/ai-search536 words

Google AI Overviews Serve the Same Password to Every Searcher

PCMag testing confirms Google AI Overviews serve one repeated password across accounts, devices and days — cached answers, not random generation, on a top query.

Don't Use Google AI Overviews to Generate Your Passwords - PCMag UK
Don't Use Google AI Overviews to Generate Your Passwords - PCMag UKPexels / pixabay
  • PCMag reproduced the same AI Overview password across multiple Google accounts, devices, Incognito mode and multiple days.
  • SEO consultant Natalia Witczyk says AI Overviews are cached for high-volume queries, serving identical output to many users.
  • PCMag's test returned a 16-character password; Witczyk reported a 15-character version with different character order.
  • Earlier this year, PCMag's Kim Key found all five Gemini-generated passwords followed the same letter-number-symbol pattern.
  • AI Overviews also surfaced password tools from LastPass, Norton and Random.org alongside the suggested code.

Google's AI Overviews (the AI-generated answer panels above search results) serve the same suggested password to multiple users searching "generate a password," PCMag UK testing has confirmed — a code that repeated across separate Google accounts, devices, Incognito mode and multiple days.

The finding, first flagged by SEO consultant Natalia Witczyk in a LinkedIn post, turns on how Google handles high-volume queries. "Google AI Overviews are cached (pre-computed) for common, high-volume queries to reduce server costs and improve load times," Witczyk wrote. That caching means the identical string can reach hundreds of searchers.

What did the testing find?

PCMag reproduced the behavior directly. Key data points from the tests:

  • A password appeared in AI Overviews when the tester was logged into a personal account; it did not appear from an enterprise account or when signed out.
  • When it did appear, the same password showed up across multiple Google accounts and multiple devices, in both normal searches and Incognito mode, and across different connectivity options.
  • The same password repeated across multiple days, raising the risk that a regular user could quietly reuse one credential across accounts.

The caching explanation is tool-and-source-level reasoning from an SEO practitioner, not a confirmed statement from Google; PCMag's contribution is the direct reproduction, which is confirmed testing.

How bad is it?

The core problem is silent reuse. A searcher sees what looks like a freshly generated random string and assumes it's unique. It isn't. Witczyk put it bluntly: "Imagine hundreds, if not thousands, of users worldwide thinking this has been randomly generated. Imagine 'k8#Pz!vL9$mQ2wX*' becoming the new 'password.'" That hypothetical string — echoing the all-too-real "password123" problem — illustrates the scale she's describing.

It's poor password hygiene (the practice of avoiding reused or predictable credentials), but not the worst result on record. PCMag's Kim Key tested Google Gemini earlier this year and found that all five of its "random" passwords followed the same pattern — letters, numbers and special characters in an identical order — which she called "a hacker's dream."

The AI Overview results aren't pattern-locked in the same way. PCMag received a 16-character password, versus the 15-character version Witczyk reported, and the character order didn't match. The Overviews also surfaced third-party password-generation tools — LastPass, Norton and Random.org — alongside the answer.

What should users do instead?

The practical takeaway is straightforward. Don't take a credential from a cached search result. Use a dedicated random password generator, and store credentials in a trusted password manager rather than typing AI-suggested strings into account sign-up forms.

For the search industry, the incident is a reminder that AI Overviews are built for speed on common queries, and caching is the mechanism that delivers it. Any query where the user expects a unique, per-request output — credentials, security codes, tokens — is a poor fit for a pre-computed answer.

Worth watching next: whether Google adjusts or suppresses AI Overview outputs for password-generation queries now that the repetition has been publicly documented, and whether similar cached-duplication behavior surfaces on other high-volume queries where users assume per-request randomness.

via Google News: AI Overviews (Source)

More from James Calloway

James Calloway

Show full bio

Market editor covering money, funding and market shifts around Search rankings and algorithm updates, generative engine optimization at SERP Journal.

231 articles