Over the past several months, our O-1 practice has seen a pattern we had not seen before at this frequency: petitions in which USCIS concedes that the beneficiary satisfies three, four, sometimes five of the regulatory evidentiary criteria — and then denies the case anyway.

The denial notices and Requests for Evidence all use the same phrase. The beneficiary has not been shown to be at the “top of the field.”

Clients understandably read this as USCIS moving the goalposts. It is a fair reaction, but it is not quite what happened. The standard did not change. What changed is that USCIS is now enforcing a second step in the analysis that has always existed on paper and that, for years, most petitions were never seriously tested against. Three things happened in the last eighteen months to bring that second step to the front of every adjudication — and the numbers confirm the effect.

An O-1 petition is adjudicated in two steps, not one.

Step one is the gate. The petitioner shows either a major internationally recognized award, or evidence satisfying at least three enumerated regulatory criteria — eight of them for O-1A, six for O-1B (Arts), four for O-1B in motion picture and television. This step asks a narrow question: does the evidence formally fit inside a criterion?

Step two is the merits. Having cleared the gate, USCIS then weighs the record as a whole to decide whether the beneficiary actually possesses sustained national or international acclaim — whether they are, in the regulation’s words for O-1A, “one of the small percentage who have arisen to the very top of the field of endeavor.”

Satisfying three criteria gets you to step two. It does not get you through it. USCIS has said so in its Policy Manual for years, and the regulatory preamble said so in 1994: the evidence “is not the standard for the classification, but merely the mechanism to establish whether the standard has been met.”

What has changed is that officers are now performing step two explicitly, in writing, in case after case — and are far less willing to accept a well-organized checklist as a substitute for evidence of actual acclaim.

This is not a matter of impression. It shows up in USCIS’s own numbers.

O-classification denial rates (USCIS reports O-1 and O-2 together):

The denial rate in the first half of FY 2026 is 46% higher than the FY 2025 full-year rate and 62% higher than FY 2024. In the fourth quarter of FY 2025 alone, the denial rate ran 46% above the same quarter a year earlier.

Requests for Evidence moved in the same direction. The O-category RFE rate sat between 18.7% and 19.7% across FY 2025. It jumped to 25.8% in the first quarter of FY 2026 and stood at 23.9% in the second — roughly a third higher than the prior-year baseline.

Those are meaningful shifts, but they understate the story. The clearest signal comes from EB-1A, the immigrant classification that uses the same statutory “extraordinary ability” language, the same “risen to the very top of the field” regulatory test, and the same two-step framework:

EB-1A denial rates doubled in a single year. By the second quarter of FY 2026, fewer than 42% of EB-1A petitions were being approved, down from roughly 67% for FY 2025 as a whole.

The O-1 is not moving as violently — it never has, and the O-1B “distinction” standard is textually more forgiving than O-1A’s “very top of the field.” But it is moving in the same direction, on the same curve, driven by the same analytical step. EB-1A is the leading indicator. O-1 is following it.

Here is the part of this story that has not been widely reported, and it is the most consequential.

On January 28, 2026, the U.S. District Court for the District of Nebraska decided Mukherji v. Miller. The court held that USCIS’s two-step framework — specifically the “final merits determination” at step two — is unlawful, because the agency adopted it as a substantive rule without going through the notice-and-comment rulemaking the Administrative Procedure Act requires. The beneficiary, a journalist, had satisfied multiple regulatory criteria and been denied solely at step two. The court did not merely send the case back for reconsideration. It vacated the denial and remanded with instructions to approve the petition.

USCIS appealed to the Eighth Circuit. Then, on June 10, 2026, it withdrew that appeal.

The agency dropped the case rather than risk a published appellate decision that would have bound it across seven states. The result is that the framework survives everywhere except in that one petitioner’s file — and remains unreviewed at the circuit level.

Six days later, on June 16, 2026, the Administrative Appeals Office issued a decision expressly rejecting a Mukherji-based challenge in a different extraordinary-ability appeal. The AAO defended the two-step process as an “interpretive” rule requiring no rulemaking, argued that the district court had misread both the 1995 regulatory history and the 2010 guidance memorandum that followed Kazarian v. USCIS, and held that the petitioner had raised the argument too late in any event.

Read those two moves together and the practical consequence is clear. Having chosen to defend the framework through litigation posture rather than through rulemaking, USCIS now has every institutional reason for its officers to write the step-two analysis out explicitly and at length in every case — building a defensible administrative record each time. That is precisely why the RFEs and denials landing on our desks now march through “top of the field” criterion by criterion, in a way they simply did not two years ago.

On August 5, USCIS issued Policy Alert PA-2026-05, effective immediately and applied to pending cases as well as new filings. Four changes matter for O-1 petitioners:

That fourth item deserves particular attention. Expert opinion letters and recommendation letters carry the argument in most O-1 petitions — they are how petitioners establish original contributions of major significance under O-1A criterion five, and lead or critical roles under O-1B criterion three. A rule that relegates that evidence to third tier, layered on top of a body of case law that already discounts generalized praise, removes the very bridge most petitions have been using to get from “three criteria satisfied” to “top of the field.”

Note also what to expect next in the data: RFE rates that fall while denial rates stay elevated or climb. Fewer requests, more direct denials. The Q1-to-Q2 dip in FY 2026 RFE rates may already be the leading edge of that.

If you are considering an O-1 filing: the honest threshold conversation now has to happen before filing, not after an RFE. Meeting three criteria is no longer a reasonable predictor of approval. The question is whether the record, taken as a whole, would persuade a skeptical officer that this person stands among a small percentage at the top of their field — and whether the evidence to prove that exists or can be built before filing.

If you have an O-1 extension coming up: an approval granted in 2023 or 2024 is not a reliable guide to how the same record will be received now. Extensions are being adjudicated under the current posture, not the one that produced the original approval.

If you are building toward a filing: the evidence that moves step two is comparative and quantified. Citation rates and h-index measured against others at the same career stage. Journal impact factors relative to alternatives in the field. Selection rates for awards, fellowships, and grants. The size of the practitioner population against the number holding a comparable credential. Unsolicited invitations to speak. Independently documented reputation of the organizations involved. None of that is required by any single criterion — which is exactly why it carries weight at the stage where cases are now being lost.

And front-load everything. Under PA-2026-05, the petition you file may be the only submission USCIS ever reads.

None of this makes the O-1 unavailable. Approval rates remain above 91%, which is still far stronger than EB-1A, EB-2 NIW, or most other high-skilled routes. But the margin for a thinly built petition has narrowed considerably, and it narrowed fast. Petitions that would have been approved in 2024 on a competent criteria-by-criteria showing are being denied in 2026 on the totality.

The standard did not move. The enforcement of it did.