PAMELA HARRIS, Circuit Judge:

Musician Devin Copeland ("Copeland"), together with his songwriting partner, appeals the dismissal of his copyright infringement claim against recording artists Justin Bieber and Usher Raymond IV. Copeland alleges that three recorded songs by the defendants, each titled "Somebody to Love," infringe upon his copyright over his own, earlier song of the same name. The district court granted the defendants' motions to dismiss on the ground that no reasonable jury could find Copeland's song and the defendants' songs sufficiently similar to give rise to liability for infringement. We disagree, and therefore vacate the district court's order and remand the case for further proceedings.

I.

A.

Because Copeland appeals from an order granting a motion to dismiss under Rule 12(b)(6) of the Federal Rules of Civil Procedure, we recount the facts as alleged by Copeland, accepting them as true for purposes of this appeal. See Jackson v. Lightsey, 775 F.3d 170, 173 (4th Cir.2014).

Copeland is a Virginia-based R & B singer and songwriter who performs under the name "De Rico." In 2008, together with his songwriting partner Mareio Overton, Copeland began writing and recording songs to perform on his upcoming album, My Story II. Among them was "Somebody to Love," the song that is the subject matter of this case (the "Copeland song"). Copeland registered a copyright for the My Story II songs, including "Somebody to Love," later that year.

In late 2009, Copeland entered into discussions with Sangreel Media ("Sangreel"), a company that recruits artists for record labels including Island Records, Sony Music, and RCA Records. Sangreel was interested in promoting Copeland's music, and Copeland turned over copies of My Story II so that Sangreel could provide promotional copies to its clients. Among the figures to whom Sangreel presented Copeland's music was Usher Raymond IV, a world-famous recording artist who performs under the name "Usher."

According to Copeland's complaint, Usher liked what he heard. Usher's mother and manager, Jonetta Patton ("Patton"), scheduled a conference call with Copeland, during which she informed Copeland that both she and Usher had listened to My Story II, and that they were interested in having Copeland re-record the album and join Usher on tour. Yet the plans never materialized, and that was the last Copeland heard from anyone in Usher's camp.

Within a few months of Copeland's phone conversation with Patton, however, Usher had recorded and posted on his YouTube channel a demo song also titled "Somebody to Love" (the "Usher demo song"). Usher did not commercially release this song, but instead allegedly brought it to his protégé and fellow recording artist, Justin Bieber ("Bieber"). Bieber recorded his own "Somebody to Love" (the "Bieber album song") and released it on his debut album, My World 2.0, in the spring of 2010. Bieber's "Somebody to Love" was a hit, peaking at number 15 on the U.S. Billboard Hot 100 chart. Finally, Bieber released a fourth and final "Somebody to Love," a remix with lead vocals by both himself and Usher (the "Bieber-Usher remix song") in June 2010. Bieber has continued to perform live versions of those songs while on tour.

B.

Alleging that Bieber and Usher had access to the Copeland song via Sangreel and that their songs bear a striking resemblance to his own work, Copeland filed suit for copyright infringement against Bieber, Usher, and other associated defendants (collectively, "Bieber and Usher"). Bieber and Usher moved to dismiss the action under Rule 12(b)(6), contending, as relevant here, that no reasonable jury could find that the Copeland song and the Bieber and Usher songs were "substantially similar," as required to make out an infringement claim.

After a hearing, the district court agreed with Bieber and Usher, and entered an order granting their motions to dismiss. The court applied our precedent requiring copyright plaintiffs to prove two distinct forms of similarity: "extrinsic" similarity, an objective match between the copyright-protectable elements of an original work and a purported copy, often based on expert testimony; and "intrinsic" similarity, a more subjective and "essentially aesthetic judgment" as to whether the intended audience of two works would experience them as similar in overall effect. While acknowledging that substantial similarity is "largely a matter of fact," J.A. 249, the district court understood our precedent to allow for dismissal on the pleadings under either prong where no reasonable jury could find substantial similarity of the requisite kind.

The court began its analysis with intrinsic similarity, and the overall appeal of the two works to their intended audience. Relying again on Fourth Circuit precedent, the court held that the relevant "intended audience" in this case is the general public, as the expected ultimate market for Copeland's song. And the general public, the court concluded, would not "construe the aesthetic appeal of the songs as being similar," for despite some shared elements, the "mood, tone, and subject matter" of the songs differ "significantly." J.A. 253-54. Having decided that no reasonable jury could find the songs substantially similar under the intrinsic prong, the district court granted defendants' motions to dismiss without reaching extrinsic similarity. This appeal followed.

II.

To establish a claim for copyright infringement under the Copyright Act of 1976, 17 U.S.C. § 101 et seq., a plaintiff must prove that it possesses a valid copyright and that the defendant copied elements of its work that are original and protectable. See Feist Publ'ns, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340, 361, 111 S.Ct. 1282, 113 L.Ed.2d 358 (1991); Universal Furniture Int'l, Inc., v. Collezione Europa USA, Inc., 618 F.3d 417, 435 (4th Cir.2010). Absent direct proof of copying, which is hard to come by, a plaintiff may prove copying indirectly, with evidence showing that the defendant had access to the copyrighted work and that the purported copy is "substantially similar" to the original. See Universal Furniture, 618 F.3d at 435. It is that final step in the analysis that is at issue here. Bieber and Usher do not challenge Copeland's copyright in his song nor their access to that song. Instead, this case turns on whether Copeland can show the "substantial similarity" that would give rise, together with undisputed access, to a presumption that Bieber and Usher copied his song.

We begin by laying out the standard under which we consider that question, and addressing Copeland's arguments for changes or refinements to that standard. As the district court correctly explained, in this circuit, a plaintiff's substantial similarity showing has two components: extrinsic and intrinsic similarity. And while both bear, obviously, on the likeness between a copyrighted work and a supposed copy, they are different in important ways.

The "extrinsic inquiry is an objective one," looking to specific and "external criteria" of substantial similarity between the original elements (and only the original elements) of a protected work and an alleged copy. Id. at 435-36; see also Dawson v. Hinshaw Music, Inc., 905 F.2d 731, 732-33 (4th Cir.1990). Because the inquiry is objective, expert testimony often will be relevant. Universal Furniture, 618 F.3d at 435. And because it is focused only on the original elements of the copyrighted work, a court examining extrinsic similarity must first engage in a process we sometimes call "analytic dissection," separating out those parts of the work that are original and protected from those that are not. See id. at 436-37.

Intrinsic similarity, by contrast, is a subjective inquiry that centers on the impressions of a work's "intended audience," usually the general public. See Lyons P'ship, L.P. v. Morris Costumes, 243 F.3d 789, 801 (4th Cir.2001). So under the intrinsic prong, we analyze works as cohesive wholes, without distinguishing between protected and unprotected elements, just as the works' intended audiences likely would encounter them in the marketplace. See Universal Furniture, 618 F.3d at 437. We often have described intrinsic similarity as measuring the "total concept and feel" of the works in question. Id. at 436; Lyons, 243 F.3d at 801.

Copeland asks us to modify this approach in two respects. First, pointing to Ninth Circuit case law, see Apple Computer, Inc. v. Microsoft Corp., 35 F.3d 1435, 1442-43 (9th Cir.1994), Copeland suggests that we revisit our precedent and insist that comparison under the intrinsic prong, as well as the extrinsic, be confined to original elements and preceded by analytic dissection. We can put to one side whether the Ninth Circuit actually has adopted the rule that Copeland endorses, because our court has held just the opposite: In Universal Furniture, we joined the Second and Eighth Circuits in finding expressly that analytic dissection is inapplicable to the intrinsic analysis, because a work's intended audience "does not make th[e] distinction" between protectable and unprotectable elements and instead encounters a work "as one object." 618 F.3d at 437.

Even if we were free to reconsider that holding, Copeland has offered no compelling reason to do so. Indeed, as Bieber and Usher point out, our rule, allowing for comparison of entire works under the intrinsic prong, generally advantages rather than disadvantages copyright plaintiffs like Copeland, by broadening the grounds upon which a court may find intrinsic similarity. The district court committed no error by declining to engage in analytic dissection before conducting its inquiry into intrinsic similarity.

Second, Copeland urges us to adopt the rule from Shaw v. Lindheim, 919 F.2d 1353 (9th Cir.1990), under which intrinsic similarity is a question reserved for the trier of fact, and only the extrinsic similarity prong can be grounds for dismissal at the summary judgment and pleading stages. Because the intrinsic similarity inquiry turns on an inherently subjective appreciation of a work's tone and feel, Copeland argues, it is unfair to resolve it as a matter of law, with one judge's personal opinion trumping what could be overwhelming expert evidence showing the substantial similarity of two songs under the extrinsic prong.

We have not squarely addressed whether a district court may grant an infringement defendant's motion to dismiss, or motion for summary judgment, under the intrinsic prong alone.1 Nor have we analyzed the precise scope of the Shaw rule in the Ninth Circuit, or decided whether to adopt it in our own. We need not resolve those issues today, however, because in our view, even assuming that a motion to dismiss may be granted on the ground that no reasonable jury could find intrinsic similarity, the district court erred in doing so here. It is to that question that we now turn.

III.

This court reviews de novo a district court's decision to grant a motion to dismiss under Rule 12(b)(6), see Jackson, 775 F.3d at 177-78, and also conducts de novo the analysis of whether works are substantially similar, see Peters v. West, 692 F.3d 629, 632 (7th Cir.2012). Like the district court, we may examine all four of the songs at issue at the pleading stage, because all were "integral to and explicitly relied on in the complaint" and because Bieber and Usher "do not challenge [their] authenticity." Phillips v. LCI Int'l, Inc., 190 F.3d 609, 618 (4th Cir.1999).

A.

As noted above, intrinsic similarity is assessed from the perspective of a work's intended audience. See Universal Furniture, 618 F.3d at 435. That means that the first step in undertaking an analysis of intrinsic similarity is identifying the right audience. The district court concluded that the general public was the intended audience for the Copeland song, and we agree.

In Dawson, we clarified our intrinsic similarity analysis by introducing the "intended audience" formulation. Because a primary purpose of copyright law is to "protect[] a creator's market," we reasoned, the intrinsic similarity inquiry should be keyed to the impressions of the intended audience for a creator's work — the impressions that count for purposes of marketability. 905 F.2d at 734. So where the market for a given work consists of a discrete and specialized class, the reactions of a generic ordinary observer will not be particularly relevant. See id. But in most cases, we cautioned, the general public is in fact the intended audience, and "a court should be hesitant" to find otherwise. Id. at 737; see Lyons, 243 F.3d at 801.

Copeland argues that this case is the exception to the ordinary rule. According to Copeland, the intended audience for his song was not the general public but instead the "industry professionals" to whom he distributed his song by way of Sangreel. The "market" Copeland was trying to reach, in other words, was the Ushers of the world, and Copeland would be harmed if industry professionals believed his song was substantially similar to those of the defendants even if the general public saw no resemblance.

Like the district court, we are unpersuaded. It may be that Copeland intended to promote his music directly to industry professionals. But "[i]f ... industry professionals reject [Copeland's] song because it is too similar to the [d]efendants' songs, it would be because those companies fear that the public will find the songs to be overly similar." J.A. 252 (emphasis in original). There is a reason that the Dawson formulation uses the word "audience," rather than "buyer" or "recipient": Ultimate marketability is not always determined by the impressions of a first-hand purchaser or recipient, but may sometimes rest on the impressions of third parties — the work's actual "audience" — whose preferences the buyer or recipient has in mind when acquiring the work.

Our decision in Lyons illustrates the point. There, we considered whether the intended audience for a purple dinosaur costume resembling the character "Barney" from the television series Barney & Friends was the adult performers who would buy the costumes or the young children they sought to entertain. 243 F.3d at 802. We concluded that it was the children's reactions that mattered to the intrinsic similarity inquiry, because even though they were not themselves the intended purchasers of the costumes, it was their impressions (or misimpressions) that could lead adults to buy the infringing costumes. Id. at 802-03. Adults might discern differences between the two costumes, but if children could not, then there would be no reason for adults to insist on the original — with the result that the "knock-off" costumes would cut into Barney's market and the profits of Barney's owner. Id. at 803. The same reasoning applies here. Though industry professionals may have been the intended direct recipients of Copeland's music, the impressions that matter are those of the general public that constitutes the market for popular music — because, as Copeland admits, J.A. 252, those are the impressions that industry professionals would have in mind in choosing whether to do business with Copeland.

Again, this should come as no surprise. When we left open in Dawson the possibility that the intended audience for a choral arrangement of a spiritual song was more specialized than the general public and might be limited to choral directors, we also made clear that we were crafting a narrow rule for exceptional circumstances. We specifically distinguished the subject matter there from popular music, for which we noted approvingly that courts "routinely" apply the lay observer test. Dawson, 905 F.2d at 737. That is because the intended audience for popular music is usually an ordinary listener or, put differently, the general public. And indeed, the entire premise of Copeland's case is that his song is substantially similar to one that appears on a multi-platinum album by one of the world's most recognizable popular music stars. This is not a case about niche audience appeal, and there is no reason to think of the "intended audience" as anything other than the general public.

B.

Finally, we come to the question at the heart of this case: Whether the songs at issue, assessed from the perspective of the intended audience — here, the general public — and taking into account their "total concept and feel," Lyons, 243 F.3d at 801, are sufficiently intrinsically similar to give rise to a valid infringement claim. The district court answered in the negative, holding that no reasonable jury could find the requisite intrinsic similarity. But under the applicable de novo standard of review, see Peters, 692 F.3d at 632, we must listen for ourselves and come to our own conclusion. And because the general public typically encounters popular music songs by hearing them from start to finish, we undertake that analysis by listening to the songs in their entirety and side by side, to determine whether a reasonable jury could find that they are subjectively similar.

1.

As a preliminary matter, we should clarify that the "songs" to which we refer include all three of the defendants' versions of "Somebody to Love": the Usher demo version, the Bieber album version, and the Bieber-Usher remix version. At oral argument, Copeland suggested that each of those songs must be considered individually, and separately compared to the Copeland song. We disagree. In our view, the defendants' three songs are sufficiently similar to each other that they may be grouped together, and the same intrinsic analysis applied to all. If any one of them fails to meet the threshold for intrinsic similarity, then all of them do.

The Bieber album song and the Bieber-Usher remix are to our ears identical; the only difference we can hear is that Bieber is the only singer featured on the album song, whereas Usher provides lead vocals in the second verse and backing vocals elsewhere on the remix. On the Usher demo song, Usher is the only singer featured, and that song is in a different key than the others, presumably to accommodate his different vocal range. But the Usher demo song is otherwise in lock-step with the others down to minor details — everything from the lead singer's exclamation of "oh" in the introductory section to the distinctive synthesizer chords in the verses and the bass line in the pre-chorus.2 By the unscientific intrinsic standard, the three Bieber and Usher songs are not just substantially similar to one another; they are the same.

2.

We turn now to a comparison of the Copeland song with, collectively, the three Bieber and Usher songs. The district court acknowledged that the Usher and Bieber songs "have some elements in common" with the Copeland song. J.A. 253. But for the district court, what was dispositive was a significant difference in the overall "aesthetic appeal" of the respective songs. J.A. 254. We cannot agree. In our view, that analysis attaches too much weight to what the district court termed a difference in "mood" and "tone," and too little to similarities between the "element" of the songs — their choruses — that is most important.

First, if by "mood" and "tone" the district court meant genre, then we agree with this much: The Copeland song belongs to a different genre than the three Bieber and Usher songs. Though all fall under the same broad umbrella of popular music, the Copeland song is squarely within the R & B subgenre, while the Bieber and Usher songs would be labeled dance pop, perhaps with hints of electronica. Indeed, that difference is striking upon first listen, and at least as a linguistic matter, the very fact of these different genres might be thought to make the songs different in "concept and feel," Lyons, 243 F.3d at 801, or, in the words of the district court, in "aesthetic appeal," J.A. 254.

But as Bieber's counsel conceded at oral argument, while genre may be relevant to intrinsic analysis of musical works, it cannot be dispositive under copyright law. For if a difference in genre were enough by itself to preclude intrinsic similarity, then nothing would prevent someone from translating, say, the Beatles' songbook into a different genre, and then profiting from an unlicensed reggae or heavy metal version of "Hey Jude" on the ground that it is different in "concept and feel" than the original. From Copeland's perspective, it may be true that the "aesthetic appeal" of an R & B song is different, in some sense, than that of a dance pop song — but if there is going to be a dance pop version of his R & B "Somebody to Love," then it is his to record or to license, so that he can reap the full return on his creative efforts. Cf. Castle Rock Entm't, Inc. v. Carol Pub. Grp., Inc., 150 F.3d 132, 140 (2d Cir.1998) ("total concept and feel" analysis must take account of fact that works from different genres "must necessarily have a different concept and feel"). And by the same token, of course, were we to put too much stock in identity of genre at the intrinsic stage, we would risk deeming each successive work in a genre — whether it be R & B, ragtime, or bossa nova — an appropriation of the same-genre works that came before it.

Second, we do not doubt that the songs at issue here are in many respects dissimilar. And if substantial similarity were a purely quantitative inquiry, asking only whether the majority of the works in question overlapped, we would agree with the district court that no reasonable jury could find the requisite intrinsic similarity. For instance, while the Copeland song concludes with a repeated instrumental figure, the Bieber and Usher songs end more abruptly, after ad-libbed vocal lines. The Bieber and Usher songs include a post-chorus interval, with the lyric "I-I need somebody" sung in a syncopated manner, that has no equivalent in the Copeland song. And perhaps most significantly, the songs' verses feature different vocal melodies and beats3 as well as different lyrical content, with the Copeland verses lamenting the end of a relationship gone sour and the Bieber and Usher verses conveying the hope and optimism of the start of a relationship with an unidentified love interest. The district court may have had some or all of these in mind when it referred to differences in "mood, tone, and subject matter," J.A. 253, and we agree that taken numerically, the points of dissimilarity may well exceed the points of similarity.

But what that analysis fails to account for, we think, is the relative importance of these differences as compared to what the songs reasonably could be heard to have in common: their choruses. Even when quantitative majorities of two works bear little resemblance, courts routinely permit a finding of substantial similarity where the works share some especially significant sequence of notes or lyrics. See Swirsky v. Carey, 376 F.3d 841, 851 (9th Cir.2004) (overlap in first measure of chorus — seven total notes — enough to make pop songs substantially similar); Fisher v. Dees, 794 F.2d 432, 434 & n. 2 (9th Cir.1986) (similarity in first six measures of songs, amounting to twenty-nine seconds on a forty-minute album, enough to constitute appropriation of album); Elsmere Music, Inc. v. Nat'l Broad. Co., 482 F.Supp. 741, 744 (S.D.N.Y.), aff'd, 623 F.2d 252 (2d Cir. 1980) (four-note phrase accompanying lyrics "I love New York" protectable because it is "the heart of the composition"); Santrayll v. Burrell, No. 91 Civ. 3166, 1996 WL 134803, at *1-2 (S.D.N.Y. Mar. 25, 1996) (repetition of the phrase "uh-oh" four times in a distinctive rhythm for one measure is protectable). And we think it is clear that when it comes to popular music, a song's chorus may be the kind of key sequence that can give rise to intrinsic similarity, even when works differ in other respects.

It is the chorus — often termed the "hook," in recognition of its power to keep a listener coming back for more — that many listeners will recognize immediately or hear in their minds when a song title is mentioned. As the part of a song that is most often repeated and remembered, a chorus hook is important not only aesthetically but also commercially, where it may be central to a song's economic success. See, e.g., Gary Burns, A Typology of `Hooks' in Popular Records, 6 Popular Music 1 (1987) (cataloging characteristics and definitions of term "hook," and noting that "the hook is `what you're selling'" and that hooks are "the foundation of commercial songwriting, particularly hit-single writing"). From "Respect" by Aretha Franklin to "Seven Nation Army" by the White Stripes, the choruses or hooks of popular music songs are often disproportionately significant, relative to the amount of time or number of measures they occupy. See id. at 1 ("[V]irtually no hit record is without a bit of music or words so compelling that it worms its way into one's memory and won't go away.").

After listening to the Copeland song and the Bieber and Usher songs as wholes, we conclude that their choruses are similar enough and also significant enough that a reasonable jury could find the songs intrinsically similar. The most obvious similarity, of course, is the shared chorus lyric, mirrored in the songs' titles: "I [] need somebody to love." As Bieber and Usher point out, this phrase is common in popular music, appearing most famously in songs also titled "Somebody to Love" by psychedelic-rock band Jefferson Airplane and arena-rock band Queen, and common lyrical phrases generally are not copyrightable, see Peters, 692 F.3d at 635-36 (discussing rap songs' use of the maxim "what does not kill me, makes me stronger"). That might preclude consideration of this similarity under the extrinsic prong, where analysis is preceded by analytic dissection to determine which portions of a work are protectable. But as Bieber and Usher concede, under the intrinsic prong, we do not engage in analytic dissection. Instead, we examine the chorus's lyrics together with the accompanying music, taking the works in their entirety, as an ordinary musical listener would.

And when we listen to the choruses that way, and in the context of the entire songs, we hear the kind of meaningful overlap on which a reasonable jury could rest a finding of substantial similarity. It is not simply that both choruses contain the lyric "somebody to love"; it is that the lyric is delivered in what seems to be an almost identical rhythm and a strikingly similar melody. To us, it sounds as though there are a couple of points in the respective chorus melodies where the Bieber and Usher songs go up a note and the Copeland song goes down a note, or vice versa. In our view, however, a reasonable jury could find that these small variations would not prevent a member of the general public from hearing substantial similarity.4

We also conclude that the choruses of the Copeland song and the Bieber and Usher songs are sufficiently important to the songs' overall effect that they may be the basis for a finding of intrinsic similarity. In both the Copeland song and the Bieber and Usher songs, the singing of the titular lyric is an anthemic, sing-along moment, delivered at a high volume and pitch. Quite simply, it is "the heart of the composition[s]," Elsmere Music, 482 F.Supp. at 744, the most prominent and memorable part of the songs, and just the sort of significant sequence that courts have found sufficient to render musical works substantially similar. Whether a member of the general public could experience these songs primarily through their choruses and thus find them substantially similar, notwithstanding the differences catalogued above, is in our view a close enough question that it cannot be disposed of as a matter of law and should instead be decided by a jury.

IV.

In summary, we hold that a reasonable jury could find that the Copeland song and the Bieber and Usher songs are intrinsically similar. Because our holding is sufficient to dispose of this appeal, we decline to reach Copeland's other arguments. For the reasons set forth above, we vacate the judgment of the district court and remand the case for further proceedings.

VACATED AND REMANDED