Section 230 Explained in Plain English, Down to the Twenty-Six Words

Section 230 of the Communications Decency Act is the most argued-about law that almost nobody has read. Cable hosts denounce it. Senators promise to repeal it. Your cousin on social media has strong views about it, delivered in capital letters. Ask any of them to recite the key sentence and you will get the same response you get when you ask a man who loves to talk about the Constitution to name the Third Amendment: a long pause, and then a change of subject.

So let's read it. The heart of the law, subsection (c)(1) of section 230 in title 47 of the United States Code, is twenty-six words long. I counted them twice, because I know my audience.

No 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.

That is it. Everything else in the argument, the shouting, the hearings, the lawsuits and the campaign speeches, hangs on that sentence and a second one next to it. Let me take you through how it got there, what the words mean, and what they do not mean. You can then decide for yourself whether to keep it, fix it or burn it down, and you will be in the small minority of people with an opinion who have read the thing.

The problem in 1995: moderate and get sued, or look away and stay safe

To understand why Congress wrote those words, you have to go back to the early commercial internet, when the services you used came with names like CompuServe and Prodigy and your modem made a noise like a cat being run through a wood chipper.

Before 1996, the law of publishers and distributors was old and clear. A newspaper is a publisher. It chooses what to print, and if it prints a libelous statement, it can be sued. A bookstore or a newsstand is a distributor. It sells what others wrote, and it is not liable for the contents of every book on the shelf unless it knew, or had reason to know, that a particular one was defamatory.

Two cases put online services into that framework, and they pointed in opposite directions.

In 1991, in Cubby, Inc. v. CompuServe, a federal court in New York held that CompuServe, which hosted forums without reviewing what was posted, was more like a distributor. It had not known about the allegedly defamatory material, and it was not liable.

In 1995, in Stratton Oakmont, Inc. v. Prodigy Services Co., a New York state trial court reached the opposite result about a different service. Someone had posted on a Prodigy bulletin board called Money Talk that the Long Island brokerage Stratton Oakmont had committed fraud. Stratton Oakmont sued Prodigy. The court found that Prodigy advertised itself as a family-friendly service, used software to screen for offensive language and had guidelines for its message board leaders. Because Prodigy exercised editorial control, the court said, it was acting like a publisher and could be held responsible for the posts it let through.

Think about what that does. Under the two rulings, the service that tried to clean up its forum was more exposed to a lawsuit than the service that did nothing at all. Moderation was a liability. The safest business plan was to look away, which is a great deal if you run a sewer and a lousy deal if you run a community.

Congress noticed, and there was nothing partisan about the noticing.

The law of 1996 and the two men behind this section

The Communications Decency Act was Title V of the Telecommunications Act of 1996, which President Clinton signed on February 8, 1996. Most of the Communications Decency Act was about something else: criminalizing the knowing transmission of indecent material to minors over the internet. The provision we are talking about was a separate amendment, written by Representative Chris Cox, a Republican from California, and Representative Ron Wyden, a Democrat from Oregon. Their idea was to reverse the incentive created by Stratton Oakmont: let services remove objectionable material without turning themselves into the legal author of everything else.

Reading the text: the two subsections that matter

The operative part of section 230 is subsection (c), titled "Protection for 'Good Samaritan' blocking and screening of offensive material." It has two pieces.

Subsection (c)(1): you are not the author of other people's posts

That is the twenty-six-word sentence quoted above. Break it into its working parts:

1. "Provider or user of an interactive computer service." A website, a forum, a comment section, a platform, an internet provider. And also a user, which means the person who shares or reposts someone else's article is covered along with the company that hosts it. People forget that part.
2. "Shall not be treated as the publisher or speaker." The law does not say the service is innocent. It says a court may not treat the service as the one who published or spoke the content. That is a rule about who gets sued.
3. "Of any information provided by another information content provider." This is the limit that people skip. If somebody else created the content, the protection applies. If the service itself created or helped develop the content, it is an "information content provider" with respect to that material, and the shield does not cover it.

So a comment from a reader on a newspaper's website belongs to the reader. The newspaper's own story belongs to the newspaper. The statute draws that line in plain English.

Subsection (c)(2): removing things in good faith is safe

The second piece protects the decision to take material down. It says that no provider or user shall be held liable for "any action voluntarily taken in good faith to restrict access to or availability of material that the provider or user considers to be obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable, whether or not such material is constitutionally protected."

Two phrases in that sentence have caused most of the trouble. One is "good faith," which opponents say platforms do not practice, and the other is "otherwise objectionable," which defenders say is the freedom to run your own house and critics say is a loophole you could drive a truck through. Hold those two phrases in your head. We are going to come back to them in the reform debate, and you will see why conservatives who agree on almost everything else disagree about them.

There is also a third small piece, (c)(2)(B), which protects the tools that let users or services block material, which is the legal foundation for a parental filter.

What the courts did next: Zeran and Reno

The statute is short. The consequences came from the courts, and two decisions in 1997 shaped the whole landscape.

Zeran v. America Online (1997)

After the Oklahoma City bombing in April 1995, an anonymous person posted messages on an America Online bulletin board advertising tasteless T-shirts and other items related to the tragedy, and listed the home telephone number of a man named Kenneth Zeran as the contact. Zeran, who had nothing to do with it, was flooded with angry calls and threats. He asked AOL to remove the posts and sued when he felt it had not acted fast enough.

The Fourth Circuit, in an opinion by Chief Judge J. Harvie Wilkinson III, held that section 230 barred the claim. The argument Zeran made was clever: even if AOL was not a publisher, it was a distributor, and a distributor is liable once it has notice that material is defamatory. The court said no. Distributor liability is just a species of publisher liability, and the statute forbids treating the service as a publisher. Notice-and-takedown liability would also have put every service in the position of having to take down whatever anyone complained about, to avoid being sued, which would chill speech.

That reading has been the foundation ever since. It is why a platform that is told about a defamatory post is generally not liable for leaving it up, though the person who wrote it can still be sued. A supporter will tell you that this is the reason the internet had room for a million small forums. A critic will tell you it means a victim often has no recourse against anyone but an anonymous poster. Both of them are describing the same reading of the same sentence.

Reno v. ACLU (1997)

The rest of the Communications Decency Act did not fare as well. In Reno v. ACLU, decided in June 1997, the Supreme Court struck down the act's provisions criminalizing "indecent" and "patently offensive" communications that might reach minors. Justice Stevens wrote for the Court that those provisions were too broad and too vague, and that the internet deserved the full protection the First Amendment gives to books and newspapers, not the lesser protection given to broadcasting.

This is a fact every discussion of section 230 should open with, and few do. The part of the law that was about censoring the internet was thrown out. The part that survived was the part that gave services protection when they let other people speak, and when they chose to clean their own space. Section 230 is not a gift from the censors. It is what remained after the Supreme Court dismantled the censors' bill.

What section 230 does not do

Now the part that defenders and critics both get wrong. The law has real limits, and they are written into the statute.

1. It does not cover the platform's own content. If a site writes a defamatory article, or materially helps create the unlawful content, the site is a content provider for that material. In Fair Housing Council of San Fernando Valley v. Roommates.com (2008), the Ninth Circuit, sitting en banc, held that a roommate-matching site that required users to answer questions that were themselves discriminatory was partly responsible for the content it elicited.
2. It does not touch federal criminal law. Subsection (e)(1) says the section has no effect on enforcement of federal criminal statutes. A federal prosecutor is not blocked.
3. It does not touch intellectual property law. Under (e)(2), copyright and trademark claims are outside the shield, and copyright has its own notice-and-takedown scheme under the Digital Millennium Copyright Act of 1998.
4. It does not cover certain communications privacy laws, under (e)(4).
5. It has an exception for sex trafficking. In 2018, Congress passed the law commonly called FOSTA, signed in April of that year, which carved out claims related to sex trafficking and the knowing promotion of prostitution.
6. It does not prevent the person who wrote the post from being sued. The speaker is on the hook. The shield protects the service, not the author.

And one thing it does not do that people often claim it does: it does not require neutrality. Nothing in section 230 says that a platform must be politically even-handed to keep its protection. The title of the subsection, "Good Samaritan" blocking, and the "good faith" words in (c)(2) have led some to read in a fairness requirement, but (c)(1) stands by itself, with no good-faith condition, and the courts have applied it that way. If you want to know why conservatives are split on what to do, that is the reason. The law as written does not say what some of its critics wish it said.

The Supreme Court has also so far steered around the biggest questions. In 2023, in Gonzalez v. Google, the justices were asked to decide whether section 230 protects a platform's algorithmic recommendations. The Court sent the case back without deciding that question, after ruling in a companion case, Twitter, Inc. v. Taamneh, that the plaintiffs had not shown the platforms aided and abetted the terrorist attack at issue. So the foundational reading from Zeran still stands, largely because Congress has not changed the text and the Court has not been eager to rewrite it for them.

Why a working family should care

You may feel this is a lawyer's quarrel with no bearing on your life. It does bear on it, and in two directions.

Consider the person who runs a neighborhood forum, a youth league's page or a church's Facebook group. If every unkind comment someone else leaves on a page you run could be your legal problem, you would shut the comments off, and probably the page. The statute is part of the reason ordinary people can host conversation without a law firm on retainer.

And consider the other side of it. When a company with a huge audience decides what stays up and what comes down, it does so with immunity that no newspaper enjoys for its own articles, and it is fair to ask whether that bargain is still working. Conservatives have heard their posts removed, accounts suspended and links throttled, and they have asked whether "otherwise objectionable" means whatever the company's staff happen to find objectionable this month. That is not a crazy question. It is the question the whole reform fight is about. I would argue that it deserves a better answer than "repeal it" or "it's perfect," and I'll make that case in Four Ways to Reform Section 230 and What Each Would Do to Real Websites. The related argument over whether platforms are simply private businesses is in Private Company Is Not the End of the Debate Over Platform Power.

Three myths to retire

Before you argue this with anyone, retire these.

"Section 230 protects platforms only if they are neutral." The text has no neutrality test. Argue for adding one, and I'll hear you out, but do not say it is already there.

"Section 230 makes platforms immune from everything." It does not. Federal crimes, intellectual property, the platform's own content and several other categories are carved out, and the author of a post remains liable.

"Without section 230, there would be no censorship." There would likely be more. A world where the host is liable for what users say is a world where a cautious lawyer deletes anything that could produce a complaint, and the first posts deleted are the ones from people with less power to fight back. That is a prediction, not a certainty, but it is the logic of the Stratton Oakmont problem, and it is why the sponsors wrote the law.

A short reading list, no law degree required

If you want to check me, and you should, the statute is at 47 U.S.C. § 230, and you can read it for free in a few minutes on the website of the U.S. Code at the House of Representatives. The cases are easy to find by name. Read Zeran for the reasoning on notice and takedown, Reno v. ACLU for what the Court said about the internet, and Roommates.com for the limits. If you want to see where Washington's proposals live, search Congress.gov for "Section 230" and read the bill texts yourself. The people who tell you what a bill says on cable are often describing a different bill. For the wider context of free speech cases, see Eight Supreme Court Cases That Explain What the First Amendment Protects and the history of how the right learned to use new media in The Internet Broke the Gatekeepers, and the Right Was Quick to See It.

The bottom line

• Read the sentence. Section 230(c)(1) is twenty-six words: a service is not treated as the publisher or speaker of what someone else posted. That is the whole shield.
• It exists because of a perverse rule. After Stratton Oakmont (1995), a service that moderated could be held liable, while one that looked away could not. Congress fixed the incentive in 1996.
• It has limits. The platform's own content, federal crimes, intellectual property and sex trafficking claims fall outside it, and the person who posted can still be sued.
• It does not require neutrality. If you want a fairness rule, you must persuade Congress to write one and then live with the consequences for every small site in the country.
• Judge any plan by who it hits. Ask what a proposed change does to a church group, a local paper and a small forum, not just to the biggest company in the headlines.