Short answer
No, not on their own. GitHub stars and package downloads can support the original contributions criterion, but USCIS asks whether your work is of major significance to the field, and raw counts do not show that. With the EB-1A denial rate at 52.5% in Q1 FY 2026, petitions need documented adoption by independent users.
- Neither the O-1A nor the EB-1A regulation mentions stars or downloads or sets any numeric threshold.
- Researchers identified 4.5 million suspected fake GitHub stars across 15,835 repositories between July 2019 and October 2024.
- npm counts every package file it serves, including mirrors and bots, so downloads are not users.
- Named production users, dependent projects, citations and independent expert letters are what turn metrics into evidence.
If you build software in the open, your GitHub profile may hold more evidence than your resume: stars, forks, package downloads and a long list of dependent projects. This guide is for engineers, founders and researchers who want to know how much of that counts toward the original contributions criterion for the O-1A and the EB-1A, and how the same work plays out in an EB-2 NIW.
It draws on the regulations, the USCIS Policy Manual, decisions of the Administrative Appeals Office (AAO), USCIS approval and denial data as analyzed by the National Foundation for American Policy and the American Immigration Council, and published research on how GitHub stars and package downloads behave. Figures checked October 2026.
Do GitHub stars count as original contributions on their own?
No. A star count shows attention, and the criterion asks for significance. The O-1A regulation at 8 CFR 214.2(o)(3)(iii)(B)(5) asks for evidence of original scientific, scholarly or business-related contributions of major significance in the field. The EB-1A version at 8 CFR 204.5(h)(3)(v) adds artistic and athletic contributions but uses the same core phrase.
Neither regulation mentions GitHub, stars, forks or downloads, and neither sets a numeric threshold. Any page that quotes a minimum star count is offering a rule of thumb, not a rule.
The two questions USCIS asks
The USCIS Policy Manual splits the criterion into two questions. First, has the person made original contributions in the field? Second, are those contributions of major significance to the field? A public repository with your commit history usually answers the first question well. It shows authorship, dates and originality. The second question is where most open source evidence falls short.
The Policy Manual is direct about this: evidence that work was funded, patented or published may show originality but will not necessarily establish, on its own, that the work is of major significance. What does work, in its examples, is research that provoked widespread commentary from others in the field, work that has been highly cited relative to others' work in that field, and patented technology that has attracted significant attention or commercialization. Translated to open source, that means adoption by others, measured against the field, and confirmed by people who are not you.
This criterion is also only one piece. An O-1A petition needs at least 3 of 8 criteria and an EB-1A petition at least 3 of 10, followed by a review of the whole record. Our guide to the Kazarian two-step explains that final review.
Why stars and download counts are weak evidence on their own
Stars and downloads are popularity signals, and the research on both shows why an officer should not treat them as proof of impact.
Stars mostly measure attention
In a survey of 791 developers published by Hudson Borges and Marco Tulio Valente, the most common reason to star a repository was to show appreciation (52.5%), followed by bookmarking (51.1%). Only 36.7% said they starred a project because they used it. At the same time, 73% of surveyed developers said they consider star counts before using or contributing to a project, which is exactly why stars are a popularity signal rather than a usage measure.
The chart shows the split: appreciation at 52.5% and bookmarking at 51.1% outrank actual use at 36.7%, and recommendations by others at 4.6%.
Stars can be bought
A study by researchers at Carnegie Mellon University, Socket and North Carolina State University analyzed GitHub activity from July 2019 to October 2024 and identified 4.5 million suspected fake stars tied to fake star campaigns on 15,835 repositories. At the July 2024 peak, those campaigns touched 15.84% of repositories with 50 or more stars that month. Most of the affected repositories were short lived malware or phishing projects. The study also found the boost from fake stars lasted less than two months.
Downloads count machines, not people
npm has explained that its download numbers are a count of HTTP 200 responses for package files, and are not the same as the number of users. Mirrors, analysis bots and build servers all add to the total. In May 2026, Tenable documented a package that reached more than 50,000 downloads in three days after attackers published more than 700 versions, with each version typically drawing 100 to 150 downloads from automated systems.
A star is often a bookmark and a download is an HTTP request. Neither is a user, and neither, on its own, is a contribution of major significance.
Scale makes context necessary
GitHub's Octoverse report counts more than 180 million developers, about 630 million repositories and more than 1.12 billion contributions to public and open source projects in the year to August 2025. At that scale, a number without a benchmark from the same ecosystem tells an officer very little. What it needs is a comparison: how your project ranks among similar projects, and who depends on it.
What turns open source metrics into evidence of major significance?
Metrics become evidence when they are tied to named, independent users and to a documented change in how the field works. The table below sets out how each common signal reads to an officer and what corroborates it.
| Signal | What it shows | What it does not show | How to corroborate |
|---|---|---|---|
| GitHub stars | Attention and visibility | Use, impact or who the users are | Pair with dependents and named users |
| Forks | Interest and possible reuse | Whether forks were changed or shipped | Identify notable forks and what they built |
| Package downloads | Distribution volume over time | Unique users; includes bots and mirrors | Dated trend data plus production users |
| Dependent projects | Other software relies on your code | Whether those projects matter | Name the significant dependents and their reach |
| Named production users | Organizations rely on the work | Nothing, if documented by the user | Letters, engineering posts, case studies |
| Citations and references | Others built on the work | Significance without a field benchmark | Compare to similar work in the field |
| Maintainer role | Trust from a project community | That the original work is yours | Commit history, design docs, maintainer letters |
Sources: USCIS Policy Manual, Volume 6, Part F, Chapter 2; npm, how download counts work; He et al., fake stars study, 2019 to 2024.
What the AAO decisions show
Two non-precedent AAO decisions show where software evidence tends to break down. In a December 2023 decision involving an AI and natural language processing engineer, the AAO dismissed the appeal in part because the record did not show the patents had been commercialized or utilized, or widely implemented throughout the field. In a February 2025 decision involving an enterprise software developer, the record lacked probative documentary evidence that his programs were being used by other members of the industry, and press releases about his vendor status did not show a significant achievement.
Neither decision turned on GitHub, but the lesson carries over directly: show use by others, with documents that come from those users. Detailed letters from independent experts are the other half, and the Policy Manual notes they are most persuasive when the record corroborates them. Our approach to expert letters and evidence is built around that pairing.
How hard is the bar right now? EB-1A data, FY 2022 to FY 2026
EB-1A denials have climbed sharply, so a petition that rests on thin open source metrics carries more risk than it did two years ago. The share of decided EB-1A petitions that were denied rose from 22.2% in FY 2022 to 33.1% in FY 2025, while denials grew 3.8 times over the same period.
| Fiscal year | Approvals | Denials | Denial rate | Approval rate |
|---|---|---|---|---|
| FY 2022 | 5,623 | 1,608 | 22.2% | 77.8% |
| FY 2023 | 10,356 | 4,156 | 28.6% | 71.4% |
| FY 2024 | 12,231 | 4,703 | 27.8% | 72.2% |
| FY 2025 | 12,468 | 6,165 | 33.1% | 66.9% |
Source: NFAP analysis of USCIS data, FY 2022 to FY 2025. Approval rate is Einstein Visa Experts analysis of the same USCIS data (approvals divided by approvals plus denials).
The quarterly data shows when the shift happened. The EB-1A denial rate held between 25.3% and 25.6% from FY 2024 Q4 through FY 2025 Q1, then rose to 27.3% in Q2, 33.5% in Q3 and 46.6% in Q4.
The American Immigration Council reports that the rate reached 52.5% in Q1 FY 2026, and that 21,157 EB-1A petitions were pending at the end of FY 2025. These are aggregate figures across all fields, not predictions for any one petition, but they explain why officers are reading original contributions evidence closely.
How does open source evidence work for the O-1A, EB-1A and EB-2 NIW?
The same repository can support all three, but each category asks a different question of it. The O-1A and EB-1A use it to prove a criterion and then acclaim; the EB-2 NIW uses it to show that your endeavor matters and that you can carry it forward.
| Factor | O-1A | EB-1A | EB-2 NIW |
|---|---|---|---|
| Where open source fits | Original contributions criterion | Original contributions criterion | Dhanasar prongs one and two |
| Threshold | 3 of 8 criteria | 3 of 10 criteria | No criteria list |
| Final review | Totality, sustained acclaim | Final merits, top of the field | Balance of factors, prong three |
| Premium processing | 15 business days | 15 business days | 45 business days |
| Premium fee | $2,965 | $2,965 | $2,965 |
Sources: 8 CFR 214.2(o); 8 CFR 204.5; 8 CFR 106.4(e); Federal Register, premium processing fee adjustment effective March 1, 2026.
For the O-1A, the bar is sustained national or international acclaim, and officers review the totality of the evidence. For the EB-1A, the final merits review asks whether you are among the small percentage at the very top of your field. For the EB-2 NIW, adoption data supports the national importance of your endeavor and shows you are well positioned to advance it, which is a different and often better fit for a widely used project led by someone earlier in their career. Our comparison of EB-1A vs EB-2 NIW covers the choice in detail.
Denial rates rose across all three categories during FY 2025, though not equally.
EB-1A denials went from 25.3% in Q1 to 46.6% in Q4, EB-2 NIW from 37.3% to 64.3%, and O petitions from 5.6% to 7.3%. Across FY 2025 as a whole, 6.1% of decided O petitions were denied, so O petitions as a group are still denied far less often than either green card category.
How to document open source impact, step by step
The goal is an exhibit an officer outside software can follow: what you built, who relies on it, and why that matters to the field.
- State the contribution in one sentence. Name the problem, what you built, and what changed for others because of it.
- Pull dated metrics from the source. Use registry and GitHub data, archive each figure with its date, and show the trend over time rather than a single snapshot.
- Name the independent users. Identify companies, projects and research groups that depend on the work, and collect documents from them: letters, engineering blog posts, documentation or case studies.
- Benchmark against the field. Compare your project with similar projects in the same ecosystem, so the officer can see where it ranks.
- Separate your work from the team's. Use commit history, design documents, release notes and maintainer letters to show which parts you originated.
- Secure independent expert letters. The most useful letters come from respected people who have not worked with you and can explain what the field did before your work existed.
- Connect it to other criteria. Coverage of the project, invited talks and invited review work can support published material, judging or scholarly authorship.
Illustrative profile
Illustrative profile: a backend engineer who created an open source data validation library now used across a programming ecosystem. The weak version of the exhibit is a screenshot of the star count and the weekly download figure. The strong version names companies that run the library in production, each confirmed by a letter or a public engineering post; lists significant projects that depend on it; shows the library referenced in later technical work; and includes letters from independent experts explaining what problem it solved.
How it changes by field
Engineers and tech professionals
For engineers, open source is often the most visible part of a career that otherwise lives behind a company firewall. The strongest petitions combine it with a critical role at a distinguished company, especially when internal work was later released publicly and adopted by others. Our page for engineers and tech professionals covers how this work maps to the other criteria as well.
Founders and startup executives
For founders of open source or open core companies, adoption of the project and adoption of the product reinforce each other. Customer contracts, investor diligence materials and press coverage can corroborate that the open source work matters commercially, which speaks directly to business-related contributions of major significance. See our page for founders and startup executives.
Researchers and scientists
For researchers, research software is often cited alongside the paper that introduced it. The Policy Manual's O-1A guidance names citation rates and a high h-index as relevant in the totality review, so citations of both the paper and the software belong in the record, compared against others in the same subfield. Reuse of the code by independent labs is strong evidence. See our page for researchers and scientists.
Mistakes that weaken open source evidence
Most weak open source exhibits fail for the same few reasons.
- Leading with the star count. It invites the question of what the stars mean, and research on fake stars makes that question easy to ask.
- Presenting downloads as users. Download totals include mirrors, bots and build servers. Call them downloads and pair them with named users.
- Using a single undated screenshot. Officers need dated, sourced figures and a trend, not one moment in time.
- Claiming team credit. Large projects have many contributors. If your specific part is not isolated, the contribution is not clearly yours.
- Relying only on collaborators' letters. Letters from people who worked with you are useful, but independent experts carry more weight on significance.
- Skipping the field benchmark. A number with no comparison gives an officer no way to judge whether it is high.
- Filing too early. If adoption is still building, mapping your record against each O-1 criterion first shows whether the evidence is ready or needs another round of documentation.
Frequently asked questions
How many GitHub stars do I need for an O-1A or EB-1A?
There is no number. Neither 8 CFR 214.2(o) nor 8 CFR 204.5(h) mentions stars, forks or downloads, and the USCIS Policy Manual sets no threshold. Officers ask whether the work is of major significance to the field. A modest star count backed by named production users and independent expert letters can outweigh a large count with nothing behind it.
Do forks or dependents count more than stars?
Dependents usually carry more weight, because each one is another project that relies on your code. Forks show interest and sometimes reuse, but many are never changed. Neither proves significance alone. The strongest exhibits identify specific dependent projects and organizations, explain why they matter in the field, and include documents from those users confirming how they rely on the work.
Can contributions to a large project I did not create count?
Yes, if you can isolate your part. Officers need to see which component, feature or design you originated, and what changed in the project and its user base because of it. Commit history, design documents, release notes that credit you, and letters from the project's maintainers describing your specific role are the usual way to separate your contribution from the team's.
Is open source evidence stronger for the EB-2 NIW than the EB-1A?
It plays a different role. The NIW has no list of criteria. Under Matter of Dhanasar, adoption data helps show that your endeavor has national importance and that you are well positioned to advance it. The EB-1A asks a harder question: whether the work is of major significance and part of sustained acclaim at the very top of the field.
Will USCIS assume my stars are fake?
Officers judge the record in front of them, and there is no rule that stars are presumed fake. But research documenting millions of suspected fake stars is public, which is one more reason not to rely on the raw count. Corroborating stars with dependents, named users and independent letters removes the question instead of leaving it open.
Can open source work satisfy criteria other than original contributions?
Often, yes. Coverage of a project in major trade or technical media can support the published material criterion if it is about you and your work. Invitations to review others' work for selective conferences or programs can support judging. A technical paper describing the project in a recognized venue can support scholarly authorship.
Talking it through
If your strongest work is open source, a consultation starts by mapping the repository, its users and its dependents against the criteria, and identifying which numbers are corroborated and which are not. We then outline the letters and user documents that would close the gaps, and which category, O-1A, EB-1A or EB-2 NIW, fits the record as it stands.
When the record is ready, we prepare the full petition, and it is reviewed, signed and filed by an independent licensed immigration attorney. To see where your open source work stands, book a consultation.
Sources
- eCFR, 8 CFR 214.2(o), O-1 classification (current)
- eCFR, 8 CFR 204.5(h), EB-1A extraordinary ability (current)
- eCFR, 8 CFR 106.4, premium processing timeframes (current)
- USCIS Policy Manual, Volume 6, Part F, Chapter 2, Extraordinary Ability (checked October 2026)
- USCIS Policy Manual, Volume 2, Part M, Chapter 4, O-1 Beneficiaries (checked October 2026)
- Federal Register, Adjustment to Premium Processing Fees (January 12, 2026)
- National Foundation for American Policy, Analysis of USCIS Denial Rates (April 2026)
- American Immigration Council, USCIS Filing Trends (August 2026)
- American Immigration Council, From Petition to Permanent Residence (August 2026)
- USCIS AAO, non-precedent EB-1A decision (December 12, 2023)
- USCIS AAO, non-precedent EB-1A decision (February 3, 2025)
- He et al., 4.5 Million (Suspected) Fake Stars in GitHub (arXiv, data July 2019 to October 2024)
- Borges and Valente, What's in a GitHub Star? (arXiv, 2018)
- npm blog, How npm download counts work (July 2014)
- Tenable, Download pumping on npm (May 2026)
- GitHub, Octoverse 2025 (September 2024 to August 2025)