Most of us guard far more doors than we can count: a NordPass study puts the average tally at around 100 passwords, and seven of the top 11 most-used were variations of "123456", crackable in under a second. A password manager promises a way out: one strong master password, and it generates, stores, and fills in a different long, random password for every site.
One honesty note: this is a research-based evaluation. No password manager was personally tested for this article; the verdict rests on published guidance, studies, and breach records, all linked below.
Source: ConsumerNotice
At its core, a password manager is an encrypted vault holding a unique password for every account, unlocked by one master password. Good ones generate those passwords — long, random strings you would never memorise — and fill them in automatically. The vault is encrypted, typically with AES-256, under a "zero-knowledge" design: the company cannot read it, because the decryption key is derived from your master password and never leaves your device.
Source: ConsumerNotice
The strongest argument is not about encryption — it is about human behaviour. Nobody can memorise 100 unique, random passwords, so without a manager people reuse them, simplify them, or write them down. Reuse is the real killer: when one site leaks your password, every account sharing it becomes vulnerable to credential-stuffing attacks, where attackers simply try the leaked pair everywhere. A manager breaks that chain — every account gets its own password, so one breach stays one breach. Official guidance agrees: CISA recommends using a password manager, and its advice is that strong passwords are long (at least 16 characters), random, and unique to one account. NIST's guidelines likewise note that managers increase the likelihood people choose stronger passwords, particularly with a built-in generator.
Sources: Fordham IT Security, MorphPass (NIST/CISA guidance summary)
The honest objection is concentration of risk: every key sits in one vault, and the master password becomes a single point of failure. Lose it under a zero-knowledge design and the company cannot reset it for you; choose a weak one and the encryption is only as strong as that password.
That is not theoretical. In 2022, attackers broke into LastPass's systems and copied backup data including customers' encrypted password vaults. Those vaults could only be decrypted with each user's master password, leaving users with weak ones exposed to offline cracking, where attackers guess relentlessly without tripping any lockout. The fallout continues: in November 2025 the UK Information Commissioner's Office fined LastPass £1.2 million over failures affecting up to 1.6 million UK users, and LastPass settled a related class action in 2025 for $24.5 million.
A second limit deserves plain statement: a password manager does not stop phishing by itself. If you willingly type your master password into a fake login page, no vault design saves you. Autofill helps — it refuses to fill credentials on a lookalike domain — but only if you let it do the filling instead of typing. The same caution applies to any tool holding your secrets, as the week AI agents went rogue showed: the more a tool can do on your behalf, the more its security model matters.
Sources: Wikipedia (2022 LastPass data breach), BleepingComputer, The Register
Why did every site once demand an uppercase letter, a number, a symbol, and a fresh password every 90 days? That traces to one appendix in a 2003 NIST document by Bill Burr, then a manager at the institute. In a 2017 Wall Street Journal interview, Burr said of that advice: "Much of what I did I now regret." He had leaned on a 1980s white paper because real password data barely existed in 2003, and the rules backfired — people meet complexity requirements in predictable ways ("Password1!" becomes "Password2!"). NIST rewrote the guidance in 2017: favour length over complexity, and stop forcing periodic changes absent evidence of compromise.
Source: ZDNet
If the case above convinces you, judge a manager on four things, not its marketing page: genuine zero-knowledge architecture with established encryption (AES-256); independent security audits and a record of fixing what they find; multi-factor authentication for the vault itself; and an export option, so you can leave if you stop trusting the company.
On cloud versus local, both can be done well: a cloud vault syncs across your devices, a local vault keeps everything on hardware you control — the same trade-off running through the open-weights war. Either way, passkey support is worth having: passkeys replace a typed secret with a device-bound key, so there is nothing to phish.
Sources: Experian, Fordham IT Security
Yes, for most people — and this verdict is research-based, not hands-on: no product was tested for this article. The expert consensus, from CISA to NIST's guidelines, is that managers beat the realistic alternative, which is not perfect memory but password reuse. The LastPass breach is the correct counterweight, not a refutation: a vault is only as strong as its master password and its vendor's engineering, which is why zero-knowledge design, audits, vault multi-factor authentication, and an export option matter more than brand names.
Use one — unless you genuinely run a small number of accounts, all behind passkeys and a second factor, with unique passwords you actually remember. Give it a long master passphrase (four or more random, unrelated words, written down somewhere physically safe), turn on multi-factor authentication for the vault, and switch on passkeys wherever offered. That combination — a manager for the passwords you still need, passkeys where you can drop them — is the most defensible setup the current evidence supports, as our reality check on AI agent browsers also concluded: trust the architecture and the evidence, not the demo.
More in this category: Beyond the Hype