
Consumers should read these announcements as a push toward clearer boundaries:
- What did you authorize the agent to do?
- What details can it share with merchants?
- What happens when something goes wrong—wrong item, wrong address, disputed charge?
Visa is effectively arguing that agent commerce must be auditable. That’s not just good governance; it’s how the system avoids collapsing under its own complexity.
- What did you authorize the agent to do?
- What details can it share with merchants?
- What happens when something goes wrong—wrong item, wrong address, disputed charge?
Visa is effectively arguing that agent commerce must be auditable. That’s not just good governance; it’s how the system avoids collapsing under its own complexity.
Mastercard Agent Pay: tokenization, acceptance specs, and the merchant side of trust
On April 29, 2025, Mastercard unveiled Mastercard Agent Pay, describing it as agentic payments technology with references to tokenization and work with checkout/acquirer partners such as Braintree and Checkout.com. (Mastercard press release, Apr. 29, 2025)
In additional commentary, Mastercard described extending bot-authentication approaches into merchant specifications so merchants can identify “trusted agentic traffic.” (Mastercard story, 2025)
By 2026, Mastercard explicitly situated agent commerce as a multi-protocol world, referencing OpenAI’s ACP and Google’s Agent Payments Protocol among others. (Mastercard, 2026 “rules of the road” commentary)
In additional commentary, Mastercard described extending bot-authentication approaches into merchant specifications so merchants can identify “trusted agentic traffic.” (Mastercard story, 2025)
By 2026, Mastercard explicitly situated agent commerce as a multi-protocol world, referencing OpenAI’s ACP and Google’s Agent Payments Protocol among others. (Mastercard, 2026 “rules of the road” commentary)
Why tokenization is central here
Mastercard’s emphasis aligns with a practical reality: agents should not need access to your raw card details. Tokenization and credential management are a way to:
- Reduce the blast radius of breaches
- Limit misuse if an agent is compromised
- Enable scoped permissions (what can be bought, when, and from whom)
The consumer benefit isn’t novelty; it’s safety. The merchant benefit is lower risk and cleaner acceptance standards for traffic that would otherwise look suspicious.
- Reduce the blast radius of breaches
- Limit misuse if an agent is compromised
- Enable scoped permissions (what can be bought, when, and from whom)
The consumer benefit isn’t novelty; it’s safety. The merchant benefit is lower risk and cleaner acceptance standards for traffic that would otherwise look suspicious.
What merchants will likely demand
Even without predicting new features, the direction is visible in Mastercard’s framing: merchants want a reliable signal that an agent is authorized and accountable. Acceptance specs and identity signals become the equivalent of “chip-and-PIN” for the agent era: infrastructure that normalizes a new kind of transaction.
Google’s Agent Payments Protocol (AP2): a competing rulebook with the same underlying problem
Axios reported that Google announced an Agent Payments Protocol (AP2) in September 2025, described as an open framework for how AI agents can transact with user authorization, emphasizing authorization, authenticity, and accountability. (Axios, Sept. 16, 2025)
AP also reported Google partnerships around agent shopping in early 2026. (AP, Jan. 11, 2026)
AP also reported Google partnerships around agent shopping in early 2026. (AP, Jan. 11, 2026)
The significance of “open rulebooks”
When multiple giants publish “open” protocols, skepticism is healthy. Yet open rulebooks can still be useful, even when companies have strategic motives. Payments and web standards often advance through messy, overlapping frameworks that slowly converge through adoption pressure.
Google’s entry underscores two realities:
1. Agent commerce is too large for a single vendor to own.
2. Platforms expect merchants to support multiple assistants and multiple protocols—eventually through intermediaries that smooth the differences.
Google’s entry underscores two realities:
1. Agent commerce is too large for a single vendor to own.
2. Platforms expect merchants to support multiple assistants and multiple protocols—eventually through intermediaries that smooth the differences.
What it means for consumers
Consumers will not choose a protocol. They’ll choose an experience. The protocols operate behind the scenes, shaping:
- Which merchants are available inside an assistant
- How much confirmation is required before a purchase
- How disputes and refunds work
If Google, OpenAI/Stripe, Visa, and Mastercard converge on similar trust primitives, consumers may see fewer “dead ends” where an assistant recommends something but can’t complete checkout.
- Which merchants are available inside an assistant
- How much confirmation is required before a purchase
- How disputes and refunds work
If Google, OpenAI/Stripe, Visa, and Mastercard converge on similar trust primitives, consumers may see fewer “dead ends” where an assistant recommends something but can’t complete checkout.
Where agent negotiation will show up first: groceries, marketplaces, travel, and subscriptions
Agent-driven purchasing works best where the consumer goal is clear and the comparison set is messy.
Case study: Instacart inside ChatGPT
OpenAI highlighted Instacart as a deeply integrated grocery experience with checkout powered by ACP. (OpenAI partnership announcement)
Grocery is a prime early category because:
- Replenishment is repetitive
- Preferences are stable (brands, dietary constraints)
- Substitutions and delivery windows matter
- Time savings are obvious
An agent can “negotiate” by finding the best combination of price, availability, and delivery slots—then executing without the user juggling carts across apps.
Grocery is a prime early category because:
- Replenishment is repetitive
- Preferences are stable (brands, dietary constraints)
- Substitutions and delivery windows matter
- Time savings are obvious
An agent can “negotiate” by finding the best combination of price, availability, and delivery slots—then executing without the user juggling carts across apps.
Marketplaces and variable seller quality
Axios and AP described Etsy as an early supported surface for Instant Checkout in ChatGPT. (Axios, Sept. 29, 2025)
Marketplace shopping involves tradeoffs: seller ratings, shipping times, return policies, and slight differences between listings. Agents are well suited to summarizing those tradeoffs—and selecting based on a user’s stated priorities.
Marketplace shopping involves tradeoffs: seller ratings, shipping times, return policies, and slight differences between listings. Agents are well suited to summarizing those tradeoffs—and selecting based on a user’s stated priorities.
Where negotiation becomes more literal
The research notes also point to B2B-like quoting behaviors creeping into consumer commerce: travel, high-ticket items, subscriptions, and some marketplaces are preparing for quote/offer flows. In these environments, “negotiation” can include structured offers: bundles, upgrades, flexible cancellation, or price matching—less haggling, more configurable terms.
Risks, frictions, and the questions no protocol can dodge
Agent commerce has a persuasive narrative: fewer tabs, fewer forms, less time wasted. The hard part is governance—figuring out how much autonomy consumers want to grant.
The authorization problem: what did you actually permit?
“User authorized” sounds clean until you test edge cases:
- Did you authorize only one purchase—or ongoing replenishment?
- Did you authorize “up to $300,” or “exactly this item at this price”?
- Did you authorize sharing your address with any seller the agent selects?
Visa’s emphasis on authorization and accountability hints at how central these questions are. (Visa press release, Oct. 14, 2025)
- Did you authorize only one purchase—or ongoing replenishment?
- Did you authorize “up to $300,” or “exactly this item at this price”?
- Did you authorize sharing your address with any seller the agent selects?
Visa’s emphasis on authorization and accountability hints at how central these questions are. (Visa press release, Oct. 14, 2025)
The merchant problem: bots, fraud, and customer service load
Merchants will ask: if an agent buys the wrong item, who eats the cost? If an agent triggers fraudulent transactions, can merchants reliably trace and block the source without blocking legitimate customers?
That’s why payment networks are building signals to distinguish trusted agents from malicious automation. The protocols are a bet that trust can be systematized.
That’s why payment networks are building signals to distinguish trusted agents from malicious automation. The protocols are a bet that trust can be systematized.
The consumer problem: transparency and control
Delegation only feels safe when it’s legible. Consumers will expect:
- Clear receipts and itemized summaries
- Easy cancellations and returns
- A way to audit why the agent chose Seller A over Seller B
- Strong default limits on spending and data sharing
Practical takeaway: the best way to use agent checkout in 2026 is to treat it like granting access to a financial app—start narrow, set constraints, and expand permissions only when the system has earned trust.
- Clear receipts and itemized summaries
- Easy cancellations and returns
- A way to audit why the agent chose Seller A over Seller B
- Strong default limits on spending and data sharing
Practical takeaway: the best way to use agent checkout in 2026 is to treat it like granting access to a financial app—start narrow, set constraints, and expand permissions only when the system has earned trust.
What to do now: practical takeaways for readers who don’t want surprises
Most readers don’t need to become protocol experts. A few habits will matter more than any acronym.
Set rules before you set it loose
If you use in-assistant checkout, define constraints in plain language:
- Maximum price and acceptable alternatives (new vs. refurbished/open-box)
- Delivery deadlines that matter
- “Never buy from unknown sellers” (or similar quality thresholds)
- Return-policy minimums
Agents are good at optimization; they’re bad at guessing what you consider unacceptable.
- Maximum price and acceptable alternatives (new vs. refurbished/open-box)
- Delivery deadlines that matter
- “Never buy from unknown sellers” (or similar quality thresholds)
- Return-policy minimums
Agents are good at optimization; they’re bad at guessing what you consider unacceptable.
Demand confirmation on the irreversible steps
A sensible pattern is: let the agent research freely, but require explicit confirmation for:
- Payment submission
- Subscription starts/renewals
- High-value items
- Shipping address changes
- Payment submission
- Subscription starts/renewals
- High-value items
- Shipping address changes
Keep receipts and summaries
As agent commerce grows, your “paper trail” becomes your leverage in disputes. Favor assistants and merchants that provide clear transaction summaries—what was chosen, why, and under what authorization.
Delegation without a paper trail isn’t convenience. It’s liability.
— — TheMurrow
A safer 2026 setup for in-assistant checkout
- ✓Set a maximum price and acceptable condition (new/refurb/open-box)
- ✓Require delivery deadlines and return-policy minimums
- ✓Block unknown sellers or require rating thresholds
- ✓Require explicit confirmation for payment, subscriptions, and address changes
- ✓Keep itemized summaries and receipts for disputes
The Murrow takeaway: the web’s next battle is over checkout, not search
For two decades, online commerce has been organized around the page: product pages, comparison pages, checkout pages. Agentic commerce reorganizes the experience around intent: what you want, under what conditions, with what permissions.
The 2025–2026 wave of announcements—Stripe and OpenAI’s ACP and Instant Checkout (Sept. 29, 2025), Visa’s Trusted Agent Protocol and Intelligent Commerce (Oct. 14, 2025), Mastercard’s Agent Pay (Apr. 29, 2025), and Google’s AP2 (September 2025 per Axios)—share a common thesis. Agents are going to transact, not just recommend, and commerce infrastructure must decide which automated actors it will trust.
That does not guarantee a frictionless future. It does suggest the argument is over: the industry is building for it. For consumers, the near-term task is simple and unglamorous—set boundaries, insist on transparency, and treat shopping delegation like any other financial permission you grant.
The companies will sell it as convenience. The real prize is control: whoever mediates your intent will shape what you buy, from whom, and on what terms. That’s negotiation—whether anyone haggles or not.
The 2025–2026 wave of announcements—Stripe and OpenAI’s ACP and Instant Checkout (Sept. 29, 2025), Visa’s Trusted Agent Protocol and Intelligent Commerce (Oct. 14, 2025), Mastercard’s Agent Pay (Apr. 29, 2025), and Google’s AP2 (September 2025 per Axios)—share a common thesis. Agents are going to transact, not just recommend, and commerce infrastructure must decide which automated actors it will trust.
That does not guarantee a frictionless future. It does suggest the argument is over: the industry is building for it. For consumers, the near-term task is simple and unglamorous—set boundaries, insist on transparency, and treat shopping delegation like any other financial permission you grant.
The companies will sell it as convenience. The real prize is control: whoever mediates your intent will shape what you buy, from whom, and on what terms. That’s negotiation—whether anyone haggles or not.
2026
The year “agentic commerce” shifted from recommendations to real transactions—driven by in-assistant checkout and new trust protocols.
Sept. 29, 2025
Stripe announced Instant Checkout in ChatGPT and released the Agentic Commerce Protocol (ACP) with OpenAI.
Oct. 14, 2025
Visa announced Trusted Agent Protocol, aiming to help merchants distinguish legitimate consumer-authorized agents from malicious bots.
Apr. 29, 2025
Mastercard unveiled Agent Pay, emphasizing tokenization and merchant acceptance specifications for trusted agentic traffic.
1) What does “negotiated by a bot” actually mean?
In most cases, it means an AI agent optimizes tradeoffs rather than haggles. The agent compares price, shipping, availability, return policies, and perks across sellers, then recommends or selects the option that best matches your rules. True “offer and counteroffer” negotiation is more likely in categories like travel, high-ticket items, and subscriptions.
2) Why is bot-negotiated checkout suddenly possible in 2026?
The big change is in-assistant checkout. Reporting in late 2025 described a shift from “search + click” toward agent-led purchase flows where discovery, confirmation, authentication, and payment can happen without sending you to a retailer’s website (Axios, Sept. 29, 2025). Protocols and payment rails are being built to support that safely.
3) What is the Agentic Commerce Protocol (ACP)?
ACP is an “open standard” Stripe said it co-developed with OpenAI, announced alongside Instant Checkout in ChatGPT on Sept. 29, 2025 (Stripe newsroom). The goal is to give merchants and assistants a shared technical language for programmatic commerce—so transactions can be completed reliably inside an AI experience instead of through custom integrations.
4) How do Visa and Mastercard fit into agent shopping?
They’re focused on trust and payment security. Visa announced a Trusted Agent Protocol on Oct. 14, 2025, aiming to help merchants distinguish legitimate, consumer-authorized agents from malicious bots (Visa press release). Mastercard announced Agent Pay on Apr. 29, 2025, emphasizing tokenization and merchant acceptance approaches for trusted agent traffic (Mastercard press release).
5) Is Google doing something similar?
Yes. Axios reported Google announced an Agent Payments Protocol (AP2) in September 2025, described as an open framework emphasizing authorization, authenticity, and accountability (Axios, Sept. 16, 2025). AP also reported Google partnerships around agent shopping in early 2026 (AP, Jan. 11, 2026), reinforcing that multiple ecosystems are competing to define the rules.
6) What should I watch for to stay safe using agent checkout?
Look for clear authorization controls and transparency:
Editor’s Note
The provided article text ends immediately after: “Look for clear authorization controls and transparency:”. No additional bullets or sentences were included to preserve.
T
About the Author
TheMurrow Editorial is a writer for TheMurrow covering trends.
Frequently Asked Questions
What does “negotiated by a bot” actually mean?
In most cases, it means an AI agent optimizes tradeoffs rather than haggles. The agent compares price, shipping, availability, return policies, and perks across sellers, then recommends or selects the option that best matches your rules. True “offer and counteroffer” negotiation is more likely in categories like travel, high-ticket items, and subscriptions.
Why is bot-negotiated checkout suddenly possible in 2026?
The big change is in-assistant checkout. Reporting in late 2025 described a shift from “search + click” toward agent-led purchase flows where discovery, confirmation, authentication, and payment can happen without sending you to a retailer’s website (Axios, Sept. 29, 2025). Protocols and payment rails are being built to support that safely.
What is the Agentic Commerce Protocol (ACP)?
ACP is an “open standard” Stripe said it co-developed with OpenAI, announced alongside Instant Checkout in ChatGPT on Sept. 29, 2025 (Stripe newsroom). The goal is to give merchants and assistants a shared technical language for programmatic commerce—so transactions can be completed reliably inside an AI experience instead of through custom integrations.
How do Visa and Mastercard fit into agent shopping?
They’re focused on trust and payment security. Visa announced a Trusted Agent Protocol on Oct. 14, 2025, aiming to help merchants distinguish legitimate, consumer-authorized agents from malicious bots (Visa press release). Mastercard announced Agent Pay on Apr. 29, 2025, emphasizing tokenization and merchant acceptance approaches for trusted agent traffic (Mastercard press release).
Is Google doing something similar?
Yes. Axios reported Google announced an Agent Payments Protocol (AP2) in September 2025, described as an open framework emphasizing authorization, authenticity, and accountability (Axios, Sept. 16, 2025). AP also reported Google partnerships around agent shopping in early 2026 (AP, Jan. 11, 2026), reinforcing that multiple ecosystems are competing to define the rules.
What should I watch for to stay safe using agent checkout?
The article text provided ends right after this prompt (“Look for clear authorization controls and transparency:”), so no further guidance was included to preserve.
