There is no threshold
Neither the regulation nor the USCIS Policy Manual sets a citation count, an h-index or any other bibliometric threshold for EB-1. Every specific figure circulating online — "you need 100 citations", "an h-index of 10" — is someone's extrapolation presented as a rule.
What citations do is evidence a claim: that the field engaged with your work. They are read alongside everything else, and their weight depends entirely on context that a bare number does not carry.
This cuts both ways. A researcher with a modest count in a small, slow-citing subfield is not disqualified. A researcher with a large count driven by self-citation and collaborators does not have the case the number suggests.
No citation count is required, and none is sufficient. The number is the start of an argument, not the end of one.
Independent citation is the figure that matters
The distinction that does most work is between independent citations and everything else. Citations from you, your co-authors, your supervisor and your own institution do not show that the field responded — they show that your immediate circle continued working.
Separating the two before filing is worth the effort, and it is straightforward: export the citing records, mark those sharing an author with the cited paper or affiliated with the same institution, and report the independent figure. Presenting that number, with the method stated, is more credible than a larger unqualified one.
Better still is showing what the citing papers did. A citation that says "we adopt the approach of [you]" is qualitatively different from one that lists you among fifteen references in an introduction, and quoting the former directly makes a point no count can.
Field norms are part of the evidence
Citation behaviour varies by an order of magnitude between disciplines. Median citation counts in mathematics and in molecular biology are not comparable, and a reviewer cannot be assumed to know your field's baseline.
Supplying that context turns a number into an argument: field or subject-category percentile data, median citation rates for comparable papers of the same age, and where available a defensible statement of where your work sits relative to that baseline.
One practical caution: databases disagree. Google Scholar, Scopus and Web of Science count different sources and will return different totals for the same paper. Pick the one appropriate to your field, name it, and use it consistently rather than quoting whichever is highest.