ResearchPolicy

LLMs can link pseudonymous accounts to real identities at scale, study shows

Illustration for the LLM deanonymization study

Your anonymous account is probably not anonymous, and researchers have the numbers to prove it.

The study

A study titled “Large-scale online deanonymization with LLMs,” first posted to arXiv in February 2026 by Simon Lermen, Daniel Paleka, Joshua Swanson, Michael Aerni, Nicholas Carlini and Florian Tramer, built a pipeline that links pseudonymous accounts to real identities at scale.

The method has three steps. An LLM extracts identity-relevant features from someone’s ordinary posts. Semantic embeddings retrieve candidate matches. Then the model reasons over the top candidates to confirm the link and cut false positives.

Note what the input is: ordinary posts. Not leaked data, not metadata, just the things people write in public under a handle.

The results

Linking Hacker News profiles to LinkedIn accounts, the pipeline reached up to 68 percent recall at 90 percent precision. Non-LLM baselines scored near zero on the same task, which is the comparison that makes the result significant rather than incremental.

The researchers also tested Reddit movie discussions and temporal Reddit profiles, extending the approach beyond a single platform pairing.

What actually changed

The researchers’ conclusion is the part to remember: the practical obscurity that protected pseudonymous users online no longer holds.

What changed is cost. Deanonymising one person always took a determined human hours of reading, cross-referencing and inference. Nothing about that was impossible, it was simply expensive enough that it only happened to people someone specifically cared about. Now it runs at scale, against everyone, on ordinary comments.

That shift redraws the practical threat model for anyone who has posted for years under a handle. Using different usernames across platforms may now matter as much as not using your real name, because the linkage signal lives in writing style, topic patterns and biographical fragments scattered across posts, not in the name at the top.

It is worth asking how many years of posts are sitting under your old handle, and what a good matcher could assemble from them.

Sources

ANOTHER News is published by ANOTHER, an AI-native content agency. Daily coverage also runs on Instagram.