On this page

EB-2 NIW appeal outcomes

EB-2 NIW

What happens when an EB-2 or NIW denial is appealed

If USCIS denies an EB-2 or national interest waiver petition, the petitioner can appeal to the USCIS Administrative Appeals Office. This page counts how those appeals actually end, and the short answer is that they almost never win. Information, not legal advice.

Decisions analysed
4,987
published 2015 to 2026
Appeal dismissed
87%
3,633 of 4,196 decided on the merits
Sent back to be redecided
10%
428 remanded, which is not a win
Won outright
3.2%
135 sustained, about one in 31
What this actually means
Appealing almost never wins. Of the 4,196 decisions decided on the merits, 3,633 were dismissed and only 135 were sustained, about one in 31. A remand is the realistic good outcome and it is not a win: it means the service center has to decide the case again properly. So getting the petition right the first time matters far more than the appeal does, and refiling is often the better move than appealing.

Counted from 4,987 decisions the Administrative Appeals Office published between 2015 and 2026, read straight out of the decision PDFs. No language model is involved at any stage. How these numbers were produced.

General information about published decisions. Not legal advice, and not a prediction about any individual case.

EB-2 NIW

How appeals actually end

Every outcome the parser found, as a share of all 4,987 decisions.

OutcomeDecisionsShareRate
dismissed3,63372.8%
motion dismissed70814.2%
remanded4288.6%
sustained1352.7%
unparsed761.5%
motion granted60.1%
abandoned10.0%

Hover any outcome for what it means. A motion to reopen or reconsider is a different procedural posture from an appeal, and harder, which is why the headline rates above exclude them.

EB-2 NIW

Which prong of the test decided it

Prong 1 is where almost all of these petitions fail, and in a fifth of decisions the AAO never evaluates prongs 2 and 3 at all.

GroundDecisionsShareRate
Prong 1: substantial merit and national importance2,91791.0%
Prong 2: well positioned to advance it80025.0%
Prong 3: benefit of waiving the job offer80425.1%
Prong 1 as the only ground given2,08265.0%
AAO stopped without reaching the later prongs1,03932.4%

Shares are of the 3,204 decisions where a prong could be identified, not of all 4,987. A decision can fail more than one prong, so these do not sum to 100%. The other 1,783 name no prong the parser could read and are left blank rather than guessed.

EB-2 NIW

What the AAO says when it refuses

The same formulations recur. Counted as the number of decisions containing each one at least once.

What the decision saysDecisionsShareRate
“not the importance of the industry, field, or profession”1,80436.2%
the Dhanasar teaching-activities comparison1,38527.8%
impact must reach “beyond his employer, company, clients, or customers”81316.3%
“inconsistent”59912.0%
“generalized”50810.2%
“material change”4258.5%
“conclusory”3837.7%
“speculative”2434.9%
“unsupported”2354.7%

The top line is the whole game

In 1,804 decisions, 36.2% of the total, the AAO had to explain that the importance of your field is not the question it was asked. The question is whether your own specific proposed work is nationally important. That is the most common way these petitions are argued wrong.
EB-2 NIW

The cases these decisions turn on

Seven authorities account for almost all of the reasoning. Each links to its full text so you can check these summaries against the original.

CaseDecisionsShareWhat it decides
Dhanasar
Matter of Dhanasar, 26 I&N Dec. 884 (AAO 2016)
4,08381.9%the three-prong NIW test
Defines the modern NIW test: the endeavor has substantial merit and national importance, you are well positioned to advance it, and on balance it benefits the US to waive the job offer and labor certification. If you read one of these, read this one.
Full text on justice.gov.
Chawathe
Matter of Chawathe, 25 I&N Dec. 369 (AAO 2010)
4,04881.2%the standard of proof
You must show eligibility is more likely than not, not beyond doubt. Nearly every decision recites it, which is why it sits at the top of this table rather than because it decides anything.
Full text on justice.gov.
Bagamasbad
INS v. Bagamasbad, 429 U.S. 24 (1976)
2,99360.0%authority to stop after one prong
A Supreme Court decision, not an immigration one: agencies need not make findings on issues whose decision is unnecessary to the result. Where the AAO cites it, it has almost certainly stopped at prong 1 and never evaluated prongs 2 and 3 at all.
Full text on U.S. Reports via the Library of Congress.
Christo’s
Matter of Christo’s, Inc., 26 I&N Dec. 537 (AAO 2015)
2,14343.0%review starts again from scratch
The AAO reviews the whole record afresh rather than checking the service center for clear error. That cuts both ways: it can find new grounds to deny, and it can withdraw a prong the service center already granted you.
Full text on justice.gov.
Katigbak
Matter of Katigbak, 14 I&N Dec. 45 (Reg’l Comm’r 1971)
50310.1%eligibility is fixed on the filing date
You must have met every requirement on the day you filed. Qualifications earned afterwards do not count, however strong. This is why filing too early is expensive.
Full text on justice.gov.
Izummi
Matter of Izummi, 22 I&N Dec. 169 (Assoc. Comm’r 1998)
3306.6%you cannot fix it after filing
The companion to Katigbak: you cannot make material changes after filing to bring a deficient petition into compliance. Changing how you describe the endeavor mid-case is the trap this catches.
Full text on justice.gov.
Furtado
Matter of Furtado, 28 I&N Dec. 794 (BIA 2024)
190.4%evidence produced only on appeal
A 2024 Board of Immigration Appeals decision, the newest authority here. It lets the AAO refuse evidence produced for the first time on appeal when you were already put on notice of the gap, usually by an RFE. Answer the RFE fully; the appeal is not a second chance to file.
Full text on justice.gov.

Read this ranking carefully

It is mostly a ranking of procedural boilerplate, not of substantive doctrine. Dhanasar and Chawathe sit within a percentage point of each other because nearly every decision recites both the test and the burden of proof. The genuinely informative entry is Bagamasbad: it is the authority for declining to reach your remaining prongs, so wherever it appears the AAO probably never evaluated prongs 2 and 3. Two of these are not AAO decisions at all — Bagamasbad is a 1976 Supreme Court case, and Furtado is a 2024 Board of Immigration Appeals decision, the newest authority here.
EB-2 NIW

Who actually appeals

Petitioners in their own words, from the first line of each decision.

Self-described asDecisionsShareRate
Business and entrepreneurship1,52339%
Not classified75919%
Engineering, other than software45011%
Medicine and health3218%
Software, IT and data2847%
Academic and research2546%
Arts, sport and other1554%
Education1323%
Law and policy662%

This is an industry docket, not a researcher’s docket

Business, engineering and software together are 2,257 of 3,944 (57%). Academic and research petitioners are 254 (6%). Whatever you have read about the NIW being a route for researchers, the appeal docket is dominated by people working in industry.

Taken from each decision’s own opening line, in 3,944 of 4,987 cases. The groupings are keyword rules written for this page, not an official taxonomy, so treat them as indicative. “Not classified” is shown rather than hidden. Petitions whose opening line says only “member of the professions” are excluded, because that describes no occupation.

EB-2 NIW

How many decisions, by year

The shape of the record. A taller bar means more decisions were published that year, not a better or worse chance of winning.

dismissedmotion dismissedremandedsustained
1,6121,2098064030201520162017201820192020202120222023202420252026

200 of 4,987 decisions carry no date the parser could read and are not in this chart. Bar height reflects both how many people appealed and how quickly the AAO published, so it is not a success rate.

EB-2 NIW

Which service center denied it

Where the original denial came from, in the decisions that name it.

Service centerDecisionsShareRate
Texas3,03374%
Nebraska1,05226%
California20%

What this does not settle

A claim that circulates in applicant forums is that one service center issues NIW requests for evidence at many times another’s rate. These counts lean the same way but nowhere near that strongly, and appeal volume is not the same measurement as an RFE rate. This neither confirms nor refutes it.

The service center is named in only some decision formats, so this covers 4,087 of 4,987 decisions (82%). A center with only a handful of decisions is shown for completeness and should not be read as a meaningful rate.

Method

How these numbers were produced

Deterministic parsing of the source PDFs, and an honest account of what is missing.

A script downloads each decision as a PDF, extracts the text, and reads the fields out using fixed text patterns. Every row keeps the exact sentence it matched, so any figure here can be traced back to its own PDF. No language model is involved at any stage, so nothing on this page is summarised or inferred.

Coverage is 4,987 decisions against about 5,122 that the USCIS listing exposes for 2015 to 2026, roughly 97%. You can browse the source listing yourself; it is the authoritative record.

The parser records nothing rather than guessing, so some fields are sparse: no prong for 35.8%, no self-described occupation for 20.9%, no service center for 18.0%, no date for 4.0%. Every share on this page is computed over the decisions where that field was actually found, and each table states which denominator it used. A sparse field means a smaller denominator, not a zero.

The decision-by-decision listing is not published here. Browse it on the USCIS listing linked above.

The one thing this cannot tell you

Every decision here is an appeal of a denied petition. This is not a sample of NIW petitions generally, so nothing on this page is an approval rate. USCIS does not publish approved petitions, and no NIW-specific approval rate exists in any official source. A three per cent success rate on appeal says nothing about your chance of approval on a first filing.
Method

What this page does not show

Gaps named on purpose, because a plausible-looking number would be worse than an admitted blank.

  • EB-1A criterion outcomes. The same listing publishes EB-1A extraordinary-ability decisions under a different category. Counting which of the ten criteria fail needs a separate parser and a separate crawl, so it is absent rather than estimated.
  • Approval rates of any kind. Nothing here can be turned into one. See the note above.
  • Why any individual case was decided. These are counts of what decisions say, not an assessment of whether the AAO was right.

Published decisions are public records. This page counts them. It is not legal advice and it is not a substitute for a licensed immigration attorney.