All articles Advanced signals

Paywall Bypass Prevention: Stop Incognito Resets

Developer in a dark room facing a glowing screen with translucent floating browser panels linked by thin blue lines

A reader hits your limit, opens a private window, and keeps reading. Nothing about that is sophisticated: the private window starts with empty local storage. A browser update, a cleared cache or a switch to another device does exactly the same. If your meter lives in client storage, the reader owns the counter.

That is the real problem behind paywall bypass prevention. Cookies measure one browser profile on one device at one moment; they know nothing about the second browser, the incognito tab, or the account that just signed in elsewhere.

The alternative: count articles on two keys — device and account — and turn that count into a policy decision rather than an automatic wall. The mechanics below come from the paywall use case in Prynt Control; the design applies to any metered model.

A cookie or localStorage entry measures one browser profile. It is a decent identifier for a returning reader on the same device and a poor identifier for the device itself. When the profile disappears, the count disappears, and you have no record that the reader was ever there.

The resets you will see in real traffic

  • Incognito or private windows, which start with empty storage by design.
  • Cleared site data, browser updates and device migrations.
  • A second browser on the same machine, or the same person on a phone.
  • Shared logins, where one account serves several readers.

None of these needs a bypass tool. Readers find them by accident, which is why a cookie-only meter both misses resets and punishes people who simply changed devices.

Count on two keys: device and account

A smartphone and a translucent glass tablet connected by fine blue lines on a black background, illustrating one device key across form factors

The device key

A device key comes from browser and device signals — the foundation covered in our guide to device fingerprinting. Prynt Identify issues a visitorId per device, so a new browser profile on the same machine still resolves to a device you have seen before.

Treat that key as probabilistic. Browser releases, hardware changes and privacy tools move the signal, and a key that never changes is a key that will eventually be wrong. The goal is not to identify a person but to recognise a device across storage resets. VPN, proxy, Tor and datacenter signals are inputs here, not verdicts: a reader on a corporate VPN is not automatically a bypasser.

The account key

Once a reader logs in or shares an email, the account is the strongest key you have. Meter on the account id and device-hopping stops being a strategy. For anonymous readers the device key carries the count until an account appears; merge the history at that point, keeping the higher count so a login never resets the meter.

Turn the count into a decision: allow, step-up, block

A meter that can only block will eventually block someone who was about to subscribe. Prynt Control answers a paywall request with one of three outcomes and a reason code:

  • allow — serve the article and increment the meter.
  • step-up — ask for a small proof (captcha, email verification, card or passkey).
  • block — show the subscription wall.

Pseudo-code for the check:

const decision = await control.evaluate({
  useCase: "paywall",
  visitorId,   // device key
  accountId,   // null if anonymous
});

switch (decision.outcome) {
  case "block":
    return renderPaywall({ code: decision.reasonCode });
  case "step-up":
    return requireStep(decision.suggestedStep);
  default:
    return renderArticle();
}

Prynt never blocks on its own. It returns a decision and your code decides what the reader sees, so the final call stays inside your product.

Reason codes make enforcement debuggable

Log the outcome, the reason code and the request id with each render. When a reader says the wall is wrong, you can see whether the device key drifted, whether they are on a VPN, or whether the account already reached the limit.

Watch it before you enforce it

Control starts with a 14-day mirror period: decisions are evaluated and logged, and nothing is blocked. Use that window to read the distribution. How many readers would have been stepped up on their first visit? How many “new” devices are the same person on a phone? Which networks look risky only because an entire office shares them?

Three rounded translucent panels in white, blue and charcoal representing allow, step-up and block decisions in a dark space

The usual conclusion is a graduated policy: allow early visits, step up when the device count passes a threshold, and save blocking for the case that combines a high article count, a reset pattern and a risky network signal.

Handle the edge cases before they become tickets

Shared devices and privacy-conscious readers

Family tablets, library machines and work laptops on a VPN will produce device collisions. Route those readers to step-up rather than block, and write the policy down where readers can find it.

Shared credentials

One login serving several readers is a different pattern with the same revenue effect. Meter on the account and pair the paywall decision with the signals in our guide to account takeover: the device history shows whether one account is being used from ten places.

Bots and scrapers

Automated readers do not run your meter at all. They fetch the HTML and leave. Pair the paywall decision with bot detection so the article count applies to humans and the crawler problem is handled before the page renders.

FAQ

Analyst watching soft blue data glows in a dark monitoring room while reviewing paywall decisions before enforcing them

Does incognito mode defeat a device-based paywall meter?

No. Incognito clears client storage; it does not change the device signals Identify reads or the account id once someone logs in. Because the meter is kept per device and per account, a private window continues the existing count instead of starting a fresh one.

Is device fingerprinting acceptable for a paywall meter?

That depends on your jurisdiction, your legal basis and what you tell readers. Fingerprinting for security and metering is common, but it needs a documented purpose, a retention limit and a clear privacy notice. Treat it as a policy decision your legal team signs off, not a default.

What if a reader looks like a resetter but is not?

That is what the mirror period and step-up are for. Both give you evidence before you lose a reader, and both are reversible decisions.

Start in mirror mode

Pick one use case — paywall — and run it in mirror mode before you enforce anything. Then set thresholds from your own data rather than from a benchmark someone quoted online.

Prynt Control handles the paywall decision next to signup, free trial, login, checkout and OTP requests, on one account and one bill. Start at https://pryntid.com and see how the products fit together.

Try it free

Prynt is device intelligence with a free plan: 5,000 identifications a month, visitor IDs with confidence and bot detection — no credit card. Upgrade to Pro for VPN, proxy, tampering and behavioral signals.

Keep reading