Before You Let AI Do Your Holiday Shopping: 7 Rules From a Security Expert
Before an AI agent shops for you, use a credit card, cap its spending, keep approval on and name the stores it may use. Seven rules for the 2026 holiday season.

Before you let an AI agent do your holiday shopping, connect a credit card instead of a debit card, cap what it can spend, keep purchase approval on, and tell it which stores it may use. Those four settings decide most of what happens when an agent buys the wrong thing or lands on a fake store. Below are the seven rules I would follow, and what to do if something still goes wrong.
This is a consumer guide, not legal advice. The payment rights below are US rules unless I say otherwise.
Key facts
- AI shopping is about to spike. Adobe forecasts that AI-driven traffic to US retail sites will rise 130% year over year this season (November 1 to December 31), and 159% on Thanksgiving. It expects a record $275.1 billion in US online spending.
- Agents can now buy for you. Amazon's Alexa for Shopping can buy items from other merchants' sites and can auto-buy at a target price. Google's agentic checkout buys on a merchant's site after you confirm.
- Fake stores are multiplying. Scamwise users report a new fake store roughly every three minutes, according to its October holiday report.
- How you pay decides your rights. In the US, credit cards carry a legal right to dispute goods "not delivered as agreed." Federal law gives debit cards no equivalent right.
What AI shopping agents can actually do this season
Ask for "a thoughtful gift under $75 for my sister who loves hiking and just moved to Denver," and a good AI agent will return options in seconds. The newer part is that some agents now finish the purchase too.
- Amazon. Rufus was renamed Alexa for Shopping on May 13, 2026. On some searches, its "Buy for Me" button lets Amazon buy an item from another merchant's website for you. Prime members can also ask it to auto-buy an item when it hits a target price. Amazon notifies you after the order is placed and gives you a free 24-hour cancellation window.
- Google. Agentic checkout in Search and AI Mode can buy from eligible US merchants with Google Pay. Google says it always asks permission first and buys only after you confirm the purchase and shipping details.
- Meta and others. Meta's Muse agent, launched in September, picks products and proceeds to checkout with your approval. I covered the wider landscape, including how Visa and Mastercard built the payment rails, in my agentic commerce overview.
Notice the difference. Google waits for your confirmation. Amazon's auto-buy places the order first and lets you cancel afterward. Know which model your tool uses before you hand it a gift list.
Why AI shopping changes the scam playbook
Holiday scams are not new. What is new is who they are aimed at.
When you shop yourself, you might notice that a store looks off, that the deal is too good, or that the web address is slightly wrong. An agent may not. It reads text and data, and a well-built fake store can have all the right text and data.
Visa has warned that AI shopping agents "can be deceived by sophisticated counterfeit merchants engineered specifically to exploit them," in its analysis of agentic commerce threats. In October, Inc. reported Visa repeating that warning for the holiday season: agents need to tell a real merchant from a fake one before any money moves.
The fake stores are also arriving faster. Savi Security's Scamwise holiday report, published October 20, gives these figures from its users' submissions:
- A new fake store is reported roughly every three minutes, about 500 a day.
- Fake holiday-themed stores reported to it grew fivefold in the previous 60 days.
- Nearly 40% were less than 90 days old when first reported, and 85% were gone within two weeks.
- Scam sites are "already being surfaced" to shoppers who hand purchases to AI tools.
These are reports from Scamwise users, not a count of every fake store online. The direction is still clear. The fake online store and non-delivery entry in my Scam Atlas explains how these shops work and the red flags to check.
The Canadian Centre for Cyber Security put the core idea in one line in its online shopping guidance: "Treat AI-generated recommendations as guidance, not assurances of legitimacy." The seven rules below are built around that idea.
Rule 1: Connect a credit card, not a debit card
This is the single most important choice you make.
In the US, the Fair Credit Billing Act lets you dispute credit card charges for things "you didn't accept or weren't delivered as agreed," as the FTC explains. The rule behind it, Regulation Z section 1026.13, gives you 60 days from the first statement showing the error to send your dispute. If an agent buys the wrong size or the wrong model, that is a real path back to your money.
Debit cards fall under a different law, the Electronic Fund Transfer Act and its Regulation E. It protects you against unauthorized transfers. But its definition of "unauthorized" excludes transfers made by someone you gave access to, unless you told your bank that access was withdrawn. Handing an agent your debit card can count as giving it access.
Fraud limits differ too. Per the FTC, the most you might owe for unauthorized credit card charges is $50. For a debit card, it can reach $500 if you report late, and more after 60 days.
American Express has gone further than the law. Fortune reported in April that Amex pledged to protect cardmembers from erroneous transactions made by agents registered with Amex.
Do this: connect a credit card to any AI shopping tool. Do not connect a debit card or link your bank account directly.
Rule 2: Set a hard spending limit
Tell the agent your budget, but do not rely on that alone. A limit in your instructions is a request. A limit enforced by the payment method is a guarantee.
Do this:
- Set a per-purchase cap inside the shopping tool's settings, not just in your chat.
- For extra safety, use a virtual card number with a spending limit, if your card issuer offers one, and use it only for AI shopping.
- Set a total holiday budget and check it weekly.
Rule 3: Keep approval on, and actually read the cart
If your tool offers a "confirm before buying" setting, keep it on this season, even if the tool offers to handle everything for you. Where a tool buys first and offers a cancellation window, as Amazon's auto-buy does, watch for the order notification and check it the same day.
Approving is only useful if you read what you are approving. Before you tap "confirm," check five things:
- The exact item (model, size, color, quantity)
- The seller (is it the store you expected, or a third-party seller you have never heard of?)
- The total price, including shipping and tax
- The delivery date (will it arrive before you need it?)
- The return policy ("final sale" is a much bigger problem when you did not pick the item yourself)
Your approval is also your evidence. If something goes wrong later, a clear record of what you approved, and what you did not, makes a dispute far easier.
Rule 4: Tell the agent where to shop
The easiest way to keep an agent away from fake stores is to tell it which stores are allowed. Inc.'s report on Visa's warning lists preferred merchants among the guardrails shoppers should set, alongside spending limits and approval requirements.
Do this:
- Add a line to every request: "Only buy from [retailers you trust], or from the brand's official website."
- If the agent suggests a store you do not recognize, look it up yourself before approving. Check how long it has existed and search its name plus "scam" or "reviews."
- Be most suspicious of brand-new stores with deep discounts. Scamwise found an impossible discount in one in four scam stores, and those blend in during Cyber Week, when real retailers are discounting too.
Rule 5: Never let an agent pay outside normal checkout
A legitimate store will take a normal card payment through a normal checkout. Scammers want something you cannot reverse.
Do this:
- Never approve payment by gift card, wire transfer, cryptocurrency, Zelle, or bank transfer for a store purchase.
- Never paste your card number into a chat with an AI or a "customer service" agent. Use saved payment methods inside trusted apps.
- If an agent asks for your full card details, a one-time code, or your password, stop.
Banks see this risk too. In September, six banks, including Bank of America, Capital One, ING and NatWest, published shared principles for trusted agentic commerce. Fox Business reported that their paper flags agents "requesting customers' card details and entering them directly into websites" as one of the risks.
Rule 6: Don't assume the AI is telling you everything
AI shopping assistants are useful, but they are not neutral referees.
A study from Columbia Law School's Center for Law and the Economy, reported by Reuters in July, tested Amazon's Alexa for Shopping and Walmart's Sparky. It found the assistants "can often detect when 'Made in USA' product labels are false," yet those labels stayed in front of shoppers. The Alliance for American Manufacturing has since asked the FTC to open an inquiry into AI shopping assistants.
The broader point applies to every claim an agent repeats: a deal, a review score, a "limited stock" warning, or a product feature.
Do this:
- For any purchase over about $100, check the product page yourself or ask a second AI engine.
- Ask the agent directly: "Has this seller or product had complaints, recalls, or safety issues?"
- Treat "only 2 left!" and countdown timers as sales tactics, whether a human or an AI shows them to you.
Rule 7: Watch what happens after you buy
The scams do not stop at checkout. "Fake delivery texts, order confirmations, and refund scams spike after big shopping days," according to the Scamwise report. The Scam Atlas walks through the package delivery text scam and the fake refund scam step by step.
Do this:
- Track orders inside the retailer's own app or website, never through links in texts or emails.
- Check your card statement weekly through the holidays. The 60-day credit card dispute clock starts when the statement is sent.
- Save your order confirmations and your instructions to the agent for anything significant.
- Know how to revoke the agent's access to your payment method. If anything looks wrong, revoke it first and investigate second.
If something goes wrong
- Revoke the agent's payment access in the shopping tool, or ask your card issuer to freeze the card.
- Call your card issuer and explain what happened. Say clearly whether the problem is fraud (you did not authorize it) or an error (the agent bought the wrong thing).
- Contact the merchant, if it is a real store, and request a return or refund.
- Report scams. In the US, report to the FTC at ReportFraud.ftc.gov. In Canada, report to the Canadian Anti-Fraud Centre.
- Secure the account. Change the password for the shopping tool and any account that shared it, and turn on passkeys or multi-factor authentication. My one-weekend security checklist covers the order to do it in.
A note for families
If you are helping older parents or teenagers use AI shopping tools this season, set things up for them: a credit card with a low limit, approval required for every purchase, and a short list of approved stores. Then show them what a fake delivery text looks like. In my experience, ten minutes of setup prevents most of the problems in this article.
Older relatives are worth the extra care. People aged 60 and over reported $7.7 billion in losses to the FBI in 2025, as I covered in the scams that actually drained Americans.
The bottom line
AI shopping agents can save you hours this holiday season, and most purchases will go fine. But the agent is only as safe as the guardrails you give it.
I founded LoginRadius in 2013 and scaled it to serve more than a billion users, so I think about this as an identity question: who is allowed to act for you, and within what limits. Use a credit card. Set a real limit. Keep approval on and read the cart. Tell the agent where to shop. Never pay outside normal checkout. Double-check claims that matter. Watch your statements.
Do those seven things and you get the convenience without handing scammers a shortcut to your wallet.
FAQ
Is it safe to let an AI agent do my holiday shopping?
It can be, with guardrails. Connect a credit card rather than a debit card, set a hard spending limit in the tool's settings, keep approval on for every purchase, and tell the agent which stores it may use. Never approve payment by gift card, wire or crypto, verify important claims yourself, and check your statement weekly.
Can AI shopping agents be fooled by fake stores?
Yes. Visa has warned that agents can be deceived by counterfeit merchants built to exploit them, and repeated the warning in October 2026. Scamwise users reported a new fake store about every three minutes in its October report. Telling your agent to buy only from retailers you trust is the simplest defense.
Should I use a credit card or debit card with an AI shopping agent?
A credit card. In the US, the Fair Credit Billing Act lets you dispute credit card charges for goods you did not accept or that were not delivered as agreed, within 60 days of the statement. Debit cards fall under Regulation E, which has no equivalent right and treats transfers by someone you gave access to as authorized.
How much AI shopping is expected this holiday season?
Adobe forecasts AI-driven traffic to US retail sites will rise 130% year over year between November 1 and December 31, 2026, and 159% on Thanksgiving. It expects total US online holiday spending to reach a record $275.1 billion.
What should I do if my AI agent bought from a scam store?
Revoke the agent's access to your card, call your card issuer and say it is a scam purchase, and report the store to the FTC at ReportFraud.ftc.gov (or the Canadian Anti-Fraud Centre in Canada). Then change your passwords and turn on passkeys or multi-factor authentication for the shopping account.
More like this
All Scams & Cybersecurity- Scams & Cybersecurity$20.9 Billion and It's Not Phishing: The Scams That Actually Drained Americans in 2025Americans reported $20.9 billion in scam losses to the FBI in 2025, and phishing was only about 1% of it. Investment and trust scams…
- Scams & CybersecurityHow to Balance UX and Online Shopping Security: A Practical Ecommerce GuidePut strong controls where ecommerce risk concentrates and keep them invisible to real shoppers: payment-page script rules, passkeys, bot…
- Scams & CybersecurityTravel Security Checklist: A Practical 2026 GuideA founder's practical travel security checklist for 2026: realistic threats, what to actually do before, during, and after a trip, and…
Get new Scams & Cybersecurity writing
Enjoyed this? Subscribe and tell us what you read most. Scams & Cybersecurity is already ticked for you. No tracking pixels, unsubscribe with one click.