Opinion by Judge BERZON. Opinion concurring in part and dissenting in part by Judge GOULD.
BERZON, Circuit Judge.There is no reason inherent in the technological features of cyberspace why First Amendment and defamation law should apply differently in cyberspace than in the brick and mortar world. Congress, however, has chosen for policy reasons to immunize from liability for defamatory or obscene speech “providers and users of interactive computer services” when the defamatory or obscene material is “provided” by someone else. This case presents the question whether and, if so, under what circumstances a moderator of a list-serv and operator of a website who posts an allegedly defamatory e-mail authored by a third party can be held liable for doing so. The case also presents a novel procedural question — whether the denial of an Anti-SLAPP motion filed pursuant to California law can be appealed prior to a final judgment in the underlying case. After recounting the unusual tale underlying this case, we address each of these questions in turn.
I.
In the summer of' 1999, sometime-handyman Robert Smith was working for Ellen Batzel, an attorney licensed to practice in California and North Carolina, at Batzel’s house in the North Carolina mountains. Smith recounted that while he was repairing Batzel’s truck, Batzel told him that she was “the granddaughter of *1021one of Adolf Hitler’s right-hand men.” Smith also maintained that as he was painting the walls of Batzel’s sitting room he overheard Batzel tell her roommate that she was related to Nazi politician Heinrich Himmler. According to Smith, Batzel told him on another occasion that some of the paintings hanging in her house were inherited. To Smith, these paintings looked old and European.
After assembling these clues, Smith used a computer to look for websites concerning stolen art work and was directed by a search engine to the Museum Security Network (“the Network”) website. He thereupon sent the following e-mail message to the Network:
From: Bob Smith [e-mail address omitted]
To: securma@museum-security.org [the Network]1
Subject: Stolen Art
Hi there,
I am a building contractor in Asheville, North Carolina, USA. A month ago, I did a remodeling job for a woman, Ellen L. Batzel who bragged to me about being the grand daughter [sic] of “one of Adolph Hitler’s right-hand men.” At the time, I was concentrating on performing my tasks, but upon reflection, I believe she said she was the descendant of Heinrich Himmler.
Ellen Batzel has hundreds of older European paintings on her walls, all with heavy carved wooden frames. She told me she inherited them.
I believe these paintings were looted during WWII and are the rightful legacy of the Jewish people. Her address is [omitted].
I also believe that the descendants of criminals should not be persecuted for the crimes of the [sic] fathers, nor should they benefit. I do not know who to contact about this, so I start with your organization. Please contact me via email [...] if you would like to discuss this matter.
Bob.
Ton Cremers, then-Director of Security at Amsterdam’s famous Rijksmuseum and (in his spare time) sole operator of the Museum Security Network (“the Network”), received Smith’s e-mail message. The nonprofit Network maintains both a website and an electronic e-mailed newsletter about museum security and stolen art. Cremers periodically puts together an electronic document containing: e-mails sent to him, primarily from Network subscribers; comments by himself as the moderator of an on-line discussion; and excerpts from news articles related to stolen works of art. He exercises some editorial discretion in choosing which of the emails he receives are included in the list-serv mailing, omitting e-mails unrelated to stolen art and eliminating other material that he decides does not merit distribution to his subscribers. The remaining amalgamation of material is then posted on the Network’s website and sent to subscribers automatically via a listserv.2 The Net*1022work’s website and listserv mailings are read by hundreds of museum security officials, insurance investigators, and law enforcement personnel around the world, who use the information in the Network posting to track down stolen art.
After receiving it, Cremers published Smith’s e-mail message to the Network, with some minor wording changes, on the Network listserv. He also posted that list-serv, with Smith’s message included, on the Network’s website. Cremers later included it on the Network listserv and posted a “moderator’s message” stating that “the FBI has been informed of the contents of [Smith’s] original message.”
After the posting, Bob Smith e-mailed a subscriber to the listserv, Jonathan Sazo-noff, explaining that he had had no idea that his e-mail would be posted to the listserv or put on the web. Smith told Sazanoff:
I [was] trying to figure out how in blazes I could have posted me [sic] email to [the Network] bulletin board. I came into MSN through the back door, directed by a search engine, and never got the big picture. I don’t remember reading anything about a message board either so I am a bit confused over how it could happen. Every message board to which I have ever subscribed required application, a password, and/or registration, and the instructions explained this is necessary to keep out the advertisers, cranks, and bumbling idiots like me.
Batzel discovered the message several months after its initial posting and complained to Cremers about the message. Cremers then contacted Smith via e-mail to request additional information about Smith’s allegations. Smith continued to insist on the truth of his statements. He also told Cremers that if he had thought his e-mail “message would be posted on an international message board [he] never would have sent it in the first place.”
Upon discovering that Smith had not intended to post his message, Cremers apologized for the confusion. He told Smith in an e-mail that “[y]ou were not a subscriber to the list and I believe that you did not realize your message would be forwarded to the mailinglist [sic].” Apparently, subscribers send messages for inclusion in the listserv to securma@x54all.nl, a different address from that to which Smith had sent his e-mail contacting the Network. Cremers further explained that he “receive[s] many e-mails each day some of which contain queries [he thinks] interesting enough to forward to the list. [Smith’s] was one of those.”
Batzel disputes Smith’s account of their conversations. She says she is not, and never said she is, a descendant of a Nazi official, and that she did not inherit any art. Smith, she charges, defamed her not because he believed her artwork stolen but out of pique, because Batzel refused to show Hollywood contacts a screenplay he had written.
Batzel claims further that because of Cremers’s actions she lost several prominent clients in California and was investigated by the North Carolina Bar Association. Also, she represents that her social reputation suffered. To redress her claimed reputational injuries she filed this lawsuit against Smith, Cremers, the Netherlands Museum Association,3 and Mosler, Inc. (“Mosler”)4 in federal court in Los Angeles, California.
*1023Cremers countered with two motions: (1) a motion to strike under the California anti-SLAPP statute,5 alleging that Batzel’s suit was meritless and that the complaint was filed in an attempt to interfere with his First Amendment rights, and (2) a motion to dismiss for lack of personal jurisdiction. The court denied both motions. Cremers then filed this appeal.
Batzel also alleged in her complaint that Mosler was vicariously liable for her reputational injuries because Cremers was acting as Mosler’s agent. This agency relationship arose, according to Batzel, because Mosler gave Cremers $8,000 for displaying Mosler’s logo and other advertisements on the Network website and in its listserv. The district court entered summary judgment in favor of Mosler, ruling that, under California law as applied to the undisputed facts, Cremers was not an agent of Mosler and Mosler could not be vicariously liable. Batzel appeals this decision as well.
II
The district court denied Cremers’s motion to dismiss for lack of personal jurisdiction on June 5, 2001. Cremers’s notice of appeal, which sought to appeal that order along with the district court’s denial of Cremers’s anti-SLAPP motion, was filed on July 27, 2001. No extension of time to appeal was sought or granted. The appeal of the June 5 order accordingly is untimely because it was not filed within thirty days of that order. Fed. R.App. P. 4(a)(1)(A).
It is understandable that Cremers did not promptly appeal the denial of his motion to dismiss for lack of personal jurisdiction, because the denial was not an appealable order. See Catlin v. United States, 324 U.S. 229, 236, 65 S.Ct. 631, 89 L.Ed. 911 (1945). Cremers argues, however, that we should review the denial of his motion to dismiss for lack of personal jurisdiction as an issue pendent to his appeal of the denial of his anti-SLAPP motion.
Although our circuit permits the exercise of appellate jurisdiction over otherwise non-appealable orders that are “inextricably intertwined” with another order that is properly appealable, see, e.g., Mendocino Envtl. Ctr. v. Mendocino County, 192 F.3d 1283, 1296-97 (9th Cir.1999), we have interpreted “inextricably intertwined” very narrowly for that purpose. See Cunningham v. Gates, 229 F.3d 1271, 1284 (9th Cir.2000). We require that the two issues: “(a) be so intertwined that we must decide the pendent issue in order to review the claims properly raised on interlocutory appeal ... or (b) resolution of the issue properly raised on interlocutory appeal necessarily resolves the pendent issue.” Id. at 1285 (internal citations omitted). The attempted appeal of the denial of the motion to dismiss for lack of personal jurisdiction fails both requirements. We can decide the anti-SLAPP issue entirely independently of the question of personal jurisdiction, and different legal standards apply to each issue. See id. We therefore dismiss Cremers’s appeal insofar as it challenges the denial of his motion to dismiss for lack of personal jurisdiction.6
Ill
A. California’s Anti-SLAPP Statute
California law provides for pre-trial dismissal of “SLAPPs”: “Strategic Lawsuits *1024against Public Participation.” Cal.Civ. Proc.Code § 425.16. These are lawsuits that “masquerade as ordinary lawsuits” but are brought to deter common citizens from exercising their political or legal rights or to punish them for doing so. Wilcox v. Superior Court, 27 Cal.App.4th 809, 816, 33 Cal.Rptr.2d 446 (1994), overruled on other grounds by Equilon Enter. v. Consumer Cause, Inc., 29 Cal.4th 53, 124 Cal.Rptr.2d 507, 52 P.3d 685 (2002) (citations omitted). “The anti-SLAPP statute was enacted to allow for early dismissal of meritless first amendment cases aimed at chilling expression through costly, time-consuming litigation.” Metabolife Int’l, Inc. v. Wornick, 264 F.3d 832, 839 (9th Cir.2001).
If the defendant files an anti-SLAPP motion to strike, all discovery proceedings are stayed. See § 425.16(g). A court may, however, permit specified discovery “on noticed motion and for good cause shown.” Id. In order to prevail on an anti-SLAPP motion, the defendant is required to make a prima facie showing that the plaintiffs suit arises from an act by the defendant made in connection with a public issue in furtherance of the defendant’s right to free speech under the United States or California Constitution. See United States ex ret. Newsham v. Lockheed Missiles & Space Co., 190 F.3d 963, 971 (9th Cir.1999); see also § 425.16(e) (defining “act in furtherance of a person’s right of ... free speech.”).
The burden then shifts to the plaintiff to establish a reasonable probability that the plaintiff will prevail on his or her defamation claim. See Lockheed, 190 F.3d at 971. “[T]he plaintiff must demonstrate that ‘the complaint is legally sufficient and supported by a prima facie showing of facts to sustain a favorable judgment if the evidence submitted by plaintiff is [sic] credited.’ ” Metabolife, 264 F.3d at 840 (quoting Wilcox v. Superior Court, 27 Cal.App.4th 809, 33 Cal.Rptr.2d 446, 454).
If the court denies an anti-SLAPP motion to strike, the parties continue with discovery. See § 425.16(g). Once the plaintiffs case has survived the motion, the anti-SLAPP statute no longer applies and the parties proceed to litigate the merits of the action.
B. Jurisdiction
We are presented with the threshold question whether we have jurisdiction over Cremers’s interlocutory appeal of the district court’s denial of his motion to strike. In California state court, a denial of an anti-SLAPP motion is immediately appealable. See § 425.16(j). The issue before us as a federal court is whether a district court’s denial of an anti-SLAPP motion is an immediately appealable “final decision” under 28 U.S.C. § 1291, so that we have jurisdiction to address Cremers’ appeal. We conclude that we have jurisdiction to review the denial of an anti-SLAPP motion pursuant to the collateral order doctrine.
In general, a party is entitled only to a single appeal, to be “deferred until final judgment has been entered.” See Digital Equip. Corp. v. Desktop Direct, Inc., 511 U.S. 863, 868, 114 S.Ct. 1992, 128 L.Ed.2d 842 (1994). The collateral order doctrine establishes “a narrow class of decisions that do not terminate the litigation, but must, in the interest of achieving a healthy legal system nonetheless be treated as final.” Id. at 867, 114 S.Ct. 1992 (internal citations and quotation marks omitted). To fall into this narrow class of immediately appealable orders, a district court decision must (1) be “conclusive,” (2) “resolve important questions completely separate from the merits,” and (3) “render such questions effectively unreviewable on *1025appeal from a final judgment in the underlying action.” Id.
In the present case all three factors are met. First, the court’s denial of Cremers’s anti-SLAPP motion is conclusive as to whether the anti-SLAPP statute required dismissal of Batzel’s suit. If an anti-SLAPP motion to strike is granted, the suit is dismissed and the prevailing defendant is entitled to recover his or her attorney’s fees and costs. See § 425.16(c). If the motion to strike is denied, the anti-SLAPP statute does not apply and the parties proceed with the litigation.
The district court’s denial of Cremers’s motion to strike resolved the threshold question whether Batzel’s defamation suit should be dismissed by pre-trial motion. Denial of an anti-SLAPP motion resolves a question separate from the merits in that it merely finds that such merits may exist, without evaluating whether the plaintiffs claim will succeed. See § 425.16(b)(3) (“If the court determines that the plaintiff has established a probability that he or she will prevail on the claim, neither that determination nor the fact of that determination shall be admissible in evidence at any later stage of the case.... ”). The purpose of an anti-SLAPP motion is to determine whether the defendant is being forced to defend against a meritless claim. The anti-SLAPP issue therefore exists separately from the merits of the defamation claim itself.
Because the anti-SLAPP motion is designed to protect the defendant from having to litigate meritless cases aimed at chilling First Amendment expression, the district court’s denial of an anti-SLAPP motion would effectively be unreviewable on appeal from a final judgment. We find it instructive that California’s anti-SLAPP statute provides that an order denying an anti-SLAPP motion may be appealed immediately. See § 425.16(j). This provision, along with the legislative history behind § 425.16, demonstrates that California lawmakers wanted to protect speakers from the trial itself rather than merely from liability. As the California Senate Judiciary Committee noted before the law’s enactment,
Without [the right of immediate appeal], a defendant will have to incur the cost of a lawsuit before having his or her right to free speech vindicated.... [Wlhen a meritorious anti-SLAPP motion is denied, the defendant, under current law, has only two options. The first is to file a writ of appeal, which is discretionary and rarely granted. The second is to defend the lawsuit. If the defendant wins, the anti-SLAPP law is useless and has failed to protect the defendant’s constitutional rights.
Cal. Sen. Judiciary Comm. Rep. on AB 1675, at 4.7
If the defendant were required to wait until final judgment to appeal the denial of a meritorious anti-SLAPP motion, a decision by this court reversing the district court’s denial of the motion would not remedy the fact that the defendant had been compelled to defend against a meritless claim brought to chill rights of free expression. Thus, a defendant’s rights under the anti-SLAPP statute are in the nature of immunity: They protect the defendant from the burdens of trial, not merely from ultimate judgments of liability.
Because California law recognizes the protection of the anti-SLAPP statute as a substantive immunity from suit, this Court, sitting in diversity, will do so as *1026well. See generally Erie R.R. Co. v. Tompkins, 304 U.S. 64, 58 S.Ct. 817, 82 L.Ed. 1188 (1938). A district court’s denial of a claim of immunity, to the extent that it turns on an issue of law, is an appealable final decision within the meaning of 28 U.S.C. § 1291 notwithstanding the absence of a final judgment. See Mitchell v. Forsyth, 472 U.S. 511, 525-27, 105 S.Ct. 2806, 86 L.Ed.2d 411 (1985); see also Estate of Kennedy v. Bell Helicopter Textron, Inc., 283 F.3d 1107, 1111 (9th Cir.2002) (applying the collateral order doctrine to an appeal from an order denying protection under federal statute of repose, on the ground that “an essential aspect of the ... statute of repose is the right to be free from the burdens of trial”). We therefore have jurisdiction to review the district court’s denial of Cremers’ anti-SLAPP motion.
C. Probability of Success
To resist a motion to strike pursuant to California’s anti-SLAPP law, Batzel must demonstrate a probability that she will prevail on the merits of her complaint. Cal.Civ.Proc.Code § 425.16. The district court held that Batzel had made such a showing, and absent 47 U.S.C. § 230,8 we would be inclined to agree.
Section 230(c)(1) specifies that “[n]o provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.” The provision thereby set limitations on liability under state law for postings on the Internet and other computer networks. The district court declined to extend the legislative grant of immunity pursuant to § 230(c) to Cremers and the Network, holding that the Network is not “an internet service provider” and therefore is not covered by the statute. We do not agree with the district court’s reading of § 230.
1. Section 230(c)
We begin with a brief survey of the background of § 230(c), as that background is useful in construing the statutory terms here at issue.
Title V of the Telecommunications Act of 1996, Pub.L. No. 104-104, is known as the “Communications Decency Act of 1996” [the “CDA” or “the Act”]. The primary goal of the Act was to control the exposure of minors to indecent material. See Pub.L. No. 104-104, Title v. (1996); see also H.R.Rep. No. 104-458, at 81-91 (1996); S.Rep. No. 104-230, at 187-193 (1996); S.Rep. No. 104-23, at 9 (1995). Parts of the Act have since been struck down as unconstitutional limitations on free speech, see Reno v. ACLU, 521 U.S. 844, 117 S.Ct. 2329, 138 L.Ed.2d 874 (1997); United States v. Playboy Ent. Group, 529 U.S. 803, 120 S.Ct. 1878, 146 L.Ed.2d 865 (2000), but the section at issue here, § 230, remains intact.
Section 230 was first offered as an amendment by Representatives Christopher Cox (R-Cal.) and Ron Wyden (D-Ore.). See 141 Cong. Rec. H4860 (August 4, 1995). The specific provision at issue here, § 230(c)(1), overrides the traditional treatment of publishers, distributors, and speakers under statutory and common law. As a matter of policy, “Congress decided not to treat providers of interactive computer services like other information providers such as newspapers, magazines or television and radio stations, all of which may be held hable for publishing or distributing obscene or defamatory material written or prepared by others.” Blumenthal v. Drudge, 992 F.Supp. 44, 49 (D.D.C.1998). Absent § 230, a person who pub*1027lished or distributed speech over the Internet could be held liable for defamation even if he or she was not the author of the defamatory text, and, indeed, at least with regard to publishers, even if unaware of the statement. See, e.g., Stratton Oakmont, Inc. v. Prodigy Services Co., 1995 WL 323710 (N.Y.Sup. May 24, 1995) (pre-Communications Decency Act case holding internet service provider hable for posting by third party on one of its electronic bulletin boards).9 Congress, however, has chosen to treat cyberspace differently.10
Congress-made this legislative choice for two primary reasons. First, Congress wanted to encourage the unfettered and unregulated development of free speech on the Internet, and to promote the development of e-commerce. Section 230(a), “Findings,” highlights that:
(3) The Internet and other interactive computer services offer a forum for a true diversity of political discourse, unique opportunities for cultural development, and myriad avenues for intellectual activity.
(4) The Internet and other interactive computer services have flourished, to the benefit of all Americans, with a minimum of government regulation.
§ 230(a). Similarly, the listed policy objectives of the section include:
(1) to promote the continued development of the Internet and other interactive computer services and other interactive media;
(2) to preserve the vibrant and competitive free market that presently exists for the Internet and other interactive computer services, unfettered by Federal or State regulation.
§ 230(b).
Consistent with these provisions, courts construing § 230 have recognized as critical in applying the statute the concern that lawsuits could threaten the “freedom of speech in the new and burgeoning Internet medium.” Zeran v. America Online, Inc., 129 F.3d 327, 330 (4th Cir.1997). “Section 230 was enacted, in part, to maintain the robust nature of Internet communication, and accordingly, to keep government interference in the medium to a minimum.” Id.; see also Ben Ezra, Weinstein, and Co. v. America Online, Inc., 206 F.3d 980, 985 n.3. Making interactive computer services *1028and their users liable for the speech of third parties would severely restrict the information available on the Internet. Section 230 therefore sought to prevent lawsuits from shutting down websites and other services on the Internet.
The second reason for enacting § 230(c) was to encourage interactive computer services and users of such services to self-police the Internet for obscenity and other offensive material, so as to aid parents in limiting their children’s access to such material. See § 230(b)(4); see also 141 Cong. Rec. H8469-70 (Statements of Representatives Cox, Wyden, and Barton); Zeran, 129 F.3d at 331; Blumenthal, 992 F.Supp. at 52. We recognize that there is an apparent tension between Congress’s goals of promoting free speech while at the same time giving parents the tools to limit the material their children can access over the Internet. As a result of this apparent tension, some commentators have suggested that the Fourth Circuit in Zeran imposed the First Amendment goals on legislation that was actually adopted for the speech-restrictive purpose of controlling the dissemination of content over the Internet. See, e.g., Ballon § 42.05[3][B][iv]. These critics fail to recognize that laws often have more than one goal in mind, and that it is not uncommon for these purposes to look in opposite directions. The need to balance competing values is a primary impetus for enacting legislation. Tension within statutes is often not a defect but an indication that the legislature was doing its job. See, e.g., United States v. Kalustian, 529 F.2d 585, 588 (9th Cir.1975) (describing dual and somewhat competing purposes of the Federal wiretap statute of both protecting individual privacy and combating crime).
So, even though the CDA overall may have had the purpose of restricting content, there is little doubt that the Cox-Wyden amendment, which added what ultimately became § 230 to the Act, sought to further First Amendment and e-commerce interests on the Internet while also promoting the protection of minors. See 141 Cong. Rec. H8469-72 (Statements of Representatives Cox, Wyden, Lofgren, and Goodlatte).11 Fostering the two ostensibly competing purposes here works because parents best can control the material accessed by their children with the cooperation and assistance of Internet service providers (“ISPs”) and other providers and users of services on the Internet.12 Section 230(b)(4) describes this goal; “It is the policy of the United States ... to remove disincentives for the development and utilization of blocking and filtering technologies that empower parents to restrict their children’s access to objectionable or inappropriate online material.” § 230(b)(4). *1029Some blocking and filtering programs depend on the cooperation of website operators and access providers who label material that appears on their services.13
Without the immunity provided in Section 230(c), users and providers of interactive computer services who review material could be found liable for the statements of third parties, yet providers and users that disavow any responsibility would be free from liability. Compare Stratton Oakmont, 1995 WL 323710 (holding a service provider liable for speech appearing on its service because it generally reviewed posted content) with Cubby, Inc. v. CompuServe, Inc., 776 F.Supp. 135 (S.D.N.Y.1991) (holding a service provider not liable for posted speech because the provider was simply the conduit through which defamatory statements were distributed).
In particular, Congress adopted § 230(c) to overrule the decision of a New York state court in Stratton Oakmont, 1995 WL 323710. Stratton Oakmont held that Prodigy, an Internet access provider that ran a number of bulletin boards, could be held responsible for libelous statements posted on its “Money Talk” bulletin board by an unidentified person. Id. The court relied on the fact that Prodigy held itself out as a service that monitored its bulletin boards for offensive content and removed such content. Id. at *2, *4. Prodigy used filtering software and assigned board leaders to monitor the postings on each bulletin board. Id. at *l-*2. Because of Prodigy’s active role in monitoring its bulletin boards, the court found, Prodigy was a publisher for purposes of state libel law and therefore could be held hable for any defamatory statements posted on the website. Id. at *4.
Although Stratton was a defamation case, Congress was concerned with the impact such a holding would have on the control of material inappropriate for minors. If efforts to review and omit third-party defamatory, obscene or inappropriate material make a computer service provider or user hable for posted speech, then website operators and Internet service providers are likely to abandon efforts to eliminate such material from their site. See S.Rep. No. 104-230, at 194 (1996) (“One of the specific purposes of [Section 230] is to overrule Stratton Oakmont v. Prodigy and any other similar decisions. ...”); H.R. Conf. Rep. No. 104-458, at 194 (1996) (“The conferees beheve that [decisions such as Stratton Oakmont] create serious obstacles to the important federal pohcy of empowering parents to determine the content of communications their children receive through interactive computer services.”); 141 Cong. Rec. at H84691-70 (statement of Rep. Cox) (referring to disincentives created by Stratton Oakmont decision); see also Zeran, 129 F.3d at 331 (emphasizing that § 230 was adopted to overrule Stratton Oakmont, and to provide incentives to self-regulate the dissemination of offensive material); *1030Harvey L. Zuckman et al., Modern Communication Law 615 (1999) (observing that it is “crystal clear that [Section 230 was] designed to change the result in future cases like Stratton Oakmont”).14
2. Application to Cremers and the Museum Security Network
To benefit from § 230(c) immunity, Cremers must first demonstrate that his Network website and listserv qualify as “provider^] or user[s] of an interactive computer service.” § 230(c)(1) (emphasis added). An “interactive computer service” is defined as “any information service, system, or access software provider that provides or enables computer access by multiple users to a computer server, including specifically a service or system that provides access to the Internet and such systems operated or services offered by libraries or educational institutions.” § 230(f)(2).
The district court concluded that only services that provide access to the Internet as a whole are covered by this definition. But the definition of “interactive computer service” on its face covers “any” information services or other systems, as long as the service or system allows “multiple users” to access “a computer server.” Further, the statute repeatedly refers to “the Internet and other interactive computer services,” (emphasis added), making clear that the statutory immunity extends beyond the Internet itself. §§ 230(a)(3), (a)(4), (b)(1), (b)(2), and (f)(3). Also, the definition of “interactive computer service” after the broad definitional language, states that the definition “includes] specifically a service or system that provides access to the Internet,” § 230(f)(2) (emphasis added), thereby confirming that services providing access to the Internet as a whole are only a subset of the services to which the statutory immunity applies.15
There is, however, no need here to decide whether a listserv or website itself fits the broad statutory definition of “interactive computer service,” because the language of § 230(c)(1) confers immunity not just on “providers” of such services, but also on “users” of such services.16 § 230(c)(1).
*1031There is no dispute that the Network uses interactive computer services to distribute its on-line mailing and to post the listserv on its website. Indeed, to make its website available and to mail out the listserv, the Network must access the Internet through some form of “interactive computer service.” Thus, both the Network website and the listserv are potentially immune under § 230.
Critically, however, § 280 limits immunity to information “provided by another information content provider.” § 230(c)(1). An “information content provider” is defined by the statute to mean “any person or entity that is responsible, in whole or in part, for the creation or development of information provided through the Internet or any other interactive computer service.” § 230(f)(3). The reference to “another information content provider” (emphasis added) distinguishes the circumstance in which the interactive computer service itself meets the definition of “information content provider” with respect to the information in question. The pertinent question therefore becomes whether Smith was the sole content provider of his e-mail, or whether Cremers can also be considered to have “creat[ed]” or “develop[ed]” Smith’s e-mail message- forwarded to the listserv.17
Obviously, Cremers did not create Smith’s e-mail. Smith composed the email entirely on his own. Nor do Crem-ers’s minor alterations of Smith’s e-mail prior to its posting or his choice to publish ■the e-mail (while rejecting other e-mails for inclusion in the listserv) rise to the level of “development.” As we have seen, a central purpose of the Act was to protect from liability service providers and users who take some affirmative steps to edit the material posted. Also, the exclusion of “publisher” liability necessarily precludes liability for exercising the usual prerogative of publishers to choose among proffered material and to edit the material published while retaining its basic form and message.
The “development of information” therefore means something more substantial than merely editing portions of an email and selecting material for publication.18 Because Cremers did no more than select and make minor alterations to Smith’s e-mail, Cremers cannot be considered the content provider of Smith’s e-mail for purposes of § 230.19
*1032The partial dissent does not register any disagreement with this interpretation of the definition of “information content provider” or with the observation that immunity for “pubhsher[s]” indicates a recognition that the immunity will extend to the selection of material supplied by others. It nonetheless simultaneously maintains that 1) a defendant who takes an active role in selecting information for publication is not immune; and 2) interactive computer service users and providers who screen the material submitted and remove offensive content are immune. See post at 1039 and n. 5. These two positions simply cannot logically coexist.
Such a distinction between deciding to publish only some of the material submitted and deciding not to publish some of the material submitted is not a viable one. The scope of the immunity cannot turn on whether the publisher approaches the selection process as one of inclusion or removal, as the difference is one of method or degree, not substance.
A distinction between removing an item once it has appeared on the Internet and screening before publication cannot fly either. For one thing, there is no basis for believing that Congress intended a one-bite-at-the-apple form of immunity. Also, Congress could not have meant to favor removal of offending material over more advanced software that screens out the material before it ever appears. If anything, the goal of encouraging assistance to parents seeking to control children’s access to offensive material would suggest a preference for a system in which the offensive material is not available even temporarily. The upshot is that the partial dissent’s posit concerning the limitations of § 230(c) immunity simply cannot be squared with the statute’s language and purposes, whatever merit it, or a variant of it, might have as a policy matter. See n. 19, swpra.
In most cases our conclusion that Crem-ers cannot be considered a content provider would end matters, but this case presents one twist on the usual § 230 analysis: Smith maintains that he never “imagined [his] message would be posted on an international message board or [he] never would have sent it in the first place.” The question thus becomes whether Smith can be said to have “provided” his e-mail in the sense intended by § 230(c). If the defamatory information is not “provided by another information content provider,” then § 230(c) does not confer immunity on the publisher of the information.
“[Provided” suggests, at least, some active role by the “provider” in supplying the material to a “provider or user of an interactive computer service.” One would not say, for example, that the author of a magazine article “provided” it to an interactive computer service provider or user *1033by allowing the article to be published in hard copy off-line. Although such an article is available to anyone with access to a library or a newsstand, it is not “provided” for use on the Internet.
The result in the foregoing example should not change if the interactive computer service provider or user has a subscription to the magazine. In that instance, the material in question is “provided” to the “provider or user of an interactive computer service,” but not in its role as a provider or user of a computer service. The structure and purpose of § 230(c)(1) indicate that the immunity applies only with regard to third-party information provided for use on the Internet or another interactive computer service. As we have seen, the section is concerned. with providing special immunity for individuals who would otherwise be publishers or speakers, because of Congress’s concern with assuring a free market in ideas and information on the Internet. If information is provided to those individuals in a capacity unrelated to their function as a provider or user of interactive computer services, then there is no reason to protect them with the special statutory immunity.
So, if, for example, an individual who happens to operate a website receives a defamatory “snail mail” letter from an old Mend, the website operator cannot be said to have been “provided” the information in his capacity as a website service.20 Section 230(c)(1) supplies immunity for only individuals or entities acting as “provider[s]” or “user[s]” of an “interactive computer service,” and therefore does not apply when it is not “provided” to such persons in their roles as providers or users.
The situation here is somewhat more complicated than our letter example, because Smith did provide his e-mail over the Internet and transmitted it to the Network, an operator of a website that is an user of an interactive computer service. Nevertheless, Smith contends that he did not intend his e-mail to be placed on an interactive computer service for public viewing.
Smith’s confusion, even if legitimate, does not matter, Cremers maintains, because the § 230(c)(1) immunity should be available simply because Smith was the author of the e-mail, without more.
We disagree. Under Cremers’s broad interpretation of § 230(c), users and providers of interactive computer services could with impunity intentionally post material they knew was never meant to be put on the Internet. At the same time, the creator or developer of the information presumably could not be held hable for unforeseeable publication of his material to huge numbers of people with whom he had no intention to communicate. The result would be nearly limitless immunity for speech never meant to be broadcast over the Internet.
Supplying a “provider or user of an interactive computer service” with immunity in such circumstances is not consistent ■with Congress’s expressly stated purposes in adopting § 230. Free speech and the development of the Internet are not “promote[d]” by affording immunity when providers and users of “interactive computer *1034service[s]” knew or had reason to know that the information provided was not intended for publication on the Internet. Quite the contrary: Users of the Internet are likely to be discouraged from sending e-mails for fear that their e-mails may be published on the web without their permission.
Such a scenario is very different from the bulletin boards that Congress had in mind when passing § 230. When a user sends a message to a bulletin board, it is obvious that by doing so, he or she will be publicly posting the message. Here, by contrast, Smith claims that he had no idea that the Network even had a listserv. His expectation, he says, was that he was simply sending a private e-mail to an organization informing it of his concern about Batzel’s artwork, and, he insists, he would not have sent the message had he known it would be sent on through the listserv. Absent an incentive for service providers and users to evaluate whether the content they receive is meant to be posted, speech over the Internet will be chilled rather than encouraged. Immunizing providers and users of “interactive computer service[s]” for publishing material when they have reason to know that the material is not intended for publication therefore contravenes the Congressional purpose of encouraging the “development of the Internet.”
Immunizing individuals and entities in such situations also interferes with Congress’s objective of providing incentives for providers and users of interactive computer services to remove offensive material, especially obscene and defamatory speech. Far from encouraging such actions, immunizing a publisher or distributor for including content not intended for Internet publication increases the likelihood that obscene and defamatory material will be widely available. Not only will on-line publishers be able to distribute such material obtained from “hard copy” sources with impunity, but, because the content provider him or herself never intended publication, there is a greater likelihood that the distributed material will in fact be defamatory or obscene. A person is much more likely to exercise care in choosing his words when he knows that those words will be widely read. This is true not only for altruistic reasons but also because liability for defamation attaches only upon publication. In the current case, Smith claimed exactly that: He told Cremers that if he had known his e-mail would be posted, he never would have sent it. The congressional objectives in passing § 230 therefore are not furthered by providing immunity in instances where posted material was clearly not meant for publication.
At the same time, Congress’s purpose in enacting § 230(c)(1) suggests that we must take great care in determining whether another’s information was “provided” to a “provider or user of an interactive computer service” for publication. Otherwise, posting of information on the Internet and other interactive computer services would be chilled, as the service provider or user could not tell whether posting was contemplated. To preclude this possibility, the focus should be not on the information provider’s intentions or knowledge when transmitting content but, instead, on the service provider’s or user’s reasonable perception of those intentions or knowledge. We therefore hold that a service provider or user is immune from liability under § 230(c)(1) when a third person or entity that created or developed the information in question furnished it to the provider or user under circumstances in which a reasonable person in the position of the service provider or user would conclude that the information was provided for publication on the Internet or other “interactive computer service.”
*1035It is not entirely clear from the record whether Smith “provided” the e-mail for publication on the Internet under this standard. There are facts that could have led Cremers reasonably to conclude that Smith sent him the information because he operated an Internet service. On the other hand, Smith was not a subscriber to the listserv and apparently sent the information to a different e-mail account from the one at which Cremers usually received information for publication. More development of the record may be necessary to determine whether, under all the circumstances, a reasonable person in Cremers’ position would conclude that the information was sent for internet publication, or whether a triable issue of fact is presented on that issue.
We therefore vacate the district court’s order denying Cremers’s anti-SLAPP motion and remand to the district court for further proceedings to develop the facts under this newly announced standard and to evaluate what Cremers should have reasonably concluded at the time he received Smith’s e-mail. If Cremers should have reasonably concluded, for example, that because Smith’s e-mail arrived via a different e-mail address it was not provided to him for possible posting on the listserv, then Cremers cannot take advantage of the § 280(c) immunities. Under that circumstance, the posted information was not “provided” by another “information content provider” within the meaning of § 230. After making such an inquiry, the district court must then evaluate whether Batzel adequately has demonstrated a probability that she will prevail on the merits of her complaint under California’s anti-SLAPP statute.21
IV
Batzel argues that the district court erred in granting Mosler’s motion for summary judgment. According to Batzel, Cremers was acting as Mosler’s agent and therefore should be held vicariously hable for Cremers’s alleged defamation. We reject this contention.22
“Agency is the fiduciary relationship that arises when one person (a ‘principal’) manifests assent to another person (an ‘agent’) that the agent shall act on the principal’s behalf and subject to the principal’s control, and the agent manifests assent or otherwise consents .so to act.” Restatement (Third) of Agency § 1.01 (Tentative Draft); Desuza v. Andersack, 63 Cal.App.3d 694, 699, 133 Cal.Rptr. 920 (1976). In order for Mosler to be held vicariously hable for the torts of Cremers on a theory of agency, Mosler must have had the ability to control Cremers’s activities. See, e.g., McCollum v. Friendly Hills Travel Ctr., 172 Cal.App.3d 83, 91, 217 Cal.Rptr. 919 (1985).
Batzel failed to present any genuine issue of material fact as to whether Mosler had control over Cremers’s management of the Network. The sponsorship agreement executed by Mosler and Cremers requires, in essence, that Cremers include Mosler’s trademark in future editions of the Network. It does not give Mosler any right to control what is published by Cremers. The agreement states that Cremers will “maintain ownership and control of ah aspects of [the Network’s] content and oper*1036ation.” The agreement further states that Mosler “understands that the Museum Security Network is editorially independent and maintains full control of All [sic] editorial content.” Cremers also posted on the Network’s website a statement that Mos-ler “understand^] that the [Network] must remain editorially independent and they respect that.” Rather than evincing Cremers’s assent to control by Mosler, these statements disclaim control. Without the element of control, Batzel’s vicarious liability argument fails. See Desuza, 63 Cal.App.3d at 699, 133 Cal.Rptr. 920 (citations omitted) (“The right of the alleged principal to control the behavior of the alleged agent is an essential element which must be factually present in order to establish the existence of agency-”).
The fact that Mosler provided Cremers with financial support does not support an inference that Mosler possessed practical control of Cremers’s editorial content. Sponsorship alone is insufficient to render the sponsor the guarantor of the truth of all statements made in a publication. See Matson v. Dvorak, 40 Cal.App.4th 539, 549, 46 Cal.Rptr.2d 880 (1995) (holding that a party whose “only contribution to a political campaign is financial, and who is not involved in the preparation, review or publication of campaign literature, cannot be subjected to liability in a defamation action for statements contained in that literature”).
We likewise reject the argument that Mosler’s continued sponsorship of the Network after Cremers published Smith’s statements should give rise to liability. Although a principal is liable when it ratifies an originally unauthorized tort,23 the principal-agent relationship is still a requisite, and ratification can have no meaning without it. Because Cremers was not acting as Mosler’s agent at the time he published Smith’s statements, Mosler’s continued funding of the Network did not amount to ratification of the tort.
The district court accordingly did not err in ruling that Mosler cannot be held vicariously liable for Cremers’ actions because there was no principal-agent relationship between Mosler and Cremers. We affirm the district court’s grant of summary judgment in favor of Mosler.
Costs on appeal are awarded to Mosler. All other parties are to bear their own costs.
VACATED IN PART, AFFIRMED IN PART, AND REMANDED.
. We are including this e-mail address because the precise e-mail address used is relevant to our later discussion.
. "A listserv is an automatic mailing list service that amounts to an e-mail discussion group.... Subscribers receive and send messages that are distributed to all others on the listserv.... Messages may be automatically posted to the listserv or filtered through the list owner (who may elect not to post messages that are off topic or inappropriate). A listserv, unlike a newsgroup, involves one-to-many messaging, rather than the use of distributed message databases." Ian C. Ballon, E-Commerce and Internet Law: Treatise with Forms, Glossary of Terms, at 30 (2001).
. The Association is described in the complaint as "a business entity of unknown form" in the Netherlands. The Association is not involved in the present appeal.
. Mosler is an Ohio-based company that manufactures security devices.
. Cal.Civ.Proc.Code § 425.16.
. Cremers’s notice of appeal requests in the alternative that we treat the notice of appeal as a petition for mandamus. So treated, we deny the petition because the question is one that may be reviewed adequately on appeal from final judgment. See Bauman v. United States Dist. Ct., 557 F.2d 650, 657 (9th Cir.1977).
. This report may be found on the internet at http://www.leginfo.ca.gov/pub/99-00/bill/ asm/ab_l 651-1700/ab_l 675_cfa_l 9990701_ 075825_sen_comm.html.
. All further references are to 47 U.S.C. unless otherwise noted.
. A bulletin board, in the Internet context, is "a computer-based system giving users access from remote terminals to text and programs contributed by one another and stored centrally." Oxford English Dictionary 642 (2d Ed. 1989). Bulletin boards allow users to post messages on the Internet and for others to view them, much like a bulletin board in the off-line world.
. We note that some commentators have suggested that Congress intended for § 230(c) to override only publisher, not distributor, liability. See, e.g., Ballon § 42.05[3][B]; Susan Friewald, Comparative Institutional Analysis in Cyberspace: The Case of Intermediary Liability for Defamation, 14 Harv. J.L. & Tech 569, 637—42 (2001) (courts should leave distributor liability intact when applying § 230); David R. Sheridan, Zeran v. AOL and the Effect of Section 230 of the Communications Decency Act upon Liability for Defamation on the Internet, 61 Alb. L. Rev. 147, 167-72 (1997) ("[W]hen Congress said ‘publisher,’ it meant ‘publisher,’ and not ‘distributor’.”). Batzel’s complaint refers to Cremers as the "publisher” of Smith's e-mail, and she has not argued that Cremers should be treated as a distributor, perhaps because the standard for distributor liability is generally less favorable for plaintiffs. See Zeran v. America Online, Inc., 129 F.3d 327, 331 (4th Cir.1997). We therefore need not decide whether § 230(c)(1) encompasses both publishers and distributors. We do note that, so far, every court to reach the issue has decided that Congress intended to immunize both distributors and publishers. See Zeran, 129 F.3d at 331-34; Ben Ezra, Weinstein, and Co. v. America Online, Inc., 206 F.3d 980, 986 (10th Cir.2000); Doe v. America Online, Inc., 783 So.2d 1010, 1013-17 (Fla.2001).
. We also note that at least some members of Congress suspected that much, if not most, of the Act would be struck down as unconstitutional. See 141 Cong. Rec. H8470 (the Cox-Wyden amendment sought to provide aid during the "flood of legal challenges” likely to prevent the rest of the Act from having any effect); see also Lawrence Lessig, Code and Other Laws of Cyberspace 174 (1999) (referring to the Act as "[a] law of extraordinary stupidity [that] practically impaled itself on the First Amendment.”). That position turned out to be prescient. See Reno, 521 U.S. 844, 117 S.Ct. 2329,; Playboy Ent. Group, 529 U.S. 803, 120 S.Ct. 1878. Quite possibly in anticipation that other aspects of the Act would not survive, the Cox-Wyden amendment attempted to control content distributed to minors in a manner consistent with the First Amendment, by encouraging Internet companies to police the Internet for themselves or to assist parents in doing so. See 141 Cong. Rec. H8469-72.
. An ISP provides its subscribers with access to the Internet. Such service providers include, for example, America Online, commonly referred to as "AOL," and Earthlink, as well as numerous other providers.
. The vast amount of material on the Internet makes it difficult for an independent board to rate all of the material, as the Motion Picture Association of America can do for the much smaller number of movies distributed. Some filtering systems therefore rely on self-labeling. See R. Polk Wagner, Filters and The First Amendment, 83 Minn. L. Rev. 755, 70-65 (1999). Zoning, another approach to controlling content on the Internet by creating child-free zones accessible only with digital identification certificates and age verification, also works best with the cooperation of website operators and service providers. See Lawrence Lessig, What Things Regulate Speech: CDA 2.0 v. Filtering, 38 Jurimetrics J. 629 (1998); see also Reno, 521 U.S. at 886-97, 117 S.Ct. 2329 (O'Connor, J., concurring in part and dissenting in part) (describing zoning approach to protecting minors from harmful Internet content).
. Although not relevant to the current case, § 230(c)(2) further encourages good Samaritans by protecting service providers and users from liability for claims arising out of the removal of potentially "objectionable” material from their services. See § 230(c)(2). This provision insulates service providers from claims premised on the taking down of a customer’s posting such as breach of contract or unfair business practices. Cf. 17 U.S.C. § 512(g)(1) (providing similar protection for service providers who take down material alleged to violate copyright laws); H.R.Rep. No. 105-551, at 25 (1998).
. Other courts construing § 230(f)(2) have recognized that the definition includes a wide range of cyberspace services, not only internet service providers. See, e.g., Gentry v. eBay, Inc., 99 Cal.App.4th 816, 831 & n. 7, 121 Cal.Rptr.2d 703 (2002) (on-line auction website is an "interactive computer service”); Schneider v. Amazon.com, 108 Wash.App. 454, 31 P.3d 37, 40-41 (2001) (on-line bookstore Amazon.com is an "interactive computer service”); Barrett v. Clark, 2001 WL 881259 at *9 (Cal.Sup.Ct.2001) (newsgroup considered an "interactive computer service”); see also Ben Ezra, 206 F.3d at 985 (parties conceded that AOL was an interactive computer service when it published an on-line stock quotation service); Zeran, 129 F.3d at 330 (AOL assumed to be interactive computer service when it operated bulletin board service for subscribers); Blumenthal, 992 F.Supp. at 49-50 (parties conceded that AOL was an "interactive computer service” even when it functioned as the publisher of an online gossip column).
.We note that several courts to reach the issue have decided that a website is an "interactive computer service.” See, e.g., Carafano v. Metrosplash.com, Inc., 207 F.Supp.2d 1055, 1065-66 (C.D.Cal.2002); Gentry, 99 Cal.App.4th at 831, 121 Cal.Rptr.2d 703 (holding *1031that website is an interactive computer service); Schneider, 31 P.3d at 40-41 (same).
. We do not separately analyze Cremers’s statement regarding contacting the FBI. Bat-zel has presented no evidence that the statement was untrue. No action for defamation can go forward absent such evidence. See Cort v. St. Paul Fire and Marine Ins. Companies, Inc., 311 F.3d 979, 985 (9th Cir.2002) ("An essential element of defamation is that the publication in question must contain a false statement of fact.") (quoting Savage v. Pac. Gas & Elec. Co., 21 Cal.App.4th 434, 444, 26 Cal.Rptr.2d 305 (1993)).
. Other courts have agreed that the "exercise of a publisher's traditional editorial functions — such as deciding whether to publish, withdraw, postpone or alter content" do not transform an individual into a "content provider" within the meaning of § 230. Zeran, 129 F.3d at 330; see also Ben Ezra, 206 F.3d at 985-86 (defendant not "information content provider” although it edited and altered stock quotations provided by third party); Blumenthal, 992 F.Supp. at 49-53 (defendant not "information content provider” although it had editorial control over content in gossip column); Schneider, 31 P.3d at 39-43 (website not "information content provider" although it had strict posting guidelines and could edit and republish posted material).
. As .other courts have pointed out, the broad immunity created by § 230 can sometimes lead to troubling results. See, e.g., Blumenthal, 992 F.Supp. at 51-52 (expressing opinion that "[i]f it were writing on a clean slate,” AOL Would be liable for defamation when it had editorial control over the defama*1032tory material). For example, a service provider that cannot be held liable for posting a defamatory message may have little incentive to take such material down even if informed that the material is defamatory.
One possible solution to this statutorily created problem is the approach taken .by Congress in the Digital Millennium Copyright Act ("Digital Act”). The Digital Act includes immunity provisions, similar to those of the Communications Decency Act, that protect service providers from liability for content provided by third parties. The Digital Act, however, unlike the Communications Decency Act, provides specific notice, take-down, and put-back procedures that carefully balance the First Amendment rights of users with the rights of a potentially injured copyright holder. See 17 U.S.C. §§ 512(c) and (g); see also H.R.Rep. No. 105-551, at 52-62 (1998) (describing the DMCA’s take-down and put-back procedures); Comm. Print, Section-By-Section Analysis of H.R. 2281, at 25-36 (Sept. 1998) (same). To date, Congress has not amended § 230 to provide for similar take-down and put-back procedures.
. We note that the partial dissent's interpretation of § 230(c)(1) has a related defect. Under the partial dissent's theory, for example, content that is sent to an interactive service provider or user labeled "for consideration for inclusion on your website” would not be "information provided by another," if the website operator makes selection decisions amongst the material submitted. Yet, almost any speaker of the English language would agree that the content was “provided" to the website by the third party.
. Cremers asks us also to evaluate Batzel's invasion of privacy and emotional distress claims. The district court, however, did not address either of these claims. They therefore are not properly before us on this interlocutory appeal.
. We review a grant of summary judgment de novo and consider the facts in the light most favorable to Batzel. See Oliver v. Keller, 289 F.3d 623, 626 (9th Cir.2002) (citing Clicks Billiards, Inc. v. Sixshooters, Inc., 251 F.3d 1252, 1257 (9th Cir.2001)).
. See, e.g., Murillo v. Rite Stuff Foods, 65 Cal.App.4th 833, 852, 77 Cal.Rptr.2d 12, (1998); Coats v. Construction & Gen. Laborers Local No. 185, 15 Cal.App.3d 908, 914, 93 Cal.Rptr. 639 (1971); McChristian v. Popkin, 75 Cal.App.2d 249, 256, 171 P.2d 85 (1946).