O-1 Visa for Artificial Intelligence (AI) and Machine Learning (ML) Researchers: O-1A Criteria and Evidence

For AI and Machine Learning researchers, the O-1A visa offers a way to work in the United States based on a record of extraordinary ability in the sciences. The visa is built for people whose work is recognized by others in the field, and much of what an active researcher already does can speak to that recognition.  

Eligibility is based on the applicant’s individual record rather than their job title, so two people in the same role can present very different cases. For an O-1 visa researcher, that record usually lives in published work, in contributions others build on, and in roles that call for real expertise.  

Can AI and Machine Learning Researchers Qualify for O-1A? 

There are two O-1 visa categories: the O-1A, for extraordinary ability in the sciences, education, business, or athletics, and the O-1B, for extraordinary ability or achievement in the arts. AI and Machine Learning researchers generally fall under the O-1A category. USCIS defines extraordinary ability for O-1A as being among the small percentage who have reached the very top of a field.    

Extraordinary ability can be demonstrated through a major, internationally recognized award, or, more commonly, through evidence satisfying at least three of the eight O-1A criteria. Meeting three criteria is the first step. USCIS also evaluates the evidence as a whole to determine whether the record establishes the required extraordinary ability and sustained recognition, as explained in its current O-1 Policy Manual guidance. 

O-1A provides temporary work authorization, granted for up to three years and renewed in one-year periods indefinitely as long as the work continues. A U.S. employer or agent files the petition. An applicant cannot self-petition, although a company the researcher owns can be the petitioner. Spouses and unmarried children under 21 can join the O-1 visa holder on O-3 status, which allows study but not work. 

For AI and Machine Learning researchers, commonly claimed criteria include published work, patents, high-salary evidence, and recognized contributions to the field, built into a record that shows sustained, field-wide recognition rather than a single achievement. 

Evidence AI and Machine Learning Researchers Commonly Use 

AI and Machine Learning careers often already generate the kind of evidence the O-1A looks for, so the task is matching what exists to the criteria it supports. For the full eight criteria and the evidence each one accepts, see the O-1 visa requirements guide.  

Below are the O-1 criteria most commonly claimed by AI and Machine Learning researchers explained: 

Authorship of scholarly articles 

This criterion looks at whether the person authored scholarly articles in the field. Papers published in peer-reviewed journals or presented at competitive AI and Machine Learning conferences, such as NeurIPS, ICML, and CVPR, can satisfy this criterion, particularly where the venue and subsequent use of the work reflect its standing in the field.  

A paper others build on often speaks to this criterion and to original contributions at the same time, whether or not the researcher is first or sole author. Citation counts and h-index figures are read in context. A recent paper will naturally have less time to gather references, so impact is not measure by citation count alone but by overall impact in the field. 

Original contributions of major significance 

This criterion examines whether the person made original scientific, scholarly, or business-related contributions of major significance to the field. In AI and Machine Learning, this often centers on models, methods, datasets, benchmarks, or systems that others adopt and build on.  

The existence of a published model or dataset does not by itself determine whether the underlying work amounts to an original contribution of major significance. Documentation such as citation records, downstream use, or independent recognition of the work’s impact may need to address that question, and USCIS then evaluates whether the proof satisfies the regulatory standard. 

A January 2025 USCIS policy update made this clearer for technical fields, treating contributions to shared repositories of software or data as examples when there is evidence of real impact. Patents are evaluated under the same distinction: a patent shows an idea is new, not that it mattered, so evidence of actual use strengthens the case. 

Judging the work of others 

This criterion considers whether the person participated as a judge of others’ work in the same or an allied field. AI and Machine Learning professionals may participate in activities that involve evaluating the work of others, including peer review for a scholarly publication, service on a conference program committee, or grant review. The regulation focuses on actual participation as a judge. An invitation that did not lead to participation does not suffice to meet this criterion. 

Awards or prizes 

This criterion examines whether the person received nationally or internationally recognized prizes or awards for excellence. Selective AI, research, or industry awards can support this criterion, though the selectivity and reputation behind the award matter as much as the award itself and must be documented as well. 

Membership in associations 

This criterion examines whether the person belongs to an association in the field that requires outstanding achievement, as judged by recognized experts, for membership. Selective fellowships, such as AAAI or IEEE Fellow status, can support this criterion when the selection process and the organization’s standing are documented, particularly when that recognition comes from outside the applicant’s own circle. 

Critical or essential roles 

This criterion considers whether the person served in a critical or essential capacity for an organization with a distinguished reputation. A senior or founding role at a respected lab or company is a common example, though the title alone does not establish this criterion. 

The analysis considers the person’s actual duties within the organization, and the organization must also have a distinguished reputation. Evidence concerning those two items serves different purposes. Proof about a lab or company’s reputation does not by itself establish that a particular researcher performed a critical role there, and documentation showing an important role does not by itself establish that the organization has a distinguished reputation. 

What Makes Evidence Persuasive 

Meeting the criteria on paper is only part of the picture. What tends to move a case is recognition from other people in the field, not the applicant’s own description of the work. A benchmark that other labs adopt, a speaking invitation from a conference the researcher does not run, or a citation rate that stands out among peers points to influence beyond a single team or employer.  

Context determines how that evidence is read. An award with few recipients or a citation count that is high for the subfield carries more weight than the same award or citation count stated on its own, without evidence of how it compares with others in the field , and because norms differ across areas of AI, the petition often has to explain what selective or highly cited means in that field. One accomplishment can support more than one criterion, but each still has to stand on its own for each of the criteria it aims to support.  

Common Weaknesses in AI and Machine Learning Researchers’ O-1A Cases 

A strong background does not guarantee a strong petition, and a few missteps recur in this field. Most share a root cause: evidence presented as a number or a title without the context that shows why it matters. There are a few patterns worth avoiding:  

• Counting papers or citations without showing how the work itself impacted the field 
• Listing peer review that was invited but never completed 
• Treating a patent as proof of  impact when it only establishes novelty
• Relying on recommendation letters without independent corroboration 

Building the Record and Next Steps 

For AI and Machine Learning researchers, an O-1A case comes down to recognition, impact, and how clearly the evidence fits the criteria. The record that supports it tends to grow out of ordinary research work, so keeping publications, review confirmations, and evidence of adoption organized makes a later petition far easier to assemble. When the record is ready, O-1 visa processing times can range widely, and premium processing may shorten the wait when it fits the circumstances, something an attorney can help determine.  

If you are considering an O-1A, a profile evaluation shows which parts of your record map most closely to the criteria and to the U.S. role you have in mind.  

See If You Qualify For The O-1

Complete our questionnaire to check your O-1 eligibility.

Evaluate your profile

Frequently Asked Questions 

Can AI and Machine Learning researchers qualify for an O-1 visa?  

Yes, in the right case. Qualifying depends on whether the record shows extraordinary ability under the O-1A standard, not on the applicant’s job title or field. 

What evidence is strongest for AI and Machine Learning researchers?  

There is no single strongest piece of evidence. Independent evidence usually carries the most weight, such as contributions others have adopted, peer review the researcher completed, and recognition decided by experts. 

How many O-1 criteria do I need to meet?  

Either one major internationally recognized award, or at least three of the eight O-1A criteria. Even after three or more criteria are met, USCIS reviews the whole record to decide whether it reflects sustained acclaim in the field. 

Does a high salary help an AI and Machine Learning researcher’s O-1 case?  

Potentially. High pay counts when it is clearly above what others in the same field earn and benchmarked to wage data. 

How many publications or citations do I need for O-1A?  

There is no set number. USCIS reads authorship and citations in context, weighing the subfield, how long the work has been out, and whether the work itself drew recognition from others.