Testing Username Search: Method and Public Case
A transparent protocol for evaluating username checks with authorized accounts, negative controls, manual verification and honest accuracy metrics.
By AliaScan · Updated · 3 min read
What this methodology establishes
A useful username-search evaluation measures account detection separately from ownership attribution. Define a test unit as one exact username on one platform, at one recorded time. A report with many links is not automatically more accurate than a shorter one.
This article is a testing protocol, not a completed product benchmark. No accuracy percentage, speed ranking or competitor superiority is claimed here. A scored benchmark requires a documented, authorized account set and observed results.
Documentation findings across four platforms
The official documentation reviewed on 17 October 2026 shows why one shared interpretation does not fit all platforms. Reddit states that finalized usernames cannot change and deleted names do not become available again. TikTok describes a 30-day change interval and a temporary restriction on reusing a released name. Discord distinguishes a unique username from display names and server nicknames. Meta explains that synchronized username changes can also change an Instagram profile URL. These are documented rules, not observations from live account testing.
The practical implication is specific: preserve the current profile link and the date, identify the correct kind of name, and avoid inferring account ownership or deletion from a name alone. This four-platform documentation review does not measure the availability of every account or the accuracy of any search product.
Build a test set before running any tool
Use accounts you own or have permission to test. Confirm known-positive profiles manually. For negative controls, choose test names and verify their absence at the time of testing; random-looking strings are not guaranteed to be unused. Include cases with changed names and access restrictions.
Record the platform, input, expected observation, verification timestamp, account authorization and any login requirement. Keep consent records private and publish only information authorized for publication.
Run comparable checks and classify every outcome
Record tool version or service date, supported platforms and test settings. Run tools in a comparable window without bypassing access restrictions. Preserve raw results before manual review. Classify true positive, false positive, true negative, false negative and inconclusive separately.
Report precision as true positives divided by all positive predictions, and recall as true positives divided by known positives on eligible test units. Publish numerator and denominator alongside percentages. Report inconclusive checks and unsupported platforms separately rather than silently removing them.
Make results reproducible without overstating them
Publish a redacted test manifest, collection window, scoring rules, per-platform observations and known limitations. Link to source documentation and describe any missing evidence. Use separate metrics for account existence and ownership; do not invent a combined identity-confidence score.
For timing, report the number of runs and observed distribution, not an unqualified speed promise. Repeat only after meaningful platform or tool changes, preserving older results so readers can see what changed. No test accounts are included with this protocol.
Observed case: the same public X URL returned different access signals
On 17 October 2026, we checked the user-supplied public URL https://x.com/elonmusk. A web-retrieval tool returned HTTP 403. A separate unauthenticated HTTP GET using Mozilla/5.0 returned HTTP 200 at the same URL, and the returned HTML mentioned the handle. These are two access observations of one URL, not two independently verified accounts.
The classification remains inconclusive: a status code and a handle appearing in HTML are insufficient to verify usable profile content or ownership. We did not run an AliaScan product scan in this observation. No precision, recall, coverage or speed score is calculated. The linked observation record includes the collection time and a command to repeat the HTTP check; results can change by time and environment.
Sources and editorial policy
Published by AliaScan. Illustrative examples are not live findings. Submit corrections through our complaints page.
Submit a correctionPut this into practice — run a free username scan across 3,000+ platforms right now.
Search a username →Frequently asked questions
Is this an AliaScan accuracy benchmark?+
No. It is a reproducible evaluation protocol. A completed benchmark needs authorized test accounts, raw observations and independently checked labels.
Should blocked checks count as accounts that do not exist?+
No. Record them as inconclusive and disclose their share of the tested sample.