The username is on Reddit, Flickr, Twitter, a photography forum, a cycling forum, and a niche local community board. At least two of those accounts are not your subject. You need to know which ones before you go any further.
The problem
Username searches routinely return more accounts than belong to the subject. Common handles — anything based on a real name, a profession, or a hobby combination — appear across dozens of platforms under independent ownership. The investigative problem is not finding the username. It is determining attribution: which instances of the username are the same person, which are coincidental name matches, and which are deliberate mimicry or squatting.
The standard approach is manual comparison: open each account, note the content, look for overlapping signals. That works at two or three accounts. At six, it becomes unreliable. An analyst working across six profiles is holding signals in working memory while switching contexts between tabs, platforms, and registration dates. The failure mode is not incompetence. It is cognitive load. Small inconsistencies that should raise flags get absorbed into a mental model that is already under pressure to close.
The specific gap this prompt addresses is the comparison step. Gathering the signals is the analyst's job. Holding all six signal sets simultaneously, assessing internal consistency, and returning a per-account confidence verdict with a stated evidence gap is where the model earns its place in the session.
