In the

    United States Court of Appeals
                For the Seventh Circuit
                    ____________________
No. 22-3170
MADELINE KRASNO,
                                                Plaintiff-Appellant,
                                 v.

JENNIFER MNOOKIN, et al.,
                                             Defendants-Appellees.
                    ____________________

        Appeal from the United States District Court for the
                   Western District of Wisconsin.
        No. 21-cv-99 — Stephen L. Crocker, Magistrate Judge.
                    ____________________

   ARGUED SEPTEMBER 21, 2023 — DECIDED AUGUST 1, 2025
                ____________________

   Before EASTERBROOK, ROVNER, and PRYOR, Circuit Judges.
    PRYOR, Circuit Judge. After the University of Wisconsin–
Madison hid Madeline Krasno’s comments on its social media
posts and restricted Krasno’s Instagram account, Krasno
sued, alleging a violation of her First Amendment rights.
Krasno claimed she earned the University’s disapproval for
calling attention to what she believes to be animal abuse in the
University’s primate testing facilities. The University de-
fended its social media moderation decisions on the ground
2                                                      No. 22-3170

that it was hiding “off-topic” comments. The district court en-
tered summary judgment for the University and denied sum-
mary judgment for Krasno. Relevant here, the district court
found the University’s comment threads were nonpublic fo-
rums and upheld the University’s social media moderation
decisions as reasonable and viewpoint neutral. It further held
that Krasno lacked standing to seek an injunction against the
University’s ongoing use of keyword filters to implement its
rule against off-topic comments.

    We disagree. We find that Krasno has standing to bring
this as-applied challenge. We also conclude that the interac-
tive comment threads attached to the University’s posts are
limited public forums, such that speech restrictions imposed
by the University on the comment threads must be reasonable
and viewpoint neutral. Because the University’s ill-defined
off-topic comment rule is neither reasonable nor viewpoint
neutral, we find it unconstitutional under the First Amend-
ment.

                      I.    BACKGROUND

    “We review de novo a district court’s decision on cross-
motions for summary judgment, construing all facts and
drawing all reasonable inferences in favor of the party against
whom the motion under consideration was filed.” Hess v. Bd.
of Trs. of S. Ill. Univ., 

 (7th Cir. 2016). The fol-
lowing facts are undisputed.
    A. Factual Background
       1. Social Media Platforms
   “Social-media platforms … have gone from unheard-of to
inescapable.” Moody v. NetChoice, LLC, 

, 716
No. 22-3170                                                  3

(2024). They now represent one of the “most important places
… for the exchange of views” in modern life. Packingham v.
North Carolina, 

 (2017). The University of Wis-
consin–Madison operates social media accounts on Instagram
and Facebook. Instagram and Facebook are social media plat-
forms that allow people and institutions to interact with one
another, including by sharing photos, videos, and messages.
Once a person or entity creates an account on either platform,
the platform generates an account-specific webpage where
the account holder can “post” (i.e., publish) content. When the
account holder posts content—whether in the form of photos,
videos, or text—the content is automatically accompanied by
an interactive space immediately adjacent to the content,
where other accounts can write comments and reply to com-
ments from others. This space is known as a “comment
thread.”
   Comment threads on Instagram and Facebook are at-
tached to, but separate from, an account holder’s original
post. Within a comment thread, a commenter’s unique Insta-
gram or Facebook username appears immediately prior to the
comment. In other words, a comment is attributed to the com-
menter.
    Instagram and Facebook offer tools for an account holder
to limit how commenters can interact in the comment threads
attached to the account holder’s posts. For instance, on Insta-
gram, an account holder can disable the comment thread en-
tirely for the account’s posts. Additionally, an Instagram ac-
count holder can “restrict” another account. If Account A “re-
stricts” Account B, then all of Account B’s comments will be
automatically hidden from the comment threads attached to
Account A’s posts, unless Account A affirmatively approves
4                                                                No. 22-3170

a specific comment. Restricted accounts are not notified that
they have been restricted.
   On Facebook, an account holder can manually hide from
public view, or delete entirely, specific comments of others. A
hidden comment can be seen only by the commenter and the
commenter’s “friends,” 1 while a deleted comment cannot be
seen by anyone.
    Finally, both Instagram and Facebook allow account hold-
ers to implement “keyword filters” that automatically hide
any comments containing specific words or phrases the ac-
count holder selects.
           2. The University’s Social Media Accounts and Comment
              Moderation
    The University’s Instagram and Facebook accounts bear
the username “UWMadison.” The University uses its ac-
counts to communicate official announcements, events, and
policies; they are considered “the official voice of the Univer-
sity.” The University chose to make both accounts “public,”
which by default allows anyone with an Instagram or Face-
book account to view and comment on the University’s posts.
The University’s audience includes current and prospective
students, faculty, staff, alumni, community members, and the
public. 2 To engage its audience, the University has created
more than 2,400 Instagram posts covering a wide range of




1
 On Facebook, accounts can “friend” one another, thereby gaining access
to one another’s non-public posts.
2
    (Dkt. 40, Joint Statement of Undisputed Facts, ¶¶ 67, 76).
No. 22-3170                                                     5

topics. It also frequently shares information, announcements,
and anecdotes on its Facebook account.
    The University’s Office of University Communications
oversees these social media accounts, including by moderat-
ing comments on the University’s posts. John Lucas is the As-
sistant Vice Chancellor for University Communications; Mike
Klein is the Director for News Content and Editorial Projects;
and Nate Moll is a Social Media Manager. Klein and Moll
make daily moderation decisions regarding comments on the
University’s posts, while Lucas oversees compliance with the
University’s social media guidelines for comment modera-
tion.
    The University has a public “Social Media Statement”
which it published to its website before this litigation began.
It provides:
       While UW–Madison does not regularly review
       content posted to social media sites, it shall have
       the right to remove any content for any reason,
       including but not limited to, content that it
       deems threatening, profane, obscene, a viola-
       tion of intellectual property rights or privacy
       laws, off-topic, commercial or promotion of or-
       ganizations or programs not related to or affili-
       ated with the university, or otherwise injurious
       or illegal. Users are fully responsible for the con-
       tent they load on any of UW–Madison’s social
       media sites.
    After the start of this litigation, the University distributed
interim guidance to its social media staff. The guidance in-
structs that the University’s “social media managers may
6                                                              No. 22-3170

engage in content moderation of social media pages based on
one criterion: whether posted content is on vs. off topic.” It
provides that social media managers have discretion to “re-
move posts that are unrelated to the topic or purpose of the
page.” As an example, it explains that “while the social media
manager of a page dedicated to UW–Madison’s animal re-
search programs may hide a post stating ‘Taylor Swift is the
B0mB,’ he or she cannot hide a post that states, ‘Taylor Swift
agrees that all universities should stop torturing animals by
using them for research. This includes UW–Madison’s animal
research programs!’”
    In addition to manual moderation, the University employs
keyword filters on its Instagram and Facebook accounts to en-
force the University’s social media guidelines. The University
programmed its keyword filter on Instagram to automatically
hide comments that include the following words or phrases:
#freebabycocoa, #releasecornelius, @peta, Cornelius, 3
WNPRC, 4 abusing, animal testing, biden, cruelty, kill animals,
killers, lab, monsters, rot in hell, shame on, testing on animals,
testing on cats, tests on cats, torture, torturing, trump, vivisec-
tion, you guys are sick. Its keyword filter on Facebook auto-
matically hides comments that include, among other phrases:
peta, macaques, animal laboratories, testing on animals, ani-
mal testing, barbaric, wnprc, primate, primates, experiment-
ing on, research animals, cruelty, torturing, torture, vivisec-
tion, monkeys, and experimenting.




3
    Cornelius is one of the primates at the University’s research labs.
4
    “WNPRC” stands for the Wisconsin National Primate Research Center.
No. 22-3170                                                  7

    Much of the University’s content moderation decisions are
left to its employees’ discretion. The University has no formal
practice for adding or removing terms from its keyword fil-
ters; the process is “situation dependent” and occurs on an
“as-needed basis.” Nor has it provided written guidance to
help staff decide whether to “ban” a user on social media; it
instead does so on a “case-by-case basis.” Similarly, Moll,
Klein, and Lucas have discretion to hide, respond to, or ignore
comments left by users. The University does not conduct au-
dits or compliance reviews regarding the application and en-
forcement of its social media guidelines.
      3. Krasno’s Advocacy and the University’s Response
    While a student at the University of Wisconsin–Madison,
Krasno worked as a primate caretaker at the Harlow Center
for Biological Psychology. Through the Harlow Center and
the Wisconsin National Primate Research Center, the Univer-
sity conducts invasive research on animals, including nonhu-
man primates.
   Krasno’s background working for the Harlow Center in-
spired her to become an animal rights advocate and educate
others about the treatment of primates. She does this, in part,
by using social media to highlight her experiences at the Har-
low Center.
    In September 2020, Krasno began commenting on the Uni-
versity’s social media posts to advocate for animal rights. For
instance, on September 18, 2020, she commented on a Univer-
sity Instagram post about dairy cows at the University’s Dairy
Cattle Center: “stop exploiting animals. Get with the future
and the future is consistent anti-oppression. Shut down the
labs and eat plants!” She also responded to another user’s
8                                                 No. 22-3170

comment, explaining that she “used to work in one of their
labs.” Krasno’s comments were hidden. Another user com-
mented on the same post, stating, “Hopefully I will be admit-
ted next year, fingers crossed!” That comment was not hid-
den.
    Ten days later, Krasno commented on a University Insta-
gram post about a new recreation center. Krasno wrote,
“Thanks for continuing to delete my comments … I will con-
tinue to share the truth about what it was like working in one
of your primate research labs and advocate for their closure.
… [T]oday is a great day to shut down the primate research
labs!” This comment, too, was hidden. At the time Krasno
posted, the post had 179 other comments.
    At the end of September 2020, the University restricted
Krasno’s Instagram account, thereby automatically hiding her
comments on the University’s posts. When questioned about
the account restriction, Moll explained that he had noticed a
“consistent pattern of off-topic comments” by Krasno. By the
time Krasno filed suit, Krasno was the only person the Uni-
versity had restricted on Instagram. Moll removed the re-
striction on Krasno’s Instagram account in late January 2021.
    While Krasno’s Instagram account was restricted, she
commented on five of the University’s posts; each comment
was hidden automatically due to the restriction. Four of the
posts she commented on concerned the University’s mascot,
COVID-19 efforts and measures, and Thanksgiving travel.
Krasno’s comments on these four posts urged the University
to stop testing on monkeys or to close its primate labs. The
fifth post concerned a dog being treated for cancer at the Uni-
versity’s veterinary hospital. On that post, Krasno com-
mented on December 22, 2020, “It is really quite hypocritical
No. 22-3170                                                     9

the compassion shown to this dog while thousands of animals
languish in laboratories at @uwmadison. I really wish you
would acknowledge this and do something about it.” None of
these comments were unhidden after the University lifted
Krasno’s account restriction in January 2021, though the Uni-
versity does not dispute that the December 22 comment was
on topic.
    The University also manually moderated Krasno’s com-
ments on Facebook. For example, on December 9, 2020,
Krasno commented on a University Facebook post about
proud alumni, stating, “University of Wisconsin–Madison,
are you really proud of all your graduates or just the ones who
don’t object to your barbaric treatment of monkeys in your
research labs?” The University manually hid Krasno’s com-
ment for being off topic.
    On December 13, 2020, Krasno commented on a Univer-
sity Facebook post about winter commencement and the fu-
ture, remarking, “I wish you would think about the future in
new ways too and stop testing on monkeys. Working in one
of those labs during college showed me how important it is
that we create a just world. Animal research is not the way to
accomplish this.” This comment, which included a word on
the University’s keyword filter list (“monkeys”), was hidden.
     Once Krasno discovered the University’s use of keyword
filters, she began modifying the spelling of certain words to
avoid the filters. On May 9, 2021, she responded to a Univer-
sity Instagram post about Mother’s Day, commenting “What
about all the mothers you have in cages on campus? Celebrat-
ing them too after ripping their babies away from them? A n
l m a l t e s t l n g is cruel.” On March 24, 2022, in response to
a University Facebook post announcing a “Cool Science
10                                                        No. 22-3170

Image contest,” Krasno commented, “Considering much of
the ‘science’ done at the university is research without con-
sent AKA t e s t l n g o*n a n l m a l s, let’s see some footage
[of] the a l n m a l s in your l a b s. After all, y’all are all about
transparency right?” Both comments remain publicly visible.
     B. Procedural History
    On February 10, 2021, Krasno sued the University of Wis-
consin–Madison’s Board of Regents alongside Jennifer
Mnookin,5 the University’s Chancellor; Charles Hoslet, the
Vice Chancellor for University Relations; John Lucas, the As-
sistant Vice Chancellor for University Communications; Mike
Klein, the Director for News Content and Editorial Projects;
and Nate Moll, a Social Media Manager. Krasno alleged the
University violated her First Amendment rights by censoring
her speech in the University’s comment threads. Krasno
sought injunctive relief, declaratory relief, and nominal dam-
ages.
   Each side moved for summary judgment. The parties dis-
puted, among other things, the nature of the forum the Uni-
versity created in its comment threads and the corresponding
standard to apply in assessing the University’s moderation
decisions. They also disputed whether Krasno had standing
and whether certain relief Krasno requested was barred by
the Eleventh Amendment. While briefing was ongoing,
Krasno voluntarily dismissed the Board of Regents as a de-
fendant, leaving only the individual defendants in the case.

5
  Krasno initially sued Rebecca Blank, then-Chancellor of the University
of Wisconsin–Madison. When Jennifer Mnookin became Chancellor dur-
ing the pendency of this lawsuit, the district court substituted Mnookin
for Blank under Federal Rule of Civil Procedure 25(d).
No. 22-3170                                                11

    The district court entered summary judgment for the de-
fendants and denied summary judgment for Krasno. It began
by applying the Supreme Court’s forum jurisprudence. Find-
ing nonpublic and limited public forums substantively equiv-
alent, the court determined that the comment threads at-
tached to the University’s posts were nonpublic forums, such
that any restrictions to speech were required to be reasonable
and viewpoint neutral. The court then held the University’s
off-topic comment rule satisfied this standard.
    Next, the district court concluded that the Eleventh
Amendment barred Krasno’s official capacity claims against
the individual defendants insofar as they related to the deci-
sion to restrict her Instagram account and hide her December
9, 2020, Facebook comment. In particular, the district court
determined that the Ex parte Young limitation on sovereign
immunity did not save her claims because any violation was
no longer ongoing. As for Krasno’s request to enjoin the Uni-
versity’s use of keyword filters, the district court held that
Krasno lacked standing because she did not have a right “to
comment about animal testing regardless of the topic of the
University’s posts” and had not shown she was likely to be
harmed by the keyword filter in the future. Finally, the court
concluded that Krasno’s individual capacity claims failed be-
cause the defendants were entitled to qualified immunity.
   Krasno timely appealed.
                     II.   ANALYSIS
   On appeal, Krasno challenges the dismissal of her official-
capacity claims against the individual defendants. Krasno ar-
gues the University’s off-topic comment rule—as imple-
mented through keyword filters, an Instagram account
12                                                     No. 22-3170

restriction, and manual moderation of individual comments
pursuant to the University’s Social Media Statement and in-
terim guidance—is unconstitutional as applied to her.
   We first address the jurisdictional issues relevant to this
appeal. We then turn to the merits of Krasno’s First Amend-
ment challenge.
     A. Jurisdictional Issues
        1. Standing
    Article III of the United States Constitution limits the judi-
cial power to deciding “cases” and “controversies.” U.S.
CONST. art. III. § 2. To establish a case or controversy, the
plaintiff must demonstrate a “personal stake” in the matter.
TransUnion LLC v. Ramirez, 

 (2021). In other
words, the plaintiff must establish standing to sue. 

 Stand-
ing requires a plaintiff to show “(i) that [s]he suffered an in-
jury in fact that is concrete, particularized, and actual or im-
minent; (ii) that the injury was likely caused by the defendant;
and (iii) that the injury would likely be redressed by judicial
relief.” 

 (citing Lujan v. Defs. of Wildlife, 

, 560–61
(1992)).
    The district court concluded that Krasno lacked standing
to seek injunctive relief against the University’s continued use
of keyword filters. In the district court’s view, Krasno was un-
able to show she faced an actual and imminent injury because
she had no “constitutionally-guaranteed right to comment
about animal testing regardless of the topic of the University’s
posts.” Nor could she show redressability, in the district
court’s view, because the possibility that Krasno would make
an off-topic comment that would be caught by the University’s
keyword filters was speculative.
No. 22-3170                                                     13

    But the district court erred by collapsing its standing in-
quiry into its assessment of the merits. Standing does not re-
quire a plaintiff to prove success on the merits. Aurora Loan
Servs., Inc. v. Craddieth, 

 (7th Cir. 2006). Ra-
ther, it requires demonstrating “a colorable claim” to a right
that the defendant allegedly infringed. 

 The conclusion that
Krasno lacked standing was based on the court’s merits de-
termination that Krasno had no right to comment on the Uni-
versity’s social media posts. But whether Krasno in fact has a
right to comment on the University’s posts cannot answer the
question of whether she raised a colorable claim to relief.
“Otherwise every losing suit would be dismissed for lack of
jurisdiction.” Owsley v. Gorbett, 

 (7th Cir.
2020).
    Krasno has standing to seek relief against the University’s
moderation of her comments. She claims the University is bar-
ring her speech from the interactive portions of its social me-
dia pages by enforcing its off-topic comment rule in a discrim-
inatory manner. That is a cognizable injury. See Members of
City Council of City of Los Angeles v. Taxpayers for Vincent, 

 (1984). The University continues to implement
its off-topic rule through keyword filters that block the speech
Krasno wants to engage in, such that the alleged harm is on-
going and could warrant injunctive relief. See Backpage.com,
LLC v. Dart, 

, 238–39 (7th Cir. 2015).
    Krasno has demonstrated that her comments have been
suppressed by the University’s implementation of its off-topic
comment rule in the past, that she wants to continue comment-
ing in the future, and that the University’s filters have largely
remained stagnant over time. As such, she has shown that the
off-topic comment rule poses an ongoing threat of actual or
14                                                     No. 22-3170

imminent injury. See Bell v. Keating, 

 (7th Cir.
2012); Lawson v. Hill, 

 (7th Cir. 2004). Such
injury is redressable because an injunction against the use of
the keyword filters would remedy the injury allegedly caused
by the filters blocking Krasno’s speech. See Corr. Servs. Corp.
v. Malesko, 

 (2001) (“[I]njunctive relief has long
been recognized as the proper means for preventing entities
from acting unconstitutionally.”).
       2. Sovereign Immunity
    As confirmed by the Eleventh Amendment, states are sov-
ereigns and are therefore immune from suit. Va. Off. for Prot.
and Advoc. v. Stewart, 

, 253 (2011). While “the Elev-
enth Amendment is not ‘jurisdictional’ in the same way as Ar-
ticle III’s case-or-controversy requirement,” it is “‘jurisdic-
tional’ in the sense that a defendant invoking its sovereign im-
munity deprives a federal court of jurisdiction over the claims
against that defendant.” McHugh v. Ill. Dept. of Trans., 

, 533 (7th Cir. 2022).
    The doctrine of Ex parte Young provides an “important
limit on the sovereign-immunity principle.” Stewart, 563 U.S.
at 254. It “allows suits … for declaratory or injunctive relief
against state officers in their official capacities.” Reed v. Goertz,

 (2023). In particular, private parties may sue
state officials in their official capacities “for prospective relief
to enjoin ongoing violations of federal law.” Sherwood v. Mar-
chiori, 

, 693 (7th Cir. 2023) (citation and quotation
omitted).
   “In determining whether the doctrine of Ex parte Young
avoids an Eleventh Amendment bar to suit, a court need only
conduct a straightforward inquiry into whether [the]
No. 22-3170                                                           15

complaint alleges an ongoing violation of federal law and
seeks relief properly characterized as prospective.” Verizon
Md., Inc. v. Pub. Serv. Com’n of Md., 

 (2002)
(alteration in original) (citation and quotation omitted).
    Ex parte Young permits Krasno’s request for an injunction
against the University’s use of keyword filters, which quali-
fies as prospective relief against an allegedly ongoing viola-
tion of federal law. Likewise, Krasno may seek injunctive re-
lief against the University’s ongoing implementation of its
off-topic comment rule through the manual moderation of its
comment threads. The sources of the University’s rule—i.e.,
its Social Media Statement and interim guidance—continue to
afford moderators discretion to manually moderate com-
ments.
    The district court, however, found barred by sovereign
immunity Krasno’s request for a declaration that the re-
striction of her Instagram account and hiding of a December
2020 Facebook comment violated the First Amendment. 6 As it
observed, the University lifted the restriction on Krasno’s In-
stagram account in January 2021 such that any violation of
federal law was not ongoing; rather, the restriction occurred
only in the past. It also reasoned that the University’s act of
hiding the December 2020 Facebook comment was past con-
duct.



6
  Although the district court and Krasno frame the University’s action on
December 9, 2020, as the deletion of a Facebook comment, the record re-
flects that the Facebook comment was hidden by the University; it was not
deleted. For accuracy, we refer to the December 9 comment as being hid-
den, not deleted.
16                                                   No. 22-3170

    We agree with the district court that Ex parte Young does
not allow Krasno to seek relief for either event. The Instagram
account restriction was lifted in January 2021. Krasno is there-
fore free to comment on the same posts on which she previ-
ously commented; this time, her comments will not be auto-
matically hidden pursuant to any account restriction. Simi-
larly, the hiding of a Facebook comment in December 2020
occurred in the past. Because these actions are not “continu-
ing conduct” that must be “change[d] to comply with federal
law,” they do not amount to alleged ongoing violations of fed-
eral law and Ex parte Young does not apply. Sherwood, 76 F.4th
at 694 (quoting Watkins v. Blinzinger, 

 (7th Cir.
1986)).
    The relief Krasno sought in relation to the account re-
striction and hiding of her December 2020 Facebook comment
also cannot be “properly characterized as prospective.” Veri-
zon Md., Inc., 

. In particular, Krasno sought an
order declaring that “Defendants’ practice of restricting [her]
Instagram account” and “deletion of [her] comments from the
UW–Madison Facebook page … [are] unconstitutional.” But
an order declaring that these past actions violated the Consti-
tution would not require the University to act prospectively.
P.R. Aqueduct and Sewer Auth. v. Metcalf & Eddy, Inc., 

 (1993) (Ex parte Young “does not permit judgments
against state officers declaring that they violated federal law
in the past”). And while Krasno proposes on appeal that she
could also be entitled to an injunction requiring the University
to “unhide” the comments that were hidden, she does not
identify where she requested this form of injunctive relief be-
fore the district court. As far as we can tell, the prospective
action she sought from the University concerned its ongoing
implementation of its off-topic comment rule through
No. 22-3170                                                    17

keyword filters and discretionary manual moderation deci-
sions. She cannot advance new claims for relief for the first
time on appeal. Gerlach v. Rokita, 

 (7th Cir.
2024).
   B. First Amendment Analysis
    Having established our jurisdiction and the relief Krasno
may seek in view of the Eleventh Amendment, we turn to the
merits of Krasno’s First Amendment challenge to the Univer-
sity’s off-topic rule, as applied to her. The First Amendment
provides, in relevant part, that “Congress shall make no law
… abridging the freedom of speech.” U.S. CONST. amend. I. It
applies to the states (and, by extension, state universities) via
the Fourteenth Amendment’s Due Process Clause. Gitlow v.
New York, 

 (1925) (states); Widmar v. Vincent,

, 267–69 (1981) (state universities).
    The parties agree the comment threads attached to the
University’s social media posts are government-controlled
property. When the government restricts private speech on
property it controls, the Supreme Court’s forum caselaw
guides our assessment of whether the restrictions are consti-
tutional. Hague v. CIO, 

, 515–16 (1939); Perry Educ.
Ass’n v. Perry Loc. Educators’ Ass’n, 

, 45–46 (1983);
Good News Club v. Milford Cent. Sch., 

, 106–112
(2001); Christian Legal Soc’y v. Martinez, 

, 679 n. 11
(2010). By contrast, “forum analysis is misplaced” when the
government is “speaking on its own behalf” on its property,
Walker v. Texas Div., Sons of Confederate Veterans, Inc., 

 (2015), as the government is “entitled … to take a po-
sition” when it speaks, 

. Accordingly, before under-
taking any forum analysis, we must assure ourselves that
comments within the comment threads are private speech
18                                                 No. 22-3170

rather than speech of the University. We do this by applying
the government-speech doctrine. Shurtleff v. City of Boston,
Mass., 

, 252–53 (2022).
       1. Government-Speech Doctrine
    Our evaluation of “whether the government intends to
speak for itself or to regulate private expression” on its prop-
erty “is not mechanical” nor guided by “the rote application
of rigid factors”; rather, “it is driven by a case’s context.”
Shurtleff, 

. Speech may be attributed to the gov-
ernment where, for example, the government has “actively
shaped or controlled” the messaging on its property to a
meaningful extent, 

 or where “the public would tend to
view the speech at issue as the government’s,” 

.
    For instance, in Pleasant Grove City, Utah v. Summum, the
Supreme Court found that Pleasant Grove City’s display of
donated monuments in its park constituted the City’s expres-
sive conduct, as the City actively selected certain monuments
“for the purpose of presenting the image of the City that it
wishe[d] to project,” and because “persons who observe[d]
donated monuments” on the City’s property would reasona-
bly “interpret them as conveying some message on the
[City’s] behalf.” 

, 471–73 (2009).
    Likewise, in Walker, the Supreme Court held that specialty
license plate designs offered by Texas for use by its citizens
conveyed Texas’ speech, even though some of the designs
were created and proposed by private groups. 576 U.S. at 204–
05, 208, 214–15. To that end, it was relevant that a proposed
design would not be offered to the public until the design re-
ceived affirmative approval from the Texas Department of
Motor Vehicle Board. Id. at 205, 213–14. In both Summum and
No. 22-3170                                                     19

Walker, then, the government was actively involved in select-
ing content to display on its property, such that the govern-
ment effectively adopted its selections as its own expression.
    In contrast, the Supreme Court in Shurtleff found that the
city of Boston was not engaged in government speech when,
“[f]or years,” it “allowed private groups to request use of the
flagpole [outside Boston City Hall] to raise flags of their
choosing.” 

. While the Court acknowledged
Boston’s tradition of displaying flags on its City Hall Plaza to
“convey the city’s messages,” 

, and further recog-
nized the possibility that the public might “ordinarily associ-
ate a flag’s message with Boston,” 

, it found “most
salient” that Boston lacked “meaningful involvement in the
selection of flags or crafting of their messages,” 

.
For instance, the Court noted that the city employee tasked
with handling flag-raising applications never viewed the
flags prior to the flag-raising events, and that Boston offered
no guidance as to the permissible messages that flags could
communicate. 

. As such, the flag raising was deemed
private speech. 

.
    Because the comment threads here operate similarly to
Boston’s flag-raising program and differently from a govern-
ment’s display of donated statues in its park or curated offer-
ing of license plate designs, we find that the comments are
private speech. Most critically, the University is not actively
involved in the selection of the comments displayed on its
comment threads, as the University does not sift through sub-
missions of comments to feature only those comments that
“present[] the image of the [University] that it wishes to pro-
ject.” Summum, 

. Were Moll or another social
media moderator to screen all comments prior to publication
20                                                        No. 22-3170

and then publish only those comments that aligned with the
University’s preferred image, the comments could more rea-
sonably be attributed to the University. But that is not how
the comment threads operate. Comments are published with-
out the University taking any action, such that “the degree of
government involvement” in shaping the comments’ messag-
ing is nearly imperceptible. Shurtleff, 

.
    We acknowledge that the keyword filters employed by the
University do amount to some government involvement in the
messages that ultimately appear on the comment threads. But
the nature and extent of that involvement is not meaningful
enough to transform the comments into the University’s
speech. The keyword filters operate passively, not actively, as
they only suppress comments that happen to contain certain
phrases or words. See Shurtleff, 

. The filters do
not evaluate the substance of a comment to display only com-
ments that convey University-approved messages. See 

. In fact, that a commenter may circumvent the filters to
publish a comment antithetical to the University’s messaging
(e.g., by editing the spacing of letters in a word) underscores
that the filters do not actively select messages the University
wishes to feature. Because the keyword filters are not mean-
ingfully or actively involved in selecting comments or shap-
ing a comment’s message, they do not supply a basis on
which we can find the comments constitute the University’s
speech. 7 See 

7
  The dissent portrays our reasoning as finding it “significant” that the
University uses an automated program rather than human curation in
suppressing comments from its comment threads. But our conclusion that
the University’s keyword filters do not transform the comments into
No. 22-3170                                                            21

    Given this absence of meaningful involvement by the Uni-
versity in selecting comments, we disagree with our dissent-
ing colleague that the comments are materially analogous to
letters to the editor featured in the University’s alumni mag-
azine. As the dissent acknowledges, in an alumni magazine,
each letter is chosen by the editor. That being the case, we take
no issue with the dissent’s position that a selection of letters
to the editor in the University’s magazine (whether in an
online or paper format) could constitute government speech.
But this case does not concern an alumni magazine, nor any
other online publication of the University. The comment
threads also do not function like an alumni magazine. Each
letter in an alumni magazine is chosen by the University; each
comment on the comment threads is not.
    We also disagree with the dissent’s application of Shurtleff
as suggesting that the mere existence of a policy governing
speech transforms any speech subject to that policy into the
government’s own speech. Were we to accept the dissent’s ap-
plication, policies restricting speech on government property
would become unreviewable: So long as the government had
some policy regulating speech, the policy would be protected
from First Amendment scrutiny under the government-
speech doctrine. We cannot agree with this result, nor do we
find this result supported by Shurtleff. In applying the govern-
ment-speech doctrine, Shurtleff did not instruct courts to ask
the binary question of whether the government had a policy
regulating speech or not. Rather, Shurtleff instructed that


government speech does not rest on the distinction between automated
and human curation. To the contrary, our conclusion rests on the fact that
the degree of the keyword filters’ involvement in the comment threads is
minimal at best. See Shurtleff, 

.
22                                                   No. 22-3170

courts conduct a “holistic inquiry” not driven by “the rote ap-
plication of rigid factors.” 

. That inquiry, it ex-
plained, may include an assessment of “the extent to which
the government has actively shaped or controlled the expres-
sion” on its property, 

 and whether that involvement is
“meaningful,” 

. Where there is no policy nor other
effort at shaping speech whatsoever, then the degree of gov-
ernment involvement will not be sufficiently meaningful to
warrant finding that the regulated speech is government
speech. See 

. But that does not mean that the existence
of a policy compels finding that any speech subject to that pol-
icy is government speech. What mattered in Shurtleff was not
the existence or absence of a policy, alone, but the extent of
government involvement in shaping speech. 

 at 256–58.
Here, the University’s minimal involvement in shaping the
comments on its comment threads weighs against finding
that the comments constitute the University’s speech.
    The University’s representations as to how it uses its com-
ment threads, and the nature of social media more generally,
reinforce that the University is not “enlist[ing] private [per-
sons] to convey [the University’s] own message” through the
private persons’ comments. Rosenberger v. Rector and Visitors
of Univ. of Va., 

 (1995). For example, the Uni-
versity represents that it does not generally “remove or re-
strict criticisms from the comment threads” on Facebook (dkt.
40 at 17 ¶ 78) and that it has replied to comments that criti-
cized the University (id. at 14 ¶ 64); (dkt. 54 at 17–18 ¶ 46).
The University also recognizes that its comment threads al-
low social media users to engage with one another, including
by debating issues such as the University’s research initia-
tives, the campus community’s willingness to wear face
masks, and the necessity and efficacy of a mobile COVID-19
No. 22-3170                                                    23

testing facility. (Dkt. 40 at 16 ¶¶ 74–75). Moreover, the Uni-
versity’s Social Media Statement places responsibility on the
commenter for the content of a comment, reflecting that the
University does not intend to “take[] ownership” of each com-
ment on its comment threads or otherwise “associate[] itself”
with the comments. Walker, 

. Collectively, these
representations illustrate that the University opened its com-
ment threads to facilitate discourse among and with its audi-
ence, but not to display its audience’s comments “for the pur-
pose of delivering a [University]-controlled message.” Sum-
mum, 

.
    The University’s discourse-oriented use of its comment
threads aligns with qualities inherent in social media, as well.
Indeed, the Supreme Court has acknowledged social media
as among “the most important places … for the exchange of
views,” likening it to a “modern public square” where indi-
viduals engage in protected First Amendment activities such
as speaking, listening, and debating topics “as diverse as hu-
man thought.” Packingham, 

. Because
individuals use social media with the expectation that they
will see and participate in debate and other discourse, we see
“little chance that observers will fail to appreciate the identity
of the speaker.” Summum, 

. The speaker is the
commenter. Should the University comment within a com-
ment thread attached to one of its posts (e.g., when replying
to a comment critical of the University), then the speaker will
be the University. But a comment published by a private per-
son constitutes the speech of that person, whose username is
displayed immediately prior to the person’s comment. We are
therefore confident that the public is unlikely to perceive
those comments as conveying a message on behalf of the Uni-
versity. See Shurtleff, 

. Accordingly, under the
24                                                 No. 22-3170

“holistic inquiry” mandated by Shurtleff, the public’s likely
perception presents another reason to find the comments con-
stitute private speech. 

    Because we find the University does not display the com-
ments of others as its own speech, but instead opens its com-
ment threads for private expression, we proceed to assess the
constitutionality of the University’s speech restrictions on its
comment threads by looking to “the nature of the relevant fo-
rum,” Cornelius v. NAACP Legal Def. & Educ. Fund, Inc., 

 (1985), and the government’s action, see Minne-
sota Voters All. v. Mansky, 

, 11–12 (2018).
       2. The Supreme Court’s Forum Jurisprudence
    The Supreme Court has identified several types of govern-
ment property, including: the traditional public forum, desig-
nated public forum, limited public forum, and nonpublic fo-
rum. Mansky, 

; Walker, 576 U.S. at 215–16; Mar-
tinez, 

 n.11.
   A traditional public forum is public property—for in-
stance, streets and parks—that “ha[s] immemorially been
held in trust for the use of the public, and, time out of mind,
ha[s] been used for purposes of assembly, communicating
thoughts between citizens, and discussing public questions.”
Perry, 

 (citation and quotation omitted). A des-
ignated public forum encompasses “government property
that has not traditionally been regarded as a public forum”
but “is intentionally opened up for that purpose.” Summum,

. In both traditional and designated public fo-
rums, “the government may impose reasonable time, place,
and manner restrictions on private speech, but restrictions
No. 22-3170                                                   25

based on content must satisfy strict scrutiny, and those based
on viewpoint are prohibited.” Mansky, 

.
    The government has considerably greater leeway in re-
stricting speech in the remaining forums: limited public fo-
rums and nonpublic forums. A limited public forum is estab-
lished when the government “open[s] property ‘limited to use
by certain groups or dedicated solely to the discussion of cer-
tain subjects.’” Martinez, 

 n.11 (quoting Sum-
mum, 

). And a nonpublic forum is a “space that
‘is not by tradition or designation a forum for public commu-
nication.’” Mansky, 

 (quoting Perry, 

). Rather, it exists “[w]here the government is acting as a
proprietor, managing its internal operations.” Int’l Soc’y for
Krishna Consciousness, Inc. v. Lee, 

 (1992); see
also Women’s Health Link, Inc. v. Fort Wayne Pub. Trans. Corp.,

 (7th Cir. 2016) (describing a nonpublic forum
as property “that could be and sometimes [is] used for private
expressive activities but [is] not primarily intended for such
use”). Speech restrictions applied to limited public forums
and nonpublic forums need only be reasonable and view-
point-neutral; there is no strict scrutiny analysis involved.
Martinez, 

; Mansky, 585 U.S. at 11–12.
       3. Determining the Type of Forum
    Krasno argues the University’s comment threads are des-
ignated public forums, while the University contends they are
either nonpublic or limited public forums.
   When considering how to classify a forum, the govern-
ment’s intent in establishing the forum matters. Cornelius, 

. Ascertaining the government’s intent requires that
we look to “the nature of the property and its compatibility
26                                                        No. 22-3170

with expressive activity,” as well as the government’s “policy
and practice.” 

    The comment threads attached to the University’s posts
are inherently compatible with expressive activity because
they are designed to facilitate that purpose. At the same time,
the University’s policy and practice demonstrate that in de-
ciding to open to the public the comment threads on its posts,
the University did not intend the public’s expressive activity
to proceed entirely unchecked. Instead, the University re-
served the right to remove any content for, among other
things, being off topic. 8 This policy was not a hidden or post
hoc rationale to support the University’s comment modera-
tion decisions. See Air Line Pilots Ass’n, Int’l v. Dep’t of Aviation
of the City of Chi., 

, 1153–54 (7th Cir. 1995). Rather,
the Social Media Statement was publicly posted on the Uni-
versity’s website before this litigation began, and it consti-
tutes “[o]bjective indicia” of the University’s intent. 

.
    The University’s Social Media Statement and efforts at en-
forcing that statement reflect that the University did not in-
tend to create a designated public forum without limits on
discussion, but instead intended to open the comment sec-
tions for the discussion of certain subjects—principally, those
that relate to its posts. See Cornelius, 473 U.S. at 802–03 (“We
will not find that a public forum has been created in the face
of clear evidence of a contrary intent” or based on mere “in-
action” by the government). Accordingly, we find the


8
 Krasno does not challenge the portions of the University’s Social Media
Statement that reserve the right to remove content for reasons other than
being off topic.
No. 22-3170                                                               27

University’s comment threads are best characterized as lim-
ited public forums. 9
    Our conclusion aligns with the few circuit courts that have
applied the Supreme Court’s forum jurisprudence to govern-
ment-run social media accounts. In general, courts have
found “that the comment threads of government social media
pages are designated public forums when the pages are open
for comment without restrictions and limited public forums
when the government prospectively sets restrictions.” People
for the Ethical Treatment of Animals v. Tabak, 

(D.C. Cir. 2024) (collecting cases). Because the University
chose to promulgate and implement its Social Media State-
ment rather than open its comment threads without re-
striction, we conclude the comment threads attached to the
University’s posts are limited public forums. And while we
acknowledge the dissent characterizes our holding as finding
that Facebook itself is a limited public forum, we emphasize
that our holding is narrower. We decide only that the com-
ment threads attached to the University’s social media
posts—and not social media platforms more broadly—are
limited public forums.


9
 Classifying the comment threads as nonpublic forums would not change
how we assess the constitutionality of the University’s speech restrictions.
We have at times treated limited public and nonpublic forums as synony-
mous, see Milestone v. City of Monroe, 

, 783 n.3 (7th Cir. 2011),
while in other cases we have treated them as distinct, see Women’s Health
Link, Inc., 

. Regardless of whether the comment threads are
deemed limited public or nonpublic forums, the University’s regulation
of speech must be reasonable and viewpoint neutral. See Summum, 

 (limited public forums); Mansky, 

 (nonpublic fo-
rums).
28                                                    No. 22-3170

       4. Applying the Reasonable-and-Viewpoint-Neutral Test
          to the Off-Topic Comment Rule
    Because the University’s comment threads are limited
public forums, we must determine whether the speech re-
strictions within these spaces are reasonable and viewpoint
neutral. See Martinez, 

 n.11; Rosenberger, 

.
    We find unconstitutional the University’s off-topic rule as
applied to Krasno because it is not viewpoint neutral, but in-
stead discriminates against Krasno’s anti-animal testing and
pro-animal rights viewpoint. The terms the University se-
lected for filtering are terms that one would reasonably expect
to be used by individuals opposed to animal testing. That the
University also chose to program its keyword filters to hide
words and phrases that are unrelated to animal testing does
not change the fact that many of the keywords selected by the
University align significantly with Krasno’s anti-animal test-
ing and pro-animal rights viewpoint. And the University’s
admission that it has hidden an on-topic comment by Krasno
suggests that the University’s moderation decisions were
spurred not by the relevance of Krasno’s commentary to the
underlying posts, but by the views her commentary con-
veyed.
    We also find the off-topic comment rule unconstitutional
because it is unreasonable. To assess reasonableness, we con-
sider the University’s justifications for its off-topic rule. Man-
sky, 

; Martinez, 

. Recall that the off-
topic comment rule, which affords the University’s social me-
dia staff discretion to moderate comments it deems off topic,
is rooted in the University’s public Social Media Statement
and was reaffirmed by the University in its interim guidance.
No. 22-3170                                                   29

The University claims the rule keeps its comment threads un-
clogged, enabling its audience to interact with its posts and
the University to respond to questions and comments. We
agree this is a permissible objective given the communicative
purpose of the comment threads.
    But lines drawn by the University to further this permissi-
ble objective must be reasonable. In other words, “the [Uni-
versity] must be able to articulate some sensible basis for dis-
tinguishing what may come in from what must stay out” of
its comment threads. See Mansky, 

. While the
University need not be perfectly clear and precise, an “inde-
terminate prohibition” enforced through sheer discretion
without “objective, workable standards” is not reasonable. 

.
     The University’s bare instruction in its Social Media State-
ment that comments may be removed if “off-topic” and its
discretion-based practice of enforcing that statement through
manual moderation decisions and keyword filters fail this
test. The University has neither clarified what it means to be
“off-topic” nor provided “objective, workable standards” to
guide its social media managers’ discretion. 

 To be sure, af-
ter this litigation began, the University’s interim guidance
provided staff a single example of what is “on vs. off topic”—
i.e., its Taylor Swift example. But even that limited guidance
refers to the “topic … of the page”—seemingly referring to the
webpages associated with the University’s accounts—which
conflicts with the University’s assertion in this litigation that
comments must relate to the topic of the “post.” Moreover, by
its own terms the guidance does not mandate consistent mod-
eration of off-topic comments, and the University admits
moderation decisions are discretionary. In sum, the
30                                                    No. 22-3170

University has not supplied a “sensible basis for distinguish-
ing what may come in from what must stay out.” 

.
    “It is ‘self-evident’ that an indeterminate prohibition car-
ries with it ‘[t]he opportunity for abuse, especially where [it]
has received a virtually open-ended interpretation.’” 

 at 21
(quoting Bd. of Airport Com’rs of City of Los Angeles v. Jews for
Jesus, Inc., 

 (1987)). The unreasonableness of
the off-topic comment rule and potential for abuse is ce-
mented by the University’s admission that it has hidden cer-
tain comments of Krasno that were on-topic comments. Such
missteps demonstrate that the University has not “respect[ed]
the lawful boundaries it has itself set.” Rosenberger, 

.
   The off-topic comment rule is also unreasonable because,
“as implemented by the keywords, … it is inflexible and un-
responsive to context”—meaning it does not really operate as
an “off-topic” rule in the first place. Tabak, 109 F.4th at 637–38.
The keyword filters apply automatically to comments on the
University’s posts, regardless of a given post’s topic. In other
words, Krasno could not today renew her comment on the
University’s December 22, 2020, Instagram post about a dog
being treated for cancer, even though the University admitted
such a comment would be on topic. The record contains no
evidence that the University manually approves on-topic
comments initially hidden by its filters.
    In short, the University’s inflexible and context-blind key-
word filters do not reasonably further its “off-topic” justifica-
tion when there is no way to know before a post is created
whether a given phrase will be on- or off-topic in relation to
the post. See id. at 636 (finding unreasonable a highly similar
off-topic comment rule that “consider[ed] words related to
No. 22-3170                                               31

animal testing categorically ‘off-topic’”). We therefore con-
clude that the University’s off-topic comment restriction, as
currently written and enforced against Krasno, is unreasona-
ble under the First Amendment.
                   III.   CONCLUSION
    For these reasons, we REVERSE the entry of summary
judgment for the University and REMAND with instruction
to the district court to enter judgment for Krasno consistent
with this opinion.
32                                                    No. 22-3170

    EASTERBROOK, Circuit Judge, dissenting. On Wisconsin is the
alumni magazine of the University of Wisconsin-Madison.
Some of its articles bear bylines of the staff, some of alumni,
some of faculty members, and some of third parties. No mat-
ter the author, each article is chosen by the editor. On Wiscon-
sin also carries letters to the editor, which appear at pages 6
and 7 of the current issue (Summer 2025). These letters discuss
the University, its sports teams, and its graduates. None men-
tions animal research.
    The magazine’s contents are governmental speech. Arti-
cles and letters may be written and signed by private parties,
but the University as the publisher decides what to print. See
Hosty v. Carter, 

 (7th Cir. 2005) (en banc) (student
paper is speech by the school if teachers act as editors, but by
the students if the school supplies only a financial grant). In
this respect an alumni magazine is no different from a news-
paper. The paper’s owner does not write the stories, but the
publisher decides what to print—something as true of the
news pages, and the letters to the editor, as it is of the editori-
als. See Miami Herald Publishing Co. v. Tornillo, 

(1974) (state cannot compel a newspaper to engage in “bal-
anced” coverage of political campaigns by running letters op-
posed to the paper’s views). No one believes that a byline on
a story or letter strips a magazine’s or newspaper’s editor of
the right to decide what appears.
   A governmental body has the same right to speak as any
private entity. “When the government wishes to state an opin-
ion, to speak for the community, to formulate policies, or to
implement programs, it naturally chooses what to say and
what not to say. That must be true for government to work.
Boston could not easily congratulate the Red Sox on a victory
No. 22-3170                                                     33

were the city powerless to decline to simultaneously transmit
the views of disappointed Yankees fans. The Constitution
therefore relies first and foremost on the ballot box, not on
rules against viewpoint discrimination, to check the govern-
ment when it speaks.” Shurtleff v. Boston, 

, 251–52
(2022) (citation omitted). See also, e.g., Walker v. Texas Divi-
sion, Sons of Confederate Veterans, Inc., 

 (2015);
Pleasant Grove v. Summum, 

, 467–69 (2009); Univer-
sity of Wisconsin v. Southworth, 

 (2000); Rosen-
berger v. University of Virginia, 

 (1995).
    In Summum the Court held that a monument was govern-
mental speech, even though a private party submitted and
paid for the monument. That’s equally true of letters to the
editor. “[E]xpressive activity includes presenting a curated
compilation of speech originally created by others.” Moody v.
NetChoice, LLC, 

 (2024). See also 

 at 728–33
(collecting cases), 781 (Alito, J., concurring). So too the selec-
tion of speakers to appear on public TV is a form of govern-
mental expression, no matter who produces or submits the
shows. Arkansas Educational Television Commission v. Forbes,

 (1998). Similarly, the choice of floats in a pa-
rade represents the speech of the organizer, not of the clubs
that submit the floats, even though each float bears its spon-
sor’s name. Hurley v. Irish-American Gay, Lesbian & Bisexual
Group of Boston, Inc., 

 (1995).
    Given the power of a magazine’s editor to choose which
letters (if any) to publish, there can’t be any problem if the
editor delegates this task to another member of the staff, with
guidelines to ensure that the staff follows the editor’s prefer-
ences. The editor might tell the staff: (1) letters collectively
cannot exceed two pages of the magazine; (2) letters must not
34                                               No. 22-3170

undermine the University (so no praise of the Ohio State foot-
ball team or requests for contributions to Northwestern); (3)
no letters on the topic of animal research, a topic the editor
wants to reserve for commissioned articles.
    Delegation to a subordinate can’t matter to constitutional
analysis, nor does the physical format of the magazine. Sup-
pose On Wisconsin appears as a PDF file on a website in addi-
tion to a printed version. (It does. See onwiscon-
sin.uwalumni.com/issues.) Suppose the University also pro-
duces the magazine as a website with clickable links to arti-
cles, letters, and other materials. (It does. See onwiscon-
sin.uwalumni.com.) The “extra space” of a website might lead
the editor to relax Guideline 1—or maybe not, as the editor
might not want to change the balance between commissioned
articles and readers’ submissions (now apt to be called “com-
ments” or “posts” rather than letters). But the magazine’s ex-
istence as a website would not be likely to change Guidelines
2 and 3, nor would the Constitution compel the editor to
change those Guidelines. If Boston need not praise the Yan-
kees, the University of Wisconsin need not publicize disgrun-
tled graduates’ proposals that alums send their donations to
Northwestern. Nor does the fact that each letter (or post, or
comment) is signed compel the University to publish it. Like-
wise the fact that On Wisconsin does not publish the Guide-
lines would not compel it to print everything submitted. Au-
thors and editors need not explain their policies to the general
public.
   Next step: Instead of hosting the magazine on its own
website, the University decides to host the magazine, or its
equivalent, on Facebook, and to use the Guidelines to select
which comments will appear there. This should not make a
No. 22-3170                                                     35

constitutional difference. The University remains the speaker
and editor. It labels its Facebook page “the official voice of the
University.” Neither the University nor Facebook censors pri-
vate speech; any would-be speakers can say whatever they
want on their own websites, Facebook pages, other social me-
dia, or letters to the editor of the Madison Capital Times.
     If there is to be a legal difference, the University ought to
have greater editorial latitude on Facebook. For Facebook is a
private entity, which controls what can be posted on its plat-
form and how content filters work. Publishing On Wisconsin,
by contrast, is a governmental operation from start to finish.
When the University uses the tools of social media, it is a state
actor (see Lindke v. Freed, 

 (2024)), but as a state
actor the University remains entitled to select its own speech.
As long as “expressive activity includes presenting a curated
compilation of speech originally created by others”, Moody,

, it is hard to see why the University’s selectiv-
ity is more problematic on Facebook than on its own website,
the PDF version of the magazine, or the paper version of the
magazine. Similarly, the fact that the Miami Herald has “more
room” on a website or Facebook page than in the printed pa-
per would not allow a state to compel the Herald to offer equal
time to all political candidates. The newspaper would retain
its right to promote its favored causes and candidates. Cf. Pa-
cific Gas & Electric Co. v. Public Utilities Commission, 

(1986) (“extra space” in a billing envelope cannot be con-
scripted by a state to undermine the mailer’s political or eco-
nomic views).
    Yet my colleagues conclude that the move from wisc.edu
or an alumni site to facebook.com flips the constitutional rule.
Instead of the editors being allowed to choose, the readers get
36                                                 No. 22-3170

to compel the University to include their submissions. This is
so, the majority says, because Facebook is a “limited public
forum.” I don’t understand why the University’s site on Face-
book is any more a public forum than is On Wisconsin’s web
page, or for that matter the printed magazine. All of these are
outlets for the University to get its ideas across. Moving from
a site under the University’s control to a private one (Face-
book) should curtail the rights of the general public to compel
the University to include what it deems anti-University
speech.
    Shurtleff is the Supreme Court’s most recent effort to dis-
tinguish public forums from governmental speech. For dec-
ades Boston allowed private groups to fly their flags over the
plaza of City Hall, but it balked at what one group called a
“Christian Flag.” In rejecting Boston’s argument that the
choice of flag was governmental speech, the Court empha-
sized that the city lacked any criteria for choosing which flags
could appear and that the tradition had been first-come-first-
served. That approach disabled the city from contending that
the flags represented its own speech. “[T]he city's lack of
meaningful involvement in the selection of flags or the craft-
ing of their messages leads us to classify the flag raisings as
private, not government, speech—though nothing prevents
Boston from changing its policies going forward.” 

. In other words, when a governmental body does have pol-
icies that shape the message being presented, the government
is entitled to claim editorial privilege to speak. And that is ex-
actly what the University of Wisconsin has done. Its policies,
designed to mold the views that appear on its Facebook page,
mean that the page is not a public forum under Shurtleff. The
University can curate the site to make the message its own.
No. 22-3170                                                       37

    My colleagues find it significant that the University uses
Facebook’s built-in filters rather than (or in addition to) hu-
man curation. That should not matter, any more than the del-
egation from On Wisconsin’s editor-in-chief to its staff should
matter. What does matter under Shurtleff is that the University
has policies about what can appear and what can’t. If these
policies are permissible in printed matter, where manual se-
lection is the norm, what makes these policies constitutionally
forbidden where computers do most of the work? As long as
people make the rules that the computers execute, the result
speaks for the University. Shurtleff tells us that having rules
about content is what separates an open forum from govern-
mental speech—and the University of Wisconsin assuredly
has rules about what can appear on its Facebook page. My
colleagues think these rules unreasonable. Maybe; maybe not;
my point is that the rules exist, and their very existence is what
nixes the assertion that the Facebook page is a public forum.
    A university’s Facebook page, designed to portray the uni-
versity in a light of the university’s choice, is no more a public
forum than is the National Portrait Gallery, which is designed
to present great art (of the curators’ selection) rather than
kitsch. People for the Ethical Treatment of Animals, Inc. v. Gittens,

 (D.C. Cir. 2005) (city can select among “party
animals” as street art even though all of them may be kitsch).
Just so, a public library chooses what to make available on the
shelves, and a decision to have Churchill’s six-volume The
Second World War does not compel it to offer Hitler’s Mein
Kampf as well. See Little v. Llano County, 

 (5th Cir.
2025) (en banc). It does not matter that the curators and librar-
ians work under open-ended guidelines (“fine art”; “useful
books”), or that the curators and librarians are not the top pol-
icy makers (they would be the Board of the Smithsonian and
38                                                 No. 22-3170

the legislature of the jurisdiction paying for the library). Using
those details to declare the museums and libraries are public
forums would prevent them from fulfilling their missions.
The constitutional point is instead that the museum and the
library do not regulate anyone’s speech and reading; kitsch
and Mein Kampf are widely available. Nor would it make a
difference if a visitor hangs a portrait that is removed by the
Portrait Gallery’s director (parallel to removing comments
that find their way onto Facebook).
     My colleagues’ approach has support from People for the
Ethical Treatment of Animals, Inc. v. Tabak, 

 (D.C.
Cir. 2024), which declares that a Facebook account established
by the National Institutes of Health is a “limited public fo-
rum” even though it implements access and topic controls
through keyword filters. I do not find the court’s analysis con-
vincing. The panel in PETA v. Tabak did not mention the same
circuit’s contrary approach in PETA v. Gittens. More im-
portant, it did not discuss Shurtleff, Moody, or many of the
other decisions on which I have relied. Summum received a
citation, but only in passing—no analysis. The panel did not
try to explain why a website differs from a newspaper or mag-
azine published by a governmental body. The decision seems
to be more a response to the Institute’s concession that the Fa-
cebook site is a public forum than it is an independent analy-
sis.
   Tabak does mention, also in passing, Knight First Amend-
ment Institute v. Trump, 

 (2d Cir. 2019), rehearing
en banc denied, 

 (2020), vacated as moot under
the name Biden v. Knight First Amendment Institute, 

 (2021). Knight held that a public official who posts or dis-
cusses official policy on social media is constitutionally
No. 22-3170                                                 39

forbidden to use the sites’ content-moderation features, such
as limits on who may respond and restrictions on off-topic
posts. This decision lacks precedential force, given the vaca-
tur, and has the same problem as PETA v. Tabak in explaining
why public officials are not entitled to express their views by
curating (and sometimes excluding) words penned by third
parties. Judge Park’s dissent from the denial of rehearing en
banc in the Second Circuit (953 F.3d at 228–31) exposes some
of these problems, and Knight has taken its lumps from aca-
demic commentators who wonder why forum analysis is ap-
propriate when a private actor controls the site. E.g., Noah
Feldman, Appeals Court Asks Wrong Question in Trump Twitter
Blocking Case, which appeared at bloomberg.com on July 9,
2019. Twitter (now X), the main forum in Knight, demon-
strated its control by excluding President Trump entirely,
which scotches any idea that it was a “public” forum con-
trolled by the government. The Supreme Court did not ad-
dress any substantive issue in Knight, given intervening moot-
ness, but this is not a good reason for us to follow the Second
Circuit down the rabbit hole.
    PETA v. Tabak and Knight lack support in the Supreme
Court’s decisions. They cannot be reconciled with Moody,
Summum, Hurley, Little, PETA v. Gittens, and many other de-
cisions, which hold that choosing what private speech to pre-
sent is a form of governmental speech. We should stick with
Moody (“expressive activity includes presenting a curated
compilation of speech originally created by others”) and not
be led astray by flawed decisions of the D.C. and Second Cir-
cuits.
  Given my colleagues’ recognition that the letters section of
On Wisconsin is not a public forum, the University may
40                                                No. 22-3170

respond to today’s decision by closing all comment options
on Facebook, vindicating its status as “the official voice of the
University”—or manually vetting all comments and posting
only those that the editors like. Neither response would pro-
mote the plaintiff’s goals, and both would harm other alums
whose views would disappear.