Blog
How chips and tokens protect your card number in different ways
Magnetic stripes, chips and wallet tokens each stop a different kind of fraud and miss others. What each does, the 2015 liability shift and why online differs.
People tend to talk about card security as though one technology guards the card number: "the chip," "tap to pay," "Apple Pay." In practice, three different protections are at work, they guard against three different threats, and none of them covers what the others do. A chip does not make an online purchase safer. A token does not make a stolen physical card harmless. And neither prevents the largest category of loss, which is a person being talked into approving a payment themselves.
This explainer takes the protections in the order they arrived, says what each one does mechanically, and sets out what each cannot do. The aim is to leave you able to answer a practical question: when something goes wrong, which layer failed or never applied?
The problem: a card number that never changes
A card number is, at heart, a long password that you hand over every time you pay and that never changes until the card is replaced. Anyone who gets a copy of it can try to use it. The history of card security is a series of ways to make a copy less useful.
There are three places a copy can be made. A thief can capture the data in person, at a reader, a terminal or a skimmer, and make a counterfeit card. A thief can capture it in transit or in storage, by breaking into a merchant's system and collecting numbers. Or a thief can take it from you directly, by tricking you into typing it, reading it out or approving a payment. Each protection below tackles one or two of those places.
Layer one: the magnetic stripe
A magnetic stripe holds your card number, expiry date and a few other fixed fields in a form that a reader can copy exactly. It was a good design for an era of paper receipts and phone checks. Its weakness is its fixed content. Whatever the reader reads is everything needed to make another card with the same stripe. A counterfeit card made from copied stripe data works at any terminal that accepts a swipe, with no way for the terminal or the issuer to tell the copy from the original, except by patterns in the spending.
That is the weakness chips were designed to remove.
Layer two: the chip
A chip, called an EMV chip after the companies that wrote the standard, is a tiny computer in the card. It does something a stripe cannot: it creates a different code for every transaction. When you insert the card into a reader, the chip and the terminal exchange messages, and the chip produces a one-time cryptogram that depends on the transaction details and on a secret key stored in the chip that never leaves it. The issuer, which holds the matching key, can verify that the cryptogram came from the real chip for this transaction.
The result is that copying the numbers off a chip card is not enough. A thief who records one transaction's data cannot replay it, because the cryptogram is good for that transaction only, and cannot extract the key to make a new one. The chip makes counterfeit card-present fraud hard, which is the very thing the stripe made easy. According to Visa (a secondary source, reported in industry material), chip technology cuts counterfeit card fraud by more than 80 percent compared with the magnetic stripe. That is a network's claim about its own system, and the exact figure depends on how it is measured. As a purely illustrative calculation, if a merchant had 1,000 counterfeit-fraud incidents a year with stripe cards and the claim held, the count would fall to about 200 (1,000 × (1 − 0.80)).
What the chip does not do is hide the card number. The number is still printed on the card, still stored by the issuer and still used when you pay online or by phone, where there is no chip to read. A chip protects the act of paying in person. It does nothing for the number once it leaves that setting.
Layer three: the liability shift
Technology only spreads if someone is paid to adopt it. In October 2015, for most U.S. merchants, the card networks moved liability for counterfeit card-present fraud onto whichever party in the transaction had not adopted chip technology. Automated fuel dispensers were given extra time: the deadline was pushed from October 2017 to October 2020 and then to April 17, 2021.
In general terms, that rule works like this.
- If the issuer gave you a card with only a stripe, and the merchant's terminal could read a chip, the issuer bears the loss from a counterfeit.
- If the issuer gave you a chip card, and the merchant did not have a chip reader, or processed it as a swipe, the merchant bears the loss.
- If both used the chip, the chip's protections apply, and the counterfeit case rarely arises.
This is a rule among the banks and merchants in the network's system. It does not decide what you owe. As a cardholder, federal law caps liability for unauthorized credit card use at $50, and if only the account number was stolen, not the physical card, you generally owe nothing. Many issuers and networks also advertise zero-liability policies that go further. The point of the liability shift was to make merchants and issuers pay for the cost of not upgrading, and it did that without moving the cost to you.
Layer four: contactless, the chip by radio
Tap-to-pay cards use the same chip and the same one-time cryptogram. The difference is only in the delivery: a short-range radio link instead of contact pins. A tap works within a few centimeters, which limits casual reading, and it generates a new cryptogram each time. contactless payment and a chip insertion offer the same cryptographic protection. A commonly repeated worry is that a stranger can "scan" a card in a crowd. The sources behind this site do not support that scenario. The reported attacks involve malware: a victim is tricked into installing an app on an Android phone, which harvests card data and relays it to a criminal's device that replays it at a terminal, an attack called "ghost tap." One security firm tracked at least $355,000 in illicit transactions from one point-of-sale vendor between November 2024 and August 2025, which says the attack is real and a long way from a pickpocket with a reader. can someone steal from your card by tapping it through your pocket? goes through the myth in more detail.
Layer five: the token
A token is the newest and most flexible tool. It replaces your real card number with a different number, issued for a specific purpose, that is useless anywhere else. The process is called tokenization. In a mobile wallet such as Apple Pay or Google Pay, the merchant receives a device-specific token, not your real number, and a one-time code is generated for each payment. The real number stays in the wallet's secure setup and with the issuer.
What does a token buy you?
- If the merchant is breached, the thief gets a token that works only with your device, or only with that merchant, and cannot be used elsewhere. The leak is far less valuable.
- If your phone is lost, the token and the card are protected by the phone's lock, and the issuer can turn off the token without replacing your whole card.
- If you use the wallet online, the checkout has no real number to type.
A virtual card number works on the same idea with a different mechanism: the issuer or a service gives you a separate number for a merchant or for a single use, linked to your account but not the same as your physical card's. The CFPB describes virtual numbers in plain terms on its site.
What a token does not do is protect a physical card that is stolen. The tokenized wallet is a different credential from the plastic card, and a lost plastic card still needs to be reported. It also does not protect you from approving a payment you were talked into. how a virtual card number works, and when it beats your real number and what actually happens when you tap your phone to pay in a store go into the practical side.
Why online and in-store are protected differently
The chip needs a reader. A website has no reader. When you buy online or by phone, the merchant gets your number, expiry date and usually a three-or-four-digit security code, the CVV, that is printed on the card but not stored in the stripe. That is what a "card-not-present" transaction is, and card-not-present fraud is the category of fraud that grew as chips cut counterfeit fraud in stores.
Online defenses are therefore different:
- The security code shows that whoever is paying has, or has seen, the physical card, but a thief with a breached database may have it too if the merchant stored it against the rules.
- An extra check such as 3-D Secure adds a step, such as a code sent to your phone or an approval in your bank's app, for some purchases.
- A token for stored cards. Many merchants that keep your card on file can store a token in place of the number. Whether a given merchant does is not visible to you.
- Fraud models at the issuer, which watch for patterns that do not fit you.
- Wallet checkout and virtual numbers, which swap the real number out altogether.
The practical upshot is that the same card number is much better protected at a terminal than on a website, and that using a wallet or virtual number online brings some of the in-store protection to the online setting.
Three short scenarios
These are hypothetical, and they show which layer matters.
Mina taps her phone at a café, and the café's payment system is later breached. The system held the token her wallet sent, not her real number. The thief has a string that works only with her device and that merchant's channel, so it is nearly worthless. Mina may never hear about it. Her real number was not in the breach.
Luis types his card into a small online shop that stores it and is later hacked. Luis typed the real number, expiry and code. The chip never entered the picture, and there was no token. His issuer may spot the use elsewhere, and his liability is limited by law, but he will need a replacement card and should update every account that uses it. Had he used a virtual number for that shop, he could have turned the number off without touching the rest.
Ana gets a call from "her bank's fraud team" and reads out a passcode. Every technical layer worked, because Ana approved the step. The protections that apply here are different ones: calling the number on the back of her card, and reporting quickly. A payment you authorize under pressure is harder to reverse than an unauthorized one.
What each protection stops and misses
| Protection | What it stops | What it cannot stop |
|---|---|---|
| Magnetic stripe | Nothing beyond the issuer's pattern checks | Copying and counterfeiting at a reader |
| Chip (contact) | Counterfeit cards made from copied data; replaying a captured transaction | Online use of the number; a lost or stolen physical card; scams |
| Contactless chip | The same as the chip, with the same cryptogram | The same; plus malware-based relay attacks that start with a tricked victim |
| Wallet token | Use of a leaked merchant number elsewhere; typing the real number online | A scam in which you approve the payment; a compromised phone or account |
| Virtual card number | A leaked number being reused at another merchant | Fraud by the merchant it was issued to; the cost of a mistake you make |
| Security code and 3-D Secure | Casual use of a number copied from a list | A thief who has your phone, or a code you give away |
What none of these stops
Three kinds of loss slip past every layer.
Authorized payments you were tricked into. If a caller says they are from your bank and persuades you to read out a one-time passcode or to approve a payment, the transaction is valid from the network's side. The chip, the token and the cryptogram all work as designed, because you did the authorizing. why you should never read a one-time passcode to someone who called you explains why, and scam payment or unauthorized charge: why your protections can differ explains how the protections differ for such payments.
Account takeover. If a thief gets into your online account, they can add a new device, reset a login or request a new card. Tokens and chips guard the card, not the account behind it. Strong logins and account alerts do.
Data you gave away. Phishing pages that ask for the number, the expiry and the code collect everything a card-not-present purchase needs.
When you see an unfamiliar charge, call the number on the back of your card promptly, and if the issuer does not resolve it, use the Consumer Financial Protection Bureau complaint process. For billing errors, written notice within 60 days after the statement showing the error was sent preserves the statutory rights. The lost or stolen card checklist lists the steps in order.
Choosing the protection for the situation
Match the layer to the place.
- In a shop: insert or tap. Use the chip or tap, not a swipe, when you have the choice.
- At an unfamiliar online store: prefer a wallet checkout or a virtual number if the card offers one, so the shop never holds your real number.
- For subscriptions and trials: a virtual number set for one merchant can be turned off without replacing the card.
- When the card is lost: report it. The token on your phone and the plastic card are separate.
- On any call or text asking for a code: hang up and call the number on the back of your card.
Key takeaways
- A stripe is a fixed copy of your data, a chip makes a one-time code per transaction, and a token substitutes a number that is useless elsewhere. Each solves a different problem.
- The chip protects the act of paying in person against counterfeit cards, and Visa says it cuts counterfeit fraud by more than 80 percent versus the stripe. It does not protect the number online.
- Since October 2015 for most U.S. merchants, counterfeit card-present fraud liability sits with whichever of the merchant or issuer did not use chip. Your own liability is set by law ($50 cap) and by zero-liability policies.
- Tokens and virtual numbers shrink the damage when a merchant is breached. They do not help if you approve a payment yourself.
- No layer stops a scam you authorize. Call the number on the back of your card, and use the CFPB complaint process if the issuer cannot resolve it.