# Case studies: first-party AI visibility and data work

Source: https://lynceus.app/case-studies/

> What we found in the data. First-party analyses of AI answers and public records, with the numbers, the method, and where it was possible, the files.

---

Case studies

# What we found  
in the data.

Every study here is first-party work: our question sets, our pulls, our cleaning, our numbers. Where the source is public, the cleaned file and the scripts are downloadable so the result can be checked rather than believed. No client, competitor or vendor is named in any of them.

[

Public records · September 2026

## The $586 million salary, and what the H-1B file actually hides

A client asked what H-1B practice actually looks like across attorneys and firms, from public records only. Answering it meant cleaning 437,496 Department of Labor filings, and the textbook fix for bad wage data turned out to be the one that fails.

437,496

cases cleaned

236x

raw max overstated

32%

of top 50 wages are E-3

Read the study → Data and scripts available

](https://lynceus.app/case-studies/h1b-wage-data-what-the-file-hides/)[

AI visibility · September 2026

## One brand, four assistants, 42 answers

The same brand was named in 73% of one assistant's answers and 18% of another's, on the same questions in the same week. A single blended score would have reported 51 and been wrong for everyone.

55 pts

spread across assistants

17 pts

branded-prompt inflation

25%

named on recommendation

Read the study →

](https://lynceus.app/case-studies/ai-visibility-one-brand-four-assistants/)