Evidence Building
How to Document Research Software Contributions as O-1A Evidence in 2026
Research software is a primary scientific output for many O-1A candidates, yet USCIS has no established template for evaluating it. This guide explains how to frame citations, adoption metrics, and expert testimony so they satisfy the original contributions criterion.
Research software in the O-1A evidence framework
Research software — computational tools, simulation frameworks, data pipelines, and analysis libraries developed by researchers — has become a central output of modern scientific work. For researchers in computational biology, climate modeling, machine learning, and astrophysics, a widely adopted software package may represent a more significant contribution to the field than any individual paper. Despite this, USCIS has historically treated research software as an ambiguous form of evidence — neither as straightforwardly legible as peer-reviewed publications nor as definitively excluded as purely commercial products. Understanding how to present software contributions so they satisfy the O-1A criteria is one of the more technically specific evidence challenges in contemporary O-1A practice.
The O-1A classification under 8 C.F.R. § 214.2(o)(3)(ii) requires the petitioner to demonstrate extraordinary ability in science, education, business, or athletics by satisfying at least three of eight regulatory criteria. The criterion most directly applicable to research software is original contributions of major significance to the field, defined at 8 C.F.R. § 214.2(o)(3)(iii)(B) as contributions evidenced by published material in professional publications, major media, or other outlets, and through testimonial letters from experts. The statute and regulations do not enumerate specific evidence types beyond this general framework, which means software contributions can satisfy the criterion if presented with evidence that meets the implied standard: independent significance that has materially influenced the direction of the field.
The USCIS Policy Manual addresses the comparable evidence provision at 8 C.F.R. § 214.2(o)(3)(iii), which allows petitioners to submit evidence comparable to the listed criteria when the standard criteria are not readily applicable to the occupation. For researchers whose primary output is software rather than text-based publications, this provision can be used to argue that software citations, download metrics, and expert adoption letters are the functional equivalent of scholarly article citations and publication records. Using the provision effectively requires a careful argument about why conventional evidence types do not apply and why the proposed substitutes are genuinely comparable in terms of the recognition they represent within the field.
What the original contributions criterion requires
The original contributions criterion under 8 C.F.R. § 214.2(o)(3)(iii)(B) requires evidence that the petitioner has made original scientific, scholarly, or business-related contributions of major significance in the field. Two elements are embedded in this standard: originality, meaning the contribution is the petitioner's own work and not merely an implementation of prior work; and major significance, meaning the contribution has had a material impact on the field rather than being incremental or localized. For research software, originality is typically straightforward to demonstrate — the code was written by the petitioner and the methods it implements are novel. Major significance is the harder element and the one where USCIS scrutiny tends to concentrate.
USCIS officers assessing the major significance of a software contribution will look for objective evidence of adoption, impact, and peer recognition. The USCIS Policy Manual notes that letters attesting to the importance of a contribution, standing alone, are typically insufficient — the letters must be supported by objective indicators of impact. For research software, the most relevant objective indicators include: citation counts in peer-reviewed literature, which demonstrate that scientists have relied on the software in published work; GitHub stars, forks, and active contributor records, which demonstrate sustained community adoption; download or installation statistics from package repositories such as PyPI, CRAN, or conda-forge; and adoption by named research institutions, national laboratories, or funding agencies as part of their standard computational toolkit.
The criterion does not require that the petitioner be the sole developer of the software. Software that began as a petitioner's doctoral project and was subsequently maintained by a broader community still supports the original contributions claim if the petitioner can demonstrate that the core algorithmic or methodological innovations that make the software significant were their own work. Expert letters from researchers who have used the software and can speak specifically to which aspects were novel and consequential are essential for making this argument when the development history is complex. Generic letters that describe the software as useful without attributing specific innovations to the petitioner are not persuasive on the major significance element.
Evidence that routinely satisfies the criterion
The most persuasive evidence file for research software original contributions combines citation data, adoption metrics, and expert testimony. On the citation side, a software paper published in journals such as the Journal of Open Source Software, SoftwareX, the Journal of Statistical Software, or a field-specific outlet provides a citable reference point that accumulates a record in Google Scholar, Scopus, and Web of Science. A software paper with several hundred citations in peer-reviewed literature places the contribution within the range that adjudicators associate with significant work. If no software paper exists, a collection of papers that cite the package in their methods sections can serve a similar evidentiary function, though this requires more assembly work.
Adoption at named, prestigious institutions is a strong objective indicator of significance. A declaration from a research group at a major national laboratory — NCAR, NOAA's Geophysical Fluid Dynamics Laboratory, a DOE national lab, or a prominent university research center — stating that the petitioner's software is an integral part of their computational workflow is the kind of specific, institution-grounded testimony that USCIS finds credible. Similarly, adoption by a major scientific collaboration or inclusion in a federally funded software sustainability initiative demonstrates that peer scientists have evaluated the software and found it important enough to integrate into institutionally funded work. Adoption letters from organizations carry significantly more weight than letters from individual researchers who simply use the tool.
A GitHub repository with substantial engagement — thousands of stars, hundreds of active forks, and a commit history showing sustained development — supports the adoption argument when combined with other evidence. USCIS does not treat GitHub metrics as decisive on their own, but they corroborate the adoption claim when the expert letters and citation record are already strong. Package repository statistics — PyPI download counts, CRAN download data, conda-forge installation figures — similarly provide objective adoption evidence that complements the citation and expert letter record. For software that serves a large but non-academic user base, such as climate scientists at government agencies or computational chemists in pharmaceutical research, download statistics may be more representative of actual impact than academic citation counts.
Evidence USCIS typically discounts
Evidence commonly submitted for research software original contributions claims but that USCIS has shown a pattern of discounting includes: generic letters that describe the software as useful, popular, or time-saving without explaining what makes the underlying methods novel; GitHub star counts in isolation, without adoption letters or citations to contextualize what the stars represent; download statistics without any expert testimony establishing that the downloads reflect genuine professional use rather than casual curiosity; and informal acknowledgments in academic presentations or conference slides without accompanying evidence that the software influenced published research.
Awards specifically designed for research software — such as best software paper awards at domain conferences — can support an original contributions argument, but they must be analyzed carefully under both the original contributions criterion and the awards criterion at 8 C.F.R. § 214.2(o)(3)(iii)(A). A best paper award at a second-tier conference for a software paper will not, on its own, satisfy either criterion. By contrast, a software package adopted as a standard tool by an NSF-funded collaboration and cited in major grant applications to NSF or NIH as a methodological prerequisite can be presented as a contribution of major significance even if it has never won an award.
Testimonial letters from colleagues who work in the same research group or who are co-developers of the software are significantly less persuasive than letters from independent experts at distinct institutions. USCIS adjudicators understand that colleagues and collaborators have an inherent interest in supporting one another's petitions, and the Policy Manual's guidance on expert letters explicitly calls for letters from experts who can speak to the petitioner's reputation in the field rather than from those with a close working relationship. For research software evidence, the most credible letter writers are researchers at independent institutions who adopted the software after evaluating alternatives — not researchers who advised the petitioner's graduate training or who contributed code to the project.
Framing borderline software contributions
Research software adopted within a specific subdiscipline but not widely known outside it presents a common framing challenge. A tool used by every active researcher in a small but technically demanding field — computational fluid dynamics for hypersonic flows, or quantum chemistry methods for actinide systems — may have a relatively modest absolute citation count while still representing genuine major significance within the relevant scientific community. The effective framing is to establish the size and structure of the relevant field first, then demonstrate that adoption within that field is comprehensive, and then explain why the methods implemented in the software represent a non-obvious advance over the prior state of the field.
The comparable evidence provision at 8 C.F.R. § 214.2(o)(3)(iii) is particularly useful for researchers in emerging computational fields where the norms around software citation are still developing. A petitioner who can demonstrate that their field does not have a tradition of software papers — and therefore cannot offer a citation record in the traditional sense — but can show that the software is used in federally funded research, cited in grant applications, and endorsed by program officers at NSF or NIH, has assembled evidence that is genuinely comparable in significance even if it does not look like a standard publication record. This argument should be made explicitly in the petition cover letter, not left implicit.
Expert letters for borderline software contributions should be written by researchers who can speak specifically to the technical difficulty of the problems the software addresses, the prior state of available tools before the petitioner's work was released, and the specific ways in which the software changed what was computationally feasible. Generic affirmations of importance are not sufficient for borderline cases. The most effective letters identify specific publications or research projects that would not have been feasible without the petitioner's software, or describe how a major funding agency or scientific collaboration changed its standard methodology after the software became available.
Auditing your research software evidence file
Before submitting an O-1A petition that relies significantly on research software evidence, the petitioner and attorney should audit the evidence file against the major significance standard. The audit should answer four questions for each piece of software evidence: Does this evidence demonstrate that the software exists and the petitioner created it? Does it demonstrate that independent researchers — at separate institutions, without a close working relationship with the petitioner — have adopted and relied on it? Does it establish that the adoption reflects a considered professional judgment that the software advances the field? And does it quantify the scale of adoption in a way that places the contribution above the threshold of incremental or local significance?
The petition should not be submitted if the evidence file can only answer yes to the first question. A GitHub repository link and a download count, without expert letters or citation data, is not an evidence file — it is a demonstration that the software exists. The file should include at minimum: a software citation record or collection of citing papers; two to four expert letters from independent researchers at distinct institutions; adoption evidence from a named institution or federal program; and a declaration or documentation establishing the petitioner's specific role in the software's design and development. Each element addresses a different dimension of the originality and major significance standard.
The software evidence should be situated within the broader O-1A evidence file rather than presented as a standalone argument. Petitioners who rely heavily on software contributions typically have at least some additional evidence in other O-1A criteria categories — a peer-reviewed paper record, even if modest; a high salary relative to the field; judging or editorial service; or a critical role in a nationally funded research program. A petition that satisfies only the original contributions criterion through software evidence, without corroborating evidence across at least two other criteria, is exposed to an RFE on the totality-of-evidence question. The software evidence is strongest when it complements and reinforces other evidence rather than carrying the entire extraordinary ability argument alone.