FLAUM, Circuit Judge.
A recent Indiana statute prohibits most registered sex offenders from using social networking websites, instant messaging services, and chat programs. John Doe, on behalf of a class of similarly situated sex offenders, challenges this law on First Amendment grounds. We reverse the district court and hold that the law as drafted is unconstitutional. Though content neutral, we conclude that the Indiana law is not narrowly tailored to serve the state’s interest. It broadly prohibits substantial protected speech rather than specifically targeting the evil of improper communications to minors.
I. Background
A. Legislative Background
Indiana Code § 35^42^4-12 prohibits certain sex offenders from “knowingly or intentionally us[ing]: a social networking web site”
or “an instant messaging or
chat room program”
that “the offender knows allows a person who is less than eighteen (18) years of age to access or use the web site or program.” § 35-42-4-12(e) (violation constitutes a Class A misdemeanor but subsequent violations constitute Class D felonies). The law applies broadly to all individuals required to register as sex offenders under Indiana Code § 11-8-8, et seq., who have committed an enumerated offense. § 35—42—4—12(b)(1)— (2). The law does not differentiate based on the age of victim, the manner in which the crime was committed, or the time since the predicate offense. Subsection (f) provides an express defense if the individual did not know the website allowed minors or upon discovering it does, immediately ceased further use. § 35-42-4-12®. Subsection (a) exempts persons convicted of so-called Romeo and Juliet relationships where the victim and perpetrator are close in age and had a consensual relationship. § 35-42-4-12(a).
B. Procedural Background
1. John Doe’s Suit
In 2000, Doe was arrested in Marion County and convicted of two counts of child exploitation. Although he was released from prison in 2003 and is not on any form of supervised release, he must register as a sex offender on Indiana’s registry. And because child exploitation is an enumerated offense, section 35-42-4-12 prohibits Doe from using the covered websites and programs. Doe filed suit against the Marion County Prosecutor alleging the law violates his First Amendment rights (as incorporated under the Fourteenth Amendment).
The district court granted his request to proceed anonymously and later granted his motion to certify a class pursuant to Federal Rule of Civil Procedure 23(b)(2). It defined the class as:
all Marion County residents required to register as sex or violent offenders pursuant to Indiana law who are not subject to any form of supervised release and who have been found to be a sexually violent predator under Indiana law or who have been convicted of one or more of the offenses noted in Indiana Code § 35-42-4-12(b)(2) and who are not within the statutory exceptions noted in Indiana Code § 35-42-4-12(a).
Doe filed a motion for preliminary injunction, but the parties agreed it should be treated as a motion for a permanent injunction and decided after a full bench trial. The district court ordered as much.
See
Fed.R.Civ.P. 65(a)(2). The parties further agreed no additional discovery was required and there would be no live evidence at trial. Accordingly, the bench trial consisted of the introduction of four affidavits—two from Doe and two from social media experts—as well as arguments from counsel.
2. Lower Court Decision
After the bench trial, the district court upheld the law and entered judgment for
the defendant. It found the law implicates Doe’s First Amendment rights but held the regulation is narrowly tailored to serve a significant state interest and leaves open ample alternative channels of communication.
Although the court noted that the statute “captures considerable conduct that has nothing to do” with the state’s legitimate interest in protecting children from predators, it asserted “Doe never furnishes the Court with workable measures that achieve the same goal (deterrence and prevention of online sexual exploitation of minors) while not violating his First Amendment rights.” The district court reasoned that the law is narrowly tailored because it “only precluded] ... using web sites where online predators have easy access” to children, but “the vast majority of the internet is still at Mr. Doe’s fingertips.” The district court concluded that the law is not “substantially broader than necessary” because social networking sites without minors, e-mail, and message boards present alternative methods to communicate as Doe wished.
Doe offered another Indiana law that already prohibits online solicitation of children as evidence that the law is not narrowly tailored. The district court rejected this argument stating that .“the statutes serve different purposes”: “[o]ne set of statutes aims to
punish
those who have
already committed
the crime of solicitation,” while the “challenged statute, by contrast, aims to
prevent and deter
the sexual exploitation of minors by barring certain sexual offenders from entering a virtual world where they have access to minors.” (emphases in original). The district court concluded by noting the statute furthers the state’s “strong interest in ensuring that sex offenders do not place themselves in these potentially dangerous situations.”
On the issue of alternative channels of communication, the district court listed several social network alternatives, namely: “the ability to congregate with others, attend civic meetings, call in to radio shows, write letters to newspapers and magazines, post on message boards, comment on online stories that do not require a Facebook [account], email friends, family, associates, politicians and other adults, publish a blog, and use social networking sites that do not allow minors.”
Doe timely appeals this decision.
II. Discussion
We review a denial of a permanent injunction for abuse of decision, accepting all factual determinations unless they are clearly erroneous.
3M v. Pribyl,
259 F.3d 587, 597 (7th Cir.2001). However, this case presents a single legal question, which we review de novo.
The statute clearly implicates Doe’s First Amendment rights as incorporated through the Fourteenth Amendment. It not only precludes expression through the medium of social media,
see Cohen v. California,
403 U.S. 15, 24, 91 S.Ct. 1780, 29 L.Ed.2d 284 (1971) (“the usual rule [is] that governmental bodies may not prescribe the form or content of individual expression”), it also limits his “right to receive information and ideas,”
Stanley v.
Georgia,
394 U.S. 557, 564, 89 S.Ct. 1243, 22 L.Ed.2d 542 (1969);
see Procunier v. Martinez,
416 U.S. 396, 408-09, 94 S.Ct. 1800, 40 L.Ed.2d 224 (1974) (“[T]he addressee as well as the sender of direct personal correspondence derive[ ] from the First and Fourteenth Amendments a protection against unjustified governmental interference with the intended communication.”). The Indiana law, however, is content neutral because it restricts speech without reference to the expression’s content.
Turner Broadcasting Sys. v. FCC,
512 U.S. 622, 641-42, 114 S.Ct. 2445, 129 L.Ed.2d 497 (1994). As such, it may impose reasonable “time, place, or manner restrictions.”
Clark v. Comm. for Creative Non-Violence,
468 U.S. 288, 293, 104 S.Ct. 3065, 82 L.Ed.2d 221 (1984). To do so, the law must satisfy a variant of intermediate scrutiny—it must be “narrowly tailored to serve a significant governmental interest” and “leave open ample alternative channels for communication of the information.”
Ward v. Rock Against Racism,
491 U.S. 781, 791, 109 S.Ct. 2746, 105 L.Ed.2d 661 (1989). Because we conclude the law is not narrowly tailored, we need not reach the alternative channel inquiry.
The state initially asserts an interest in “protecting public safety, and specifically in protecting minors from harmful online communications.” Indiana is certainly justified in shielding its children from improper sexual communication. Doe agrees, but argues the state burdens substantially more speech than necessary to serve the intended interest. Indiana naturally counters that the law’s breadth is necessary to achieve its goal. On this point, the Supreme Court’s cases on narrow tailoring are instructive.
“A complete ban [such as the social media ban at issue] can be narrowly tailored, but only if each activity within the proscription’s scope is an appropriately targeted evil.”
See Frisby v. Schultz,
487 U.S. 474, 485, 108 S.Ct. 2495, 101 L.Ed.2d 420 (1988). In
FHsby,
the Court upheld an ordinance that prohibited picketing focused on a particular residence.
Id.
at 477, 108 S.Ct. 2495. The regulation sought to stop a recent pattern of abortion protesters that surrounded abortion doctors’ homes. The Court found that the state had a significant interest in protecting “residential privacy,” and a “complete prohibition” was the only way to further this interest.
Id.
at 484-86, 108 S.Ct. 2495. The Court reasoned that “the evil of targeted residential picketing ... is created by the medium of expression itself.”
Id.
at 487-88, 108 S.Ct. 2495 (internal quotations omitted). A ban on all picketing would have gone too far because only the focused residential protests threatened the state interest.
Id.
at 486, 108 S.Ct. 2495. Similarly, in
City of Los Angeles v. Taxpayers for Vincent,
a city ordinance prohibited posting signs on public property. 466 U.S. 789, 792, 104 S.Ct. 2118, 80 L.Ed.2d 772 (1984). The Court concluded these signs constituted “visual blight” and the regulation furthered the city’s legitimate interest in esthetic values.
Id.
at 805, 104 S.Ct. 2118. The Court upheld the complete ban reasoning the “substantive evil—visual blight—is not merely a possible by-product of the activity, but is created by the medium of expression itself.”
Id.
at 810, 104 S.Ct. 2118.
In contrast to
FHsby
and
Vincent,
the Supreme Court has invalidated bans on expressive activity that are not the substantive evil if the state had alternative means of combating the evil. In
Schneider v. Town of Irvington,
the Court struck down various blanket prohibitions against distributing handbills. 308 U.S. 147, 162-64, 60 S.Ct. 146, 84 L.Ed. 155 (1939). The laws in that case furthered a legitimate interest in preventing litter. But unlike the ordinances in
FHsby
and
Vincent,
the
expressive activity—handing paper to people in public—did not produce the evil. The recipients’ incidental decision to drop the paper did. As such, the Court required the cities to prevent littering by enforcing littering laws; they could not prohibit activity that might incidentally result in littering.
Id.
at 162-63, 60 S.Ct. 146. Similarly, in
Martin v. City of Struthers,
the Court invalidated an ordinance that prohibited all door-to-door distributions or solicitations because “[t]he dangers of distribution can so easily be controlled by traditional legal methods, leaving to each householder the full right to decide whether he will receive strangers as visitors.” 319 U.S. 141, 147, 63 S.Ct. 862, 87 L.Ed. 1313 (1943). For example, those who did not want to receive a stranger could post no trespassing signs, and states could permissibly punish those who disobeyed the warnings.
Id.
Turning to the Indiana statute, the state agrees there is nothing dangerous about Doe’s use of social media as long as he does not improperly communicate with minors. Further, there is no disagreement that illicit communication comprises a minuscule subset of the universe of social network activity. As such, the Indiana law targets substantially more activity than the evil it seeks to redress. Even the district court agreed with this sentiment, stating the law “captures considerable conduct that has nothing to do” with minors. Indiana prevents Doe from using social networking sites for fear that he might, subsequent to logging on to the website or program, engage in activity that Indiana is entitled to prevent. But like the states in
Schneider
and
Martin,
Indiana has other methods to combat unwanted and inappropriate communication between minors and sex offenders. For instance, it is a felony in Indiana for persons over twenty-one to “solicit” children under sixteen “to engage in: (1) sexual intercourse; (2) deviate sexual conduct; or (3) any fondling intended to arouse or satisfy the sexual desires of either the child or the older person.” Ind. Code § 35-42^-6 (it is also a felony for person between the ages of eighteen to twenty-one to solicit children under fourteen). A separate statute goes further. It punishes mere “inappropriate communication with a child” and communication “with the intent to gratify the sexual desires of the person or the individual,” Ind.Code § 35-42-4-13 (applies to persons over twenty-one communicating with children fourteen or younger). Significantly, both statutes have enhanced penalties for using a computer network and better advance Indiana’s interest in preventing harmful interaction with children (by going beyond social networks). They also accomplish that end more narrowly (by refusing to burden benign Internet activity). That is, they are neither over- nor under-inclusive like the statute at issue here.
In conducting this analysis, however, we must be most careful not to impose too high a standard on Indiana. The Supreme Court has continually reminded us that the state’s regulation “need not be the least restrictive or least intrusive means of’ combating the state’s legitimate interests,
Ward,
491 U.S. at 798, 109 S.Ct. 2746, and post-hoc analyses, like the one we are engaging in, are particularly susceptible to running afoul of this principle. At first glance, this standard seems in tension with language in
Frisby
noting a law must “target[] and eliminate[] no more than the exact source of the ‘evil’ it seeks to remedy,”
Frisby,
487 U.S. at 485, 108 S.Ct. 2495, and indeed, that is what the dissenters in
Ward
alleged,
see Ward,
491 U.S.
at 804-07, 109 S.Ct. 2746 (Marshall, J., dissenting). However,
Ward
scales back
Frisby
in a limited number of situations. On the one hand,
Ward
adds a quantitative component to the
Frisby
language by noting the law must not be
“substantially
broader than necessary.”
Id.
at 800, 109 S.Ct. 2746 (emphasis added). On the other hand,
Ward
also embodies an administr-ability exception in stating “the requirement of narrow tailoring is satisfied ‘so long as the [state interest] would be achieved less effectively absent the regulation.’ ”
Ward,
491 U.S. at 799, 109 S.Ct. 2746 (original quotations and alterations omitted). In other words, the Constitution tolerates some over-inclusiveness if it furthers the state’s ability to administer the regulation and combat an evil.
Hill v. Colorado
is illustrative. 530 U.S. 703, 120 S.Ct. 2480, 147 L.Ed.2d 597 (2000). There, the state, concerned about abortion protests, passed a statute that prohibited approaching individuals within a 100-foot radius of a healthcare facility “for the purpose of passing a leaflet or handbill to, displaying a sign to, or engaging in oral protest, education or counseling with [an]other person” on public property.
Id.
at 707, 120 S.Ct. 2480. The Court acknowledged that in furthering the state’s interest in providing unimpeded access to healthcare facilities and shielding patients from potentially traumatic encounters, the state’s blanket prohibition on approaching individuals would “sometimes inhibit a demonstrator whose approach in fact would have proved harmless.”
Id.
at 729, 120 S.Ct. 2480. But this over-inclusiveness was “justified by the great difficulty” in creating a law that “targets and eliminates no more than the exact source of the ‘evil.’”
See id.; Frisby,
487 U.S. at 485, 108 S.Ct. 2495. The Court hypothesized that the ideal statute would prove convoluted, potentially protecting “a pregnant woman from physical harassment with legal rules that focus exclusively on the individual impact of each instance of behavior, demanding in each case an accurate characterization (as harassing or not harassing) of each individual movement within the 8-foot boundary.”
Hill,
530 U.S. at 729, 120 S.Ct. 2480. Thus, the statute at issue in
Hill
was constitutional because (1) the prohibited expression that did not further the state interest was minimal, and (2) its inclusion stemmed from the difficulty in carving a rule that covered precisely the evil contemplated by the legislature.
The Indiana statute’s over-inclusiveness, however, cannot be justified by the admin-istrability concerns described in
Hill.
With little difficulty, the state could more precisely target illicit communication, as the statutes above demonstrate.
See
Ind.Code §§ 35-42-4-6; 35-42-4-13. To be sure, other sex-offender or social media statutes might present different administrability questions. For instance, a hypothetical law banning all communication between minors and sex offenders through social media burdens less speech but nevertheless creates problems. Such a law frees most expression from regulation but still prohibits substantial harmless speech— e.g., at a very basic level, it would prohibit conversations between a parent and child if the parent is a sex offender. But as additional exceptions make a law more precise, the over-inclusiveness concerns decrease until the
Hill
administrability concerns dominate. Where that point lies, however, is for another case.
The district court also suggested the law was narrowly tailored to serve purposes different from the existing solicitation and communication laws. It stated the existing laws “aim[ ] to
punish
those who have
already committed
the crime of solicitation,” while the “challenged statute, by contrast, aims to
prevent and deter
the sexual exploitation of minors by barring certain sexual offenders from entering a virtual world where they have access to
minors” (emphases in original). The state continues this argument on appeal. The immediate problem with this suggestion is that all criminal laws generally “punish” those who have “already committed” a crime. The punishment is what “prevent[s] and deter[s]” undesirable behavior. Thus, characterizing the new statute as preventative and the existing statutes as reactive is questionable. The legislature attached criminal penalties to solicitations in order to prevent conduct in the same way decade-long sentences are promulgated to deter repeat drug offenses. Perhaps the state suggests that prohibiting social networking deprives would-be solicitors the opportunity to send the solicitation in the first place. But if they are willing to break the existing anti-solicitation law, why would the social networking law provide any more deterrence? By breaking two laws, the sex offender will face increased sentences; however, the state can avoid First Amendment pitfalls by just increasing the sentences for solicitation— indeed, those laws already have enhanced penalties if the defendant uses a computer network.
See
Ind.Code §§ 35-42-4-6(b)(3); 35^12^t-13(c).
The state also makes the conclusory assertion that “the State need not wait until a child is solicited by a sex offender on Facebook.” Of course this statement is correct, but the goal of deterrence does not license the state to restrict far more speech than necessary to target the prospective harm. Moreover, the state never explains how the social network law allows them to avoid “waiting.” “That the [state’s] asserted interests are important in the abstract does not mean ... that [its regulation] will in fact advance those interests.”
See Turner,
512 U.S. at 664, 114 S.Ct. 2445. The state “must do more than simply posit the existence of the disease sought to be cured,” and “the regulation [must] in fact alleviate these harms in a direct and material way.”
Id.
(internal quotations omitted). The state bears this burden, and it does not explain how the law furthers this interest.
Despite the infirmity of the statute in this case, we do not foreclose the possibility that keeping certain sex offenders off social networks advances the state’s interest in ways distinct from the existing justifications. For example, perpetrators may take time to seek out minors they will later solicit. This initial step requires time spent on social networking websites before the solicitation occurs. In the future, the state may argue that prohibiting the use of social networking allows law enforcement to swoop in and arrest perpetrators before they have the opportunity to send an actual solicitation. This argument remains speculative. And it is uncertain whether such a law could advance this interest without burdening a “substantially broader” than necessary group of sex offenders who will not use the Internet in illicit ways.
See Ward,
491 U.S. at 800, 109
S.Ct. 2746. But perhaps such a law could apply to certain persons that present an acute risk—those individuals whose presence on social media impels them to solicit children. Currently, the state presents no evidence that covered individuals present this sort of risk. We speculate only to make clear that this decision should not be read to limit the legislature’s ability to craft constitutional solutions to this modern-day challenge.
The district court also cited
Kansas v. Hendricks,
521 U.S. 346, 117 S.Ct. 2072, 138 L.Ed.2d 501 (1997), for the proposition that the state may “ensur[e] that sex offenders do not place themselves in these potentially dangerous situations.”
However,
Hendricks
is inapposite. It rejected a substantive due process challenge to a Kansas statute permitting the state to commit persons that are “likely to engage in sexually violent behavior.”
Id.
at 351, 117 S.Ct. 2072 (quoting Kan. Stat. Ann. § 59-29a01 (1994)). The case did not present a First Amendment challenge. Notwithstanding, the
Hendricks
statute proceeds cautiously and is far more targeted than the Indiana statute here. The
Hendricks
law provided respondents a number of procedural safeguards before a trial in which the state had to prove beyond a reasonable doubt that petitioners had a “mental abnormality” that made them “likely to engage in the predatory acts of sexual violence.”
Id.
at 352, 117 S.Ct. 2072. They were also entitled to regular review after their confinement to ensure they still met the act’s criteria.
Id.
at 353, 117 S.Ct. 2072. We do not suggest that these procedures are a prerequisite to a First Amendment deprivation;
Hen
dricks-style civil commitment presents a far greater deprivation of liberty than banning social networking. But
Hendricks
nevertheless illuminates the imprecision of the Indiana statute. Unlike the individualized assessment that ensured each respondent was “likely” to commit the redressa-ble evil, the Indiana legislature imprecisely used the sex offender registry as a universal proxy for those likely to solicit minors. There may well be an appropriate proxy, but the state has to provide some evidence, beyond conclusory assertions, to justify the regulation.
This case also differs from our decision in
Doe v. City of Lafayette,
377 F.3d 757 (2004) (en banc). That case involved an individual with an extensive history of sex offenses against children, who admitted he was going to the city parks “cruising” and “looking” for children.
Id.
at 759-60. The city issued a unilateral order banning the plaintiff from the city parks without a hearing.
Id.
at 760. Unlike this case, the regulation did not implicate the First Amendment so we upheld it under the deferential rational basis review.
Id.
at 764, 773. Other laws restricting sex offenders’ proximity to schools or parks have been similarly upheld under rational basis review because courts have found they do not implicate the First Amendment or involve a fundamental right.
See, e.g., Smith v. Doe,
538 U.S. 84, 123 S.Ct. 1140, 155 L.Ed.2d 164 (2003) (holding that the Alaska Sex Offender Registration Act did not violate the Ex Post Facto Clause);
Conn. Dep’t of Pub. Safety v. Doe,
538 U.S. 1, 123 S.Ct. 1160, 155 L.Ed.2d 98 (2003) (holding that the public disclosure provision of Connecticut’s sex offender registration law did not violate the Due Process Clause);
Doe v. Miller,
405 F.3d 700 (8th
Cir.2005) (holding residency restriction within two thousand feet of school or child care facility constitutional under rational basis review).
Finally, this opinion should not be read to affect district courts’ latitude in fashioning terms of supervised release, 18 U.S.C. § 3583(a) (“The court, in imposing a sentence to a term of imprisonment for a felony or a misdemeanor, may include as a part of the sentence a requirement that the defendant be placed on a term of supervised release after imprisonment[.]”), or states from implementing similar solutions. Our penal system necessarily implicates various constitutional rights, and we review sentences under distinct doctrines. Terms of supervised release, for instance, must be “reasonably related to the [sentencing] factors” and “involve[ ] no greater deprivation of liberty than is reasonably necessary.” § 3583(d)(l)-(2). Thus, in assessing the need for incapacitation,
see
§ 3553(a)(2)(C), a court could conceivably limit a defendant’s Internet access if full access posed too high a risk of recidivism.
United States v. Scott,
316 F.3d 733, 736-37 (7th Cir.2003). The alternative to limited Internet access may be additional time in prison, which is surely more restrictive of speech than a limitation on electronics. This option is not without limits,
see United States v. Holm,
326 F.3d 872, 878-79 (7th Cir.2003) (holding total Internet ban was too broad and compiling similar cases from other circuits), but terms of supervised release or parole may offer viable constitutional alternatives to the blanket ban—imposed outside the penal system— in this case.
We conclude by noting that Indiana continues to possess existing tools to combat sexual predators. The penal system offers speech-restrictive alternatives to imprisonment. Regulations that do not implicate the First Amendment are reviewed only for a rational basis. The Constitution even permits civil commitment under certain conditions. But laws that implicate the First Amendment require narrow tailoring. Subsequent Indiana statutes may well meet this requirement, but the blanket ban on social media in this case regrettably does not.
III. Conclusion
For the foregoing reasons, we Reverse the district court’s decision, and Remand with instructions to enter judgment in favor of Doe and issue the injunction.