A metered paywall that resets in an incognito window isn’t a paywall — it’s a suggestion. The meter lives in a cookie or in localStorage, the reader opens a private window, and the count starts at zero. The same happens after a cache clear, on a second browser, or when the browser expires the cookie on its own schedule. Each of those events is a free article.
This is not a bug you forgot to fix. Client-side storage is disposable by design: the browser belongs to the reader, and the reader decides what it keeps. You can chase the bypass with harder cookies, but the bypass stays one menu item away.
The better move is to change what you count. Instead of asking “does this browser still have a cookie?”, ask “has this device or this account already used its free articles?” Below: why the cookie meter fails, what to count instead, and how to return a paywall decision that keeps the limit honest without walling off readers who would have subscribed anyway.
Why a cookie meter resets
What a private window actually is
Incognito, Private Browsing, InPrivate — different names, same behavior: a separate, temporary storage partition. Cookies set inside the private window are dropped when it closes, and cookies from the normal window are not visible inside it. A meter that lives only in a cookie has no idea the reader has already read four articles. Clearing site data does the same thing on purpose, and an extension that wipes storage on close reproduces it without anyone thinking about your paywall.
Your browser is already expiring cookies for you
Modern browsers put their own limits on script-set cookies, cap their lifetime, and partition storage in ways that break naive counting. A meter built on “set a cookie for a year” fails quietly and unevenly, so your own QA rarely shows it: your colleagues’ browsers behave like each other, not like your readers’.
Count distinct articles, not page views
A paywall counter should answer one question: how many different articles has this reader consumed in the period? Page views are the wrong unit — a refresh shouldn’t cost an article, and neither should returning to a piece you already opened.

Per device: Prynt Identify returns a visitorId built from device and browser signals rather than a cookie the reader can delete. Your meter stores the article IDs already counted against that visitorId, so a private window, a cookie clear or another browser does not produce a fresh count. The mechanics are on our device fingerprinting page.
Per account: for logged-in readers the count belongs on the account, keyed by your own user ID. It survives device changes — the reader on a new laptop is the same reader — and it is the number that matters when they upgrade. The same device-to-account linking is what exposes one device cycling through several accounts, covered under account takeover.
Store article IDs, not titles or URLs, and hash the identifiers on your side: you need to know a reader used five of the allowance, not what they read. For a logged-in reader, take the higher of the device count and the account count.
Return a decision, not a number
Testing if (count > limit) block() inside a page template puts policy in a template. Prynt Control answers the paywall use case with allow, step-up or block, plus a reason code and a suggested step such as captcha, verify_email or card_or_passkey. Your server decides what to render; Prynt never blocks on its own. The shape below is illustrative, not a literal SDK call.
const decision = await prynt.decide({
useCase: "paywall",
visitorId,
linkedId: user?.id, // present when the reader is logged in
context: { articleId, window: "30d" }
});
Log the reason codes and watch how many readers finish the step-up: every step-up is a reader a hard block would have lost.
Keep it honest for readers you cannot count
A device counter is wrong for shared hardware — a family laptop, a library terminal, a coworking machine — and for readers who use privacy tools on purpose. Two guards keep the meter defensible.

- Treat device count and account count as evidence, not a verdict. Logged in, prefer the account. Anonymous, prefer the device and offer a step-up such as email verification, which turns an anonymous reader into an account instead of a dead end.
- Keep privacy signals out of the paywall decision. A VPN or a Tor exit node is a risk signal elsewhere; on a paywall it mostly punishes readers who care about privacy. Record it, don’t block on it.
Bots get their own line: crawlers don’t subscribe, so counting them against the allowance produces a graph and nothing else. Prynt Analytics leaves bot sessions out of every number, and Identify flags automation before the decision is made.
Run it in mirror mode first
Control starts with a 14-day mirror period in which every decision is recorded and nothing is blocked. Use it as a comparison, not a formality: put the decision you would have returned next to the count your cookie meter produced. The readers who move from “fresh count” to “over the limit” are the ones your current meter gives away. Fix the copy and the step-up choice before you enforce.
FAQ

Does incognito break a metered paywall?
Only if the meter lives in storage the private window doesn’t share. A meter keyed to a device identifier and to the reader’s account survives incognito, cookie clears and a switch to another browser.
What counts as one article?
One distinct article ID per reader per window. Refreshes, re-reads of the same piece and internal navigation shouldn’t move the counter; that is the difference between a meter and a page-view cap.
Is a device-based meter allowed under privacy law?
A device fingerprint can be personal data, and the notice you give readers matters. Prynt returns sealed results and supports hashed identifiers; how you disclose and document the practice is a question for your counsel.
Where the meter lives
The paywall meter is part of Prynt Control, the top Identify plan, and needs the Control plan — not Free, Pro or Scale. See how Prynt handles the paywall decision, or read how the pieces fit together on the platform page.
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.