*187The Immigration and Nationality Act (INA), 66 Stat. 163, 8 U.S.C. § 1101 et seq., provides that a noncitizen who has been convicted of an " aggravated felony" may be deported from this country. The INA also prohibits the Attorney General from granting discretionary relief from removal to an aggravated felon, no matter how compelling his case. Among the crimes that are classified as aggravated felonies, and thus lead to these harsh consequences, are illicit drug trafficking offenses. We must decide whether this category includes a state criminal statute that extends to the social sharing of a small amount of marijuana. We hold it does not.
I
A
The INA allows the Government to deport various classes of noncitizens, such as those who overstay their visas, and those who are convicted of certain crimes while in the United States, including drug offenses. § 1227. Ordinarily, when a noncitizen is found to be deportable on one of these grounds, he may ask the Attorney General for certain forms of discretionary relief from removal, like asylum (if he has a well-founded fear of persecution in his home country) and cancellation of removal (if, among other things, he has been lawfully present in the United States for a number of years). §§ 1158, 1229b. But if a noncitizen has been convicted of one of a narrower set of crimes classified as "aggravated felonies," then he is not only deportable, § 1227(a)(2)(A)(iii), but also ineligible for these discretionary forms of relief. See §§ 1158(b)(2)(A)(ii), (B)(i); §§ 1229b(a)(3), (b)(1)(C).1
*1683*188The INA defines "aggravated felony" to include a host of offenses. § 1101(a)(43). Among them is "illicit trafficking in a controlled substance." § 1101(a)(43)(B). This general term is not defined, but the INA states that it "includ[es] a drug trafficking crime (as defined in section 924(c) of title 18 )." Ibid. In turn, 18 U.S.C. § 924(c)(2) defines "drug trafficking crime" to mean "any felony punishable under the Controlled Substances Act," or two other statutes not relevant here. The chain of definitions ends with § 3559(a)(5), which provides that a "felony" is an offense for which the "maximum term of imprisonment authorized" is "more than one year." The upshot is that a noncitizen's conviction of an offense that the Controlled Substances Act (CSA) makes punishable by more than one year's imprisonment will be counted as an "aggravated felony" for immigration purposes. A conviction under either state or federal law may qualify, but a "state offense constitutes a 'felony punishable under the Controlled Substances Act' only if it proscribes conduct punishable as a felony under that federal law." Lopez v. Gonzales, 549 U.S. 47, 60, 127 S.Ct. 625, 166 L.Ed.2d 462 (2006).
B
Petitioner Adrian Moncrieffe is a Jamaican citizen who came to the United States legally in 1984, when he was three. During a 2007 traffic stop, police found 1.3 grams of marijuana in his car. This is the equivalent of about two or three marijuana cigarettes. Moncrieffe pleaded guilty to possession of marijuana with intent to distribute, a violation of Ga.Code Ann. § 16-13-30(j)(1) (2007). Under a Georgia statute providing more lenient treatment to first-time offenders, *189§ 42-8-60(a) (1997), the trial court withheld entering a judgment of conviction or imposing any term of imprisonment, and instead required that Moncrieffe complete five years of probation, after which his charge will be expunged altogether.2 App. to Brief for Petitioner 11-15.
Alleging that this Georgia conviction constituted an aggravated felony, the Federal Government sought to deport Moncrieffe. The Government reasoned that possession of marijuana with intent to distribute is an offense under the CSA, 21 U.S.C. § 841(a), punishable by up to five years' imprisonment, § 841(b)(1)(D), and thus an aggravated felony. An Immigration Judge agreed and ordered Moncrieffe removed. App. to Pet. for Cert. 14a-18a. The Board of Immigration Appeals (BIA) affirmed that conclusion on appeal. Id., at 10a-13a.
The Court of Appeals denied Moncrieffe's petition for review. The court rejected Moncrieffe's reliance upon § 841(b)(4), a provision that, in effect, makes marijuana distribution punishable only as a misdemeanor if the offense involves a small amount of marijuana for no remuneration. It held that in a federal criminal prosecution, "the default sentencing range for a marijuana distribution offense is the CSA's felony provision, § 841(b)(1)(D), rather than the misdemeanor provision."
*1684662 F.3d 387, 392 (C.A.5 2011). Because Moncrieffe's Georgia offense penalized possession of marijuana with intent to distribute, the court concluded that it was "equivalent to a federal felony." Ibid.
We granted certiorari, 566 U.S. ----, 132 S.Ct. 1857, 182 L.Ed.2d 642 (2012), to resolve a conflict among the Courts of Appeals with respect to whether a conviction under a statute that criminalizes conduct described by both § 841's felony provision and its misdemeanor provision, such as a statute that punishes all *190marijuana distribution without regard to the amount or remuneration, is a conviction for an offense that "proscribes conduct punishable as a felony under" the CSA.3 Lopez, 549 U.S., at 60, 127 S.Ct. 625. We now reverse.
II
A
When the Government alleges that a state conviction qualifies as an "aggravated felony" under the INA, we generally employ a "categorical approach" to determine whether the state offense is comparable to an offense listed in the INA. See, e.g., Nijhawan v. Holder, 557 U.S. 29, 33-38, 129 S.Ct. 2294, 174 L.Ed.2d 22 (2009) ; Gonzales v. Duenas-Alvarez, 549 U.S. 183, 185-187, 127 S.Ct. 815, 166 L.Ed.2d 683 (2007). Under this approach we look "not to the facts of the particular prior case," but instead to whether "the state statute defining the crime of conviction" categorically fits within the "generic" federal definition of a corresponding aggravated felony. Id., at 186, 127 S.Ct. 815 (citing Taylor v. United States, 495 U.S. 575, 599-600, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990) ). By "generic," we mean the offenses must be viewed in the abstract, to see whether the state statute shares the nature of the federal offense that serves as a point of comparison. Accordingly, a state offense is a categorical match with a generic federal offense only if a conviction of the state offense " 'necessarily' involved ... facts equating to [the] generic [federal offense]." Shepard v. United States, 544 U.S. 13, 24, 125 S.Ct. 1254, 161 L.Ed.2d 205 (2005) (plurality opinion). Whether the noncitizen's actual conduct involved such facts "is quite irrelevant." United States ex rel. Guarino v. Uhl, 107 F.2d 399, 400 (C.A.2 1939) (L. Hand, J.).
Because we examine what the state conviction necessarily involved, not the facts underlying the case, we must presume *191that the conviction " rested upon [nothing] more than the least of th[e] acts" criminalized, and then determine whether even those acts are encompassed by the generic federal offense. Johnson v. United States, 559 U.S. 133, 137, 130 S.Ct. 1265, 176 L.Ed.2d 1 (2010) ; see Guarino, 107 F.2d, at 400. But this rule is not without qualification. First, our cases have addressed state statutes that contain several different crimes, each described separately, and we have held that a court may determine which particular offense the noncitizen was convicted of by examining the charging document and jury instructions, or in the case of a guilty plea, the plea agreement, plea colloquy, or " 'some comparable judicial record' of the factual basis for the plea." Nijhawan, 557 U.S., at 35, 129 S.Ct. 2294 (quoting Shepard, 544 U.S., at 26, 125 S.Ct. 1254). Second, our focus on the minimum conduct criminalized by the state statute is not an *1685invitation to apply "legal imagination" to the state offense; there must be "a realistic probability, not a theoretical possibility, that the State would apply its statute to conduct that falls outside the generic definition of a crime." Duenas-Alvarez, 549 U.S., at 193, 127 S.Ct. 815.
This categorical approach has a long pedigree in our Nation's immigration law. See Das, The Immigration Penalties of Criminal Convictions: Resurrecting Categorical Analysis in Immigration Law, 86 N.Y.U.L.Rev. 1669, 1688-1702, 1749-1752 (2011) (tracing judicial decisions back to 1913). The reason is that the INA asks what offense the noncitizen was "convicted" of, 8 U.S.C. § 1227(a)(2)(A)(iii), not what acts he committed. "[C]onviction" is "the relevant statutory hook."4 Carachuri-Rosendo v. Holder, 560 U.S. ----, ----, 130 S.Ct. 2577, 2588, 177 L.Ed.2d 68 (2010); see United States ex rel. Mylius v. Uhl, 210 F. 860, 862 (C.A.2 1914).
B
*192The aggravated felony at issue here, "illicit trafficking in a controlled substance," is a "generic crim[e]." Nijhawan, 557 U.S., at 37, 129 S.Ct. 2294. So the categorical approach applies. Ibid. As we have explained, supra, at 1682 - 1683, this aggravated felony encompasses all state offenses that "proscrib[e] conduct punishable as a felony under [the CSA]." Lopez, 549 U.S., at 60, 127 S.Ct. 625. In other words, to satisfy the categorical approach, a state drug offense must meet two conditions: It must "necessarily" proscribe conduct that is an offense under the CSA, and the CSA must "necessarily" prescribe felony punishment for that conduct.
Moncrieffe was convicted under a Georgia statute that makes it a crime to "possess, have under [one's] control, manufacture, deliver, distribute, dispense, administer, purchase, sell, or possess with intent to distribute marijuana." Ga.Code Ann. § 16-13-30(j)(1). We know from his plea agreement that Moncrieffe was convicted of the last of these offenses. App. to Brief for Petitioner 11; Shepard, 544 U.S., at 26, 125 S.Ct. 1254. We therefore must determine whether possession of marijuana with intent to distribute is "necessarily" conduct punishable as a felony under the CSA.
We begin with the relevant conduct criminalized by the CSA. There is no question that it is a federal crime to "possess with intent to ... distribute ... a controlled substance," 21 U.S.C. § 841(a)(1), one of which is marijuana, § 812(c).5 So far, the state and federal provisions correspond. But this is not enough, because the generically defined federal crime is "any felony punishable under the Controlled Substances Act," 18 U.S.C. § 924(c)(2), not just any "offense under the *193CSA." Thus we must look to what punishment the CSA imposes for this offense.
Section 841 is divided into two subsections that are relevant here: (a), titled "Unlawful acts," which includes the offense just described, and (b), titled "Penalties." Subsection (b) tells us how "any person who violates subsection (a)" shall be punished, *1686depending on the circumstances of his crime (e.g., the type and quantity of controlled substance involved, whether it is a repeat offense).6 Subsection (b)(1)(D) provides that if a person commits a violation of subsection (a) involving "less than 50 kilograms of marihuana," then "such person shall, except as provided in paragraphs (4) and (5) of this subsection, be sentenced to a term of imprisonment of not more than 5 years," i.e., as a felon. But one of the exceptions is important here. Paragraph (4) provides, "Notwithstanding paragraph (1)(D) of this subsection, any person who violates subsection (a) of this section by distributing a small amount of marihuana for no remuneration shall be treated as" a simple drug possessor, 21 U.S.C. § 844, which *194for our purposes means as a misdemeanant.7 These dovetailing provisions create two mutually exclusive categories of punishment for CSA marijuana distribution offenses: one a felony, and one not. The only way to know whether a marijuana distribution offense is "punishable as a felony" under the CSA, Lopez, 549 U.S., at 60, 127 S.Ct. 625, is to know whether the conditions described in paragraph (4) are present or absent.
A conviction under the same Georgia statute for "sell[ing]" marijuana, for example, would seem to establish remuneration. The presence of remuneration would mean that paragraph (4) is not implicated, and thus that the conviction is necessarily for conduct punishable as a felony under the CSA (under paragraph (1)(D)). In contrast, the fact of a conviction for possession with intent to distribute marijuana, standing alone, does not reveal whether either remuneration or more than a small amount of marijuana was involved. It is possible neither was; we know that Georgia prosecutes this offense when a defendant possesses only a small amount of marijuana, see, e.g., Taylor v. State, 260 Ga.App. 890, 581 S.E.2d 386, 388 (2003) (6.6 grams), and that "distribution" does not require remuneration, see, e.g., Hadden v. State, 181 Ga.App. 628, 628-629, 353 S.E.2d 532, 533-534 (1987). So Moncrieffe's *1687conviction could correspond to either the CSA felony or the CSA misdemeanor. Ambiguity on this point means that the conviction did not "necessarily" involve *195facts that correspond to an offense punishable as a felony under the CSA. Under the categorical approach, then, Moncrieffe was not convicted of an aggravated felony.
III
A
The Government advances a different approach that leads to a different result. In its view, § 841(b)(4)'s misdemeanor provision is irrelevant to the categorical analysis because paragraph (4) is merely a "mitigating exception," to the CSA offense, not one of the "elements" of the offense. Brief for Respondent 12. And because possession with intent to distribute marijuana is "presumptive[ly]" a felony under the CSA, the Government asserts, any state offense with the same elements is presumptively an aggravated felony. Id., at 37. These two contentions are related, and we reject both of them.
First, the Government reads our cases to hold that the categorical approach is concerned only with the "elements" of an offense, so § 841(b)(4)"is not relevant" to the categorical analysis. Id., at 20. It is enough to satisfy the categorical inquiry, the Government suggests, that the "elements" of Moncrieffe's Georgia offense are the same as those of the CSA offense: (1) possession (2) of marijuana (a controlled substance), (3) with intent to distribute it. But that understanding is inconsistent with Carachuri-Rosendo, our only decision to address both "elements" and "sentencing factors." There we recognized that when Congress has chosen to define the generic federal offense by reference to punishment, it may be necessary to take account of federal sentencing factors too. See 560 U.S., at ----, 130 S.Ct., at 2581-2582. In that case the relevant CSA offense was simple possession, which "becomes a 'felony punishable under the [CSA]' only because the sentencing factor of recidivism authorizes additional punishment beyond one year, the criterion for a felony." Id., at ----, 130 S.Ct., at 2590 (SCALIA, J., concurring in judgment). We therefore called *196the generic federal offense "recidivist simple possession," even though such a crime is not actually "a separate offense" under the CSA, but rather an " 'amalgam' " of offense elements and sentencing factors. Id., at ----, and n. 3, ----, 130 S.Ct., at 2581-2582, and n. 3, 2583-2584 (majority opinion).
In other words, not only must the state offense of conviction meet the "elements" of the generic federal offense defined by the INA, but the CSA must punish that offense as a felony. Here, the facts giving rise to the CSA offense establish a crime that may be either a felony or a misdemeanor, depending upon the presence or absence of certain factors that are not themselves elements of the crime. And so to qualify as an aggravated felony, a conviction for the predicate offense must necessarily establish those factors as well.
The Government attempts to distinguish Carachuri-Rosendo on the ground that the sentencing factor there was a "narrow" aggravating exception that turned a misdemeanor into a felony, whereas here § 841(b)(4) is a narrow mitigation exception that turns a felony into a misdemeanor. Brief for Respondent 40-43. This argument hinges upon the Government's second assertion: that any marijuana distribution conviction is "presumptively" a felony. But that is simply incorrect, and the Government's argument collapses as a result. Marijuana distribution is neither a felony nor a misdemeanor until we know whether the conditions in paragraph (4)
*1688attach: Section 841(b)(1)(D) makes the crime punishable by five years' imprisonment "except as provided" in paragraph (4), and § 841(b)(4) makes it punishable as a misdemeanor "[n]otwithstanding paragraph (1)(D)" when only "a small amount of marihuana for no remuneration" is involved. (Emphasis added.) The CSA's text makes neither provision the default. Rather, each is drafted to be exclusive of the other.
Like the BIA and the Fifth Circuit, the Government believes the felony provision to be the default because, in practice, that is how federal criminal prosecutions for marijuana *197distribution operate. See 662 F.3d, at 391-392; Matter of Aruna, 24 I. & N. Dec. 452, 456-457 (2008) ; Brief for Respondent 18-23. It is true that every Court of Appeals to have considered the question has held that a defendant is eligible for a 5-year sentence under § 841(b)(1)(D) if the Government proves he possessed marijuana with the intent to distribute it, and that the Government need not negate the § 841(b)(4) factors in each case. See, e.g., United States v. Outen, 286 F.3d 622, 636-639 (C.A.2 2002) (describing § 841(b)(4) as a "mitigating exception"); United States v. Hamlin, 319 F.3d 666, 670-671 (C.A.4 2003) (collecting cases). Instead, the burden is on the defendant to show that he qualifies for the lesser sentence under § 841(b) (4). Cf. id., at 671.
We cannot discount § 841's text, however, which creates no default punishment, in favor of the procedural overlay or burdens of proof that would apply in a hypothetical federal criminal prosecution. In Carachuri-Rosendo, we rejected the Fifth Circuit's " 'hypothetical approach,' " which examined whether conduct " 'could have been punished as a felony' 'had [it] been prosecuted in federal court.' " 560 U.S., at ----, ----, 130 S.Ct., at 2584, 2585-2586.8 The outcome in a hypothetical prosecution is not the relevant inquiry. Rather, our "more focused, categorical inquiry" is whether the record of conviction of the predicate *198offense necessarily establishes conduct that the CSA, on its own terms, makes punishable as a felony. Id., at ----, 130 S.Ct., at 2588-2589.
The analogy to a federal prosecution is misplaced for another reason. The Court of Appeals cases the Government cites distinguished between elements and sentencing factors to determine which facts must be proved to a jury, in light of the Sixth Amendment concerns addressed in Apprendi v. New Jersey, 530 U.S. 466, 120 S.Ct. 2348, 147 L.Ed.2d 435 (2000). The courts considered which "provision ... states a complete crime upon the fewest facts, " Outen, 286 F.3d, at 638, which was significant after Apprendi to identify what a jury had to find before a defendant could receive § 841(b)(1)(D)'s maximum 5-year sentence. But those concerns do not apply in this context. Here we consider a "generic"
*1689federal offense in the abstract, not an actual federal offense being prosecuted before a jury. Our concern is only which facts the CSA relies upon to distinguish between felonies and misdemeanors, not which facts must be found by a jury as opposed to a judge, nor who has the burden of proving which facts in a federal prosecution.9
Because of these differences, we made clear in Carachuri-Rosendo that, for purposes of the INA, a generic federal offense may be defined by reference to both " 'elements' in the traditional sense" and sentencing factors. 560 U.S., at ----, n. 3, ----, 130 S.Ct., at 2581-2582, and n. 3, 2583-2584; see also id., at ----, 130 S.Ct., at 2581-2582 (SCALIA, J., concurring in judgment) (describing the generic federal offense there as "the Controlled Substances Act felony of possession-plus-recidivism"). Indeed, the distinction between "elements" and "sentencing factors" did not exist when Congress added illicit drug trafficking to the list of aggravated felonies, Anti-Drug Abuse Act of 1988, 102 *199Stat. 4469-4470, and most courts at the time understood both § 841(b)(1)(D) and § 841(b)(4) to contain sentencing factors that draw the line between a felony and a misdemeanor. See, e.g., United States v. Campuzano, 905 F.2d 677, 679 (C.A.2 1990). Carachuri-Rosendo controls here.
Finally, there is a more fundamental flaw in the Government's approach: It would render even an undisputed misdemeanor an aggravated felony. This is "just what the English language tells us not to expect," and that leaves us "very wary of the Government's position." Lopez, 549 U.S., at 54, 127 S.Ct. 625. Consider a conviction under a New York statute that provides, "A person is guilty of criminal sale of marihuana in the fifth degree when he knowingly and unlawfully sells, without consideration, [marihuana] of an aggregate weight of two grams or less ; or one cigarette containing marihuana." N.Y. Penal Law Ann. § 221.35 (West 2008) (emphasis added). This statute criminalizes only the distribution of a small amount of marijuana for no remuneration, and so all convictions under the statute would fit within the CSA misdemeanor provision, § 841(b)(4). But the Government would categorically deem a conviction under this statute to be an aggravated felony, because the statute contains the corresponding "elements" of (1) distributing (2) marijuana, and the Government believes all marijuana distribution offenses are punishable as felonies.
The same anomaly would result in the case of a noncitizen convicted of a misdemeanor in federal court under § 841(a) and (b)(4) directly. Even in that case, under the Government's logic, we would need to treat the federal misdemeanor conviction as an aggravated felony, because the conviction establishes elements of an offense that is presumptively a felony. This cannot be. "We cannot imagine that Congress took the trouble to incorporate its own statutory scheme of felonies and misdemeanors," only to have courts presume felony treatment and ignore the very factors that distinguish felonies from misdemeanors. Lopez, 549 U.S., at 58, 127 S.Ct. 625.
*200Recognizing that its approach leads to consequences Congress could not have intended, the Government hedges its argument by proposing a remedy: Noncitizens should be given an opportunity during immigration proceedings to demonstrate that their predicate marijuana distribution convictions involved only a small amount of marijuana and no remuneration, just as a federal criminal defendant could do at sentencing. Brief for Respondent 35-39. This is the procedure adopted by the BIA in Matter of Castro Rodriguez, 25 I. & N. Dec. 698, 702 (2012), and endorsed by Justice ALITO's dissent, post, at 1701 - 1702.
This solution is entirely inconsistent with both the INA's text and the categorical approach. As noted, the relevant INA provisions ask what the noncitizen was "convicted of," not what he did, and the inquiry in immigration proceedings is limited accordingly. 8 U.S.C. §§ 1227(a)(2)(A)(iii), 1229b(a)(3) ; see Carachuri-Rosendo, 560 U.S., at ----, 130 S.Ct., at 2585-2586. The Government cites no statutory authority for such case-specific factfinding in immigration court, and none is apparent in the INA. Indeed, the Government's main categorical argument would seem to preclude this inquiry: If the Government were correct that "the fact of a marijuana-distribution conviction alone constitutes a CSA felony," Brief for Respondent 37, then all marijuana distribution convictions would categorically be convictions of the drug trafficking aggravated felony, mandatory deportation would follow under the statute, and there would be no room for the Government's follow-on factfinding procedure. The Government cannot have it both ways.
Moreover, the procedure the Government envisions would require precisely the sort of post hoc investigation into the facts of predicate offenses that we have long deemed undesirable. The categorical approach serves "practical" purposes: It promotes judicial and administrative efficiency by precluding the relitigation of past convictions in minitrials conducted *201long after the fact. Chambers v. United States, 555 U.S. 122, 125, 129 S.Ct. 687, 172 L.Ed.2d 484 (2009) ; see also Mylius, 210 F., at 862-863. Yet the Government's approach would have our Nation's overburdened immigration courts entertain and weigh testimony from, for example, the friend of a noncitizen who may have shared a marijuana cigarette with him at a party, or the local police officer who recalls to the contrary that cash traded hands. And, as a result, two noncitizens, each "convicted of" the same offense, might obtain different aggravated felony determinations depending on what evidence remains available or how it is perceived by an individual immigration judge. The categorical approach was designed to avoid this "potential unfairness." Taylor, 495 U.S., at 601, 110 S.Ct. 2143; see also Mylius, 210 F., at 863.
Furthermore, the minitrials the Government proposes would be possible only if the noncitizen could locate witnesses years after the fact, notwithstanding that during removal proceedings noncitizens are not guaranteed legal representation and are often subject to mandatory detention, § 1226(c)(1)(B), where they have little ability to collect evidence. See Katzmann, The Legal Profession and the Unmet Needs of the Immigrant Poor, 21 Geo. J. Legal Ethics 3, 5-10 (2008) ; Brief for National Immigrant Justice Center et al. as Amici Curiae 5-18; Brief for Immigration Law Professors as Amici Curiae 27-32. A noncitizen in removal proceedings is not at all similarly situated to a defendant in a federal criminal prosecution. The Government's suggestion that the CSA's procedures could readily be replicated in immigration *1691proceedings is therefore misplaced. Cf. Carachuri-Rosendo, 560 U.S., at ----, 130 S.Ct., at 2587-2588 (rejecting the Government's argument that procedures governing determination of the recidivism sentencing factor could "be satisfied during the immigration proceeding").
The Government defends its proposed immigration court proceedings as "a subsequent step outside the categorical approach in light of Section 841(b)(4)'s 'circumstance-specific'
*202nature." Brief for Respondent 37. This argument rests upon Nijhawan, in which we considered another aggravated felony, "an offense that ... involves fraud or deceit in which the loss to the victim or victims exceeds $10,000." 8 U.S.C. § 1101(a)(43)(M) (i). We held that the $10,000 threshold was not to be applied categorically as a required component of a generic offense, but instead called for a " circumstance-specific approach" that allows for an examination, in immigration court, of the "particular circumstances in which an offender committed the crime on a particular occasion." Nijhawan, 557 U.S., at 38-40, 129 S.Ct. 2294. The Government suggests the § 841(b)(4) factors are like the monetary threshold, and thus similarly amenable to a circumstance-specific inquiry.
We explained in Nijhawan, however, that unlike the provision there, "illicit trafficking in a controlled substance" is a "generic crim[e]" to which the categorical approach applies, not a circumstance-specific provision. Id., at 37, 129 S.Ct. 2294; see also Carachuri-Rosendo, 560 U.S., at ----, n. 11, 130 S.Ct., at 2586-2587 n. 11. That distinction is evident in the structure of the INA. The monetary threshold is a limitation, written into the INA itself, on the scope of the aggravated felony for fraud. And the monetary threshold is set off by the words "in which," which calls for a circumstance-specific examination of "the conduct involved 'in ' the commission of the offense of conviction." Nijhawan, 557 U.S., at 39, 129 S.Ct. 2294. Locating this exception in the INA proper suggests an intent to have the relevant facts found in immigration proceedings. But where, as here, the INA incorporates other criminal statutes wholesale, we have held it "must refer to generic crimes," to which the categorical approach applies. Id., at 37, 129 S.Ct. 2294.
Finally, the Government suggests that the immigration court's task would not be so daunting in some cases, such as those in which a noncitizen was convicted under the New York statute previously discussed or convicted directly under § 841(b)(4). True, in those cases, the record of conviction might reveal on its face that the predicate offense was *203punishable only as a misdemeanor. But most States do not have stand-alone offenses for the social sharing of marijuana, so minitrials concerning convictions from the other States, such as Georgia, would be inevitable.10 The Government suggests that even in these other States, the record of conviction may often address the § 841(b)(4) factors, because noncitizens "will be advised of the immigration *1692consequences of a conviction," as defense counsel is required to do under Padilla v. Kentucky, 559 U.S. 356, 130 S.Ct. 1473, 176 L.Ed.2d 284 (2010), and as a result counsel can build an appropriate record when the facts are fresh. Brief for Respondent 38. Even assuming defense counsel "will" do something simply because it is required of effective counsel (an assumption experience does not always bear out), this argument is unavailing because there is no reason to believe that state courts will regularly or uniformly admit evidence going to facts, such as remuneration, that are irrelevant to the offense charged.
In short, to avoid the absurd consequences that would flow from the Government's narrow understanding of the categorical approach, the Government proposes a solution that largely undermines the categorical approach. That the only cure is worse than the disease suggests the Government is simply wrong.
C
The Government fears the consequences of our decision, but its concerns are exaggerated. The Government observes *204that, like Georgia, about half the States criminalize marijuana distribution through statutes that do not require remuneration or any minimum quantity of marijuana. Id., at 26-28. As a result, the Government contends, noncitizens convicted of marijuana distribution offenses in those States will avoid "aggravated felony" determinations, purely because their convictions do not resolve whether their offenses involved federal felony conduct or misdemeanor conduct, even though many (if not most) prosecutions involve either remuneration or larger amounts of marijuana (or both).
Escaping aggravated felony treatment does not mean escaping deportation, though. It means only avoiding mandatory removal. See Carachuri-Rosendo, 560 U.S., at ----, 130 S.Ct., at 2589. Any marijuana distribution offense, even a misdemeanor, will still render a noncitizen deportable as a controlled substances offender. 8 U.S.C. § 1227(a)(2)(B)(i). At that point, having been found not to be an aggravated felon, the noncitizen may seek relief from removal such as asylum or cancellation of removal, assuming he satisfies the other eligibility criteria. §§ 1158(b), 1229b(a)(1)-(2). But those forms of relief are discretionary. The Attorney General may, in his discretion, deny relief if he finds that the noncitizen is actually a member of one "of the world's most dangerous drug cartels," post, at 1696 (opinion of ALITO, J.), just as he may deny relief if he concludes the negative equities outweigh the positive equities of the noncitizen's case for other reasons. As a result, "to the extent that our rejection of the Government's broad understanding of the scope of 'aggravated felony' may have any practical effect on policing our Nation's borders, it is a limited one." Carachuri-Rosendo, 560 U.S., at ----, 130 S.Ct., at 2589.
In any event, serious drug traffickers may be adjudicated aggravated felons regardless, because they will likely be convicted under greater "trafficking" offenses that necessarily establish that more than a small amount of marijuana was involved. See, e.g., Ga.Code Ann. § 16-13-31(c)(1) (Supp.2012)
*205(separate provision for trafficking in more than 10 pounds of marijuana). Of course, some offenders' conduct will fall between § 841(b)(4) conduct and the more serious conduct required to trigger a "trafficking" statute. Brief for Respondent 30. Those offenders may avoid aggravated felony status by operation of the categorical approach. But the Government's objection to that underinclusive result is little more than an attack on the *1693categorical approach itself.11 We prefer this degree of imperfection to the heavy burden of relitigating old prosecutions. See supra, at 1690 - 1691. And we err on the side of underinclusiveness because ambiguity in criminal statutes referenced by the INA must be construed in the noncitizen's favor. See Carachuri-Rosendo, 560 U.S., at ----, 130 S.Ct., at 2589 ; Leocal v. Ashcroft, 543 U.S. 1, 11, n. 8, 125 S.Ct. 377, 160 L.Ed.2d 271 (2004).
Finally, the Government suggests that our holding will frustrate the enforcement of other aggravated felony provisions, like § 1101(a)(43)(C), which refers to a federal firearms statute that contains an exception for "antique firearm[s]," 18 U.S.C. § 921(a)(3). The Government fears that a conviction under any state firearms law that lacks such an exception will be deemed to fail the categorical inquiry. But Duenas-Alvarez *206requires that there be "a realistic probability, not a theoretical possibility, that the State would apply its statute to conduct that falls outside the generic definition of a crime." 549 U.S., at 193, 127 S.Ct. 815. To defeat the categorical comparison in this manner, a noncitizen would have to demonstrate that the State actually prosecutes the relevant offense in cases involving antique firearms. Further, the Government points to § 1101 (a)(43)(P), which makes passport fraud an aggravated felony, except when the noncitizen shows he committed the offense to assist an immediate family member. But that exception is provided in the INA itself. As we held in Nijhawan, a circumstance-specific inquiry would apply to that provision, so it is not comparable. 557 U.S., at 37-38, 129 S.Ct. 2294.
* * *
This is the third time in seven years that we have considered whether the Government has properly characterized a low-level drug offense as "illicit trafficking in a controlled substance," and thus an "aggravated felony." Once again we hold that the Government's approach defies "the 'commonsense conception' " of these terms. Carachuri-Rosendo, 560 U.S., at ----, 130 S.Ct., at 2584-2585 (quoting Lopez, 549 U.S., at 53, 127 S.Ct. 625). Sharing a small amount of marijuana for no remuneration, let alone possession with intent to do so, "does not fit easily into the 'everyday understanding' " of "trafficking," which " 'ordinarily ... means some sort of commercial dealing.' " Carachuri-Rosendo, 560 U.S., at ----, 130 S.Ct., at 2584-2585 (quoting Lopez, 549 U.S., at 53-54, 127 S.Ct. 625). Nor is it sensible that a state statute that criminalizes conduct that the CSA treats as a misdemeanor should be designated an "aggravated felony." We hold that it may not be. If a noncitizen's conviction for a marijuana distribution offense fails to establish that the offense involved either remuneration or more than *1694a small amount of marijuana, the conviction is not for an aggravated felony under the INA. The contrary judgment of the Court of Appeals *207is reversed, and the case is remanded for further proceedings consistent with this opinion.
It is so ordered.
In addition to asylum, a noncitizen who fears persecution may seek withholding of removal, 8 U.S.C. § 1231(b)(3)(A), and deferral of removal under the Convention Against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment (CAT), Art. 3, Dec. 10, 1984, S. Treaty Doc. No. 100-20, p. 20, 1465 U.N.T.S. 85; 8 CFR § 1208.17(a) (2012). These forms of relief require the noncitizen to show a greater likelihood of persecution or torture at home than is necessary for asylum, but the Attorney General has no discretion to deny relief to a noncitizen who establishes his eligibility. A conviction of an aggravated felony has no effect on CAT eligibility, but will render a noncitizen ineligible for withholding of removal if he "has been sentenced to an aggregate term of imprisonment of at least 5 years" for any aggravated felonies. 8 U.S.C. § 1231(b)(3)(B).
The parties agree that this resolution of Moncrieffe's Georgia case is nevertheless a "conviction" as the INA defines that term, 8 U.S.C. § 1101(a)(48)(A). See Brief for Petitioner 6, n. 2; Brief for Respondent 5, n. 2.
Compare 662 F.3d 387 (C.A.5 2011) (case below), Garcia v. Holder, 638 F.3d 511 (C.A.6 2011) (is an aggravated felony), and Julce v. Mukasey, 530 F.3d 30 (C.A.1 2008) (same), with Martinez v. Mukasey, 551 F.3d 113 (C.A.2 2008) (is not an aggravated felony), and Wilson v. Ashcroft, 350 F.3d 377 (C.A.3 2003) (same).
Carachuri-Rosendo construed a different provision of the INA that concerns cancellation of removal, which also requires determining whether the noncitizen has been "convicted of any aggravated felony." 8 U.S.C. § 1229b(a)(3) (emphasis added). Our analysis is the same in both contexts.
In full, 21 U.S.C. § 841(a)(1) provides,
"Except as authorized by this subchapter, it shall be unlawful for any person knowingly or intentionally-
"(1) to manufacture, distribute, or dispense, or possess with intent to manufacture, distribute, or dispense, a controlled substance...."
In pertinent part, § 841(b)(1)(D) and (b)(4) (2006 ed. and Supp. V) provide,
"Except as otherwise provided in section 849, 859, 860, or 861 of this title, any person who violates subsection (a) of this section shall be sentenced as follows:
. . . . .
"[ (1) ](D) In the case of less than 50 kilograms of marihuana, except in the case of 50 or more marihuana plants regardless of weight, 10 kilograms of hashish, or one kilogram of hashish oil, such person shall, except as provided in paragraphs (4) and (5) of this subsection, be sentenced to a term of imprisonment of not more than 5 years, a fine not to exceed the greater of that authorized in accordance with the provisions of title 18 or $250,000 if the defendant is an individual or $1,000,000 if the defendant is other than an individual, or both....
. . . . .
"(4) Notwithstanding paragraph (1)(D) of this subsection, any person who violates subsection (a) of this section by distributing a small amount of marihuana for no remuneration shall be treated as provided in section 844 of this title and section 3607 of title 18."
Although paragraph (4) speaks only of "distributing" marijuana, the parties agree that it also applies to "the more inchoate offense of possession with intent to distribute that drug." Matter of Castro Rodriguez, 25 I. & N. Dec. 698, 699, n. 2 (BIA 2012) ; see Brief for Petitioner 6, n. 2; Brief for Respondent 8, n. 5.
The CSA does not define "small amount." The BIA has suggested that 30 grams "serve[s] as a useful guidepost," Castro Rodriguez, 25 I. & N. Dec., at 703, noting that the INA exempts from deportable controlled substances offenses "a single offense involving possession for one's own use of 30 grams or less of marijuana," 8 U.S.C. § 1227(a)(2)(B)(i). The meaning of "small amount" is not at issue in this case, so we need not, and do not, define the term.
Justice ALITO states that the statute "obviously" requires examination of whether "conduct associated with the state offense ... would have supported a qualifying conviction under the federal CSA." Post, at 1697 (dissenting opinion) (emphasis added); see also post, at 1699. But this echoes the Fifth Circuit's approach in Carachuri-Rosendo . As noted in the text, our opinion explicitly rejected such reasoning based on conditional perfect formulations. See also, e.g., Carachuri-Rosendo, 560 U.S., at ----, 130 S.Ct., at 2588-2589 (criticizing approach that "focuses on facts known to the immigration court that could have but did not serve as the basis for the state conviction and punishment" (emphasis altered)). Instead, as we have explained, supra, at 1687 - 1688, our holding depended upon the fact that Carachuri-Rosendo's conviction did not establish the fact necessary to distinguish between misdemeanor and felony punishment under the CSA. The same is true here.
The Government also cites 21 U.S.C. § 885(a)(1), which provides that the Government need not "negative any exemption or exception set forth" in the CSA, and instead "the burden of going forward with the evidence with respect to any such exemption or exception shall be upon the person claiming its benefit." Brief for Respondent 21. Even assuming § 841(b)(4) is such an "exception," § 885(a)(1) applies, by its own terms, only to "any trial, hearing, or other proceeding under" the CSA itself, not to the rather different proceedings under the INA.
In addition to New York, it appears that 13 other States have separate offenses for § 841(b)(4) conduct. See Cal. Health & Safety Code Ann. § 11360(b) (West Supp.2013); Colo.Rev.Stat. Ann. § 18-18-406(5) (2012) ; Fla. Stat. § 893.13(2)(b)(3) (2010) ; Ill. Comp. Stat., ch. 20, §§ 550/3, 550/4, 550/6 (West 2010); Iowa Code § 124.410 (2009); Minn.Stat. § 152.027(4)(a) (2010) ; N.M. Stat. Ann. § 30-31-22(E) (Supp.2011); Ohio Rev.Code Ann. § 2925.03(C)(3)(h) ( Lexis 2012 Cum.Supp.); Ore.Rev.Stat. § 475.860(3) (2011); Pa. Stat. Ann., Tit. 35, § 780-113(a)(31) (Purdon Supp.2012); S.D. Codified Laws § 22-42-7 (Supp.2012); Tex. Health & Safety Code Ann. § 481.120(b)(1) (West 2010); W. Va.Code Ann. § 60A-4-402(c) (Lexis 2010).
Similarly, Justice ALITO's dissent suggests that he disagrees with the first premises of the categorical approach. He says it is a "strange and disruptive resul[t]" that "defendants convicted in different States for committing the same criminal conduct" might suffer different collateral consequences depending upon how those States define their statutes of conviction. Post, at 9. Yet that is the longstanding, natural result of the categorical approach, which focuses not on the criminal conduct a defendant "commit[s]," but rather what facts are necessarily established by a conviction for the state offense. Different state offenses will necessarily establish different facts. Some will track the "uniform" federal definition of the generic offense, and some will not. Taylor v. United States, 495 U.S. 575, 590, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990). Whatever disparity this may create as between defendants whose real-world conduct was the same, it ensures that all defendants whose convictions establish the same facts will be treated consistently, and thus predictably, under federal law. This was Taylor 's chief concern in adopting the categorical approach. See id., at 599-602, 110 S.Ct. 2143.