The arbitration clause in your terms of service: what actually makes it hold up when an advertiser sues

When a banned advertiser calls a lawyer instead of a competitor

Every classifieds operator eventually bans an advertiser: for a stolen photo, a chargeback pattern that looks like fraud, a listing that violates the content policy, or a verification check that came back wrong. Most of those advertisers move on to a competing site within the week. A few call a lawyer instead, and what happens next depends on a single clause almost nobody thinks about until the day it is tested: the arbitration clause in the terms of service the advertiser agreed to when they signed up.

If that clause holds up, the dispute becomes a private, individual proceeding, usually resolved faster and cheaper than a lawsuit, and it stays that way even if nine other advertisers were banned for the same reason last month. If it does not hold up, the operator is defending a lawsuit in whatever court the advertiser's lawyer picked, possibly a small claims court two states away, possibly a court that lets the case be filed as a class action on behalf of every advertiser the platform ever banned under the same policy.

Nearly every classifieds site already has arbitration language somewhere in its terms, usually copied from a template along with the rest of the document. That language is not the problem. Courts almost never strike down an arbitration clause because of a specific word choice inside it. They strike it down because the operator cannot prove the advertiser actually agreed to it, in a way a judge will accept as real consent rather than a sentence nobody read.

This piece is about that proof: what makes a court treat an arbitration clause as binding, what a real ruling from last year shows about the specific evidence that decided it, and what a classifieds operator's signup flow needs to produce so the same evidence exists the day someone tests it.

Why the wording is rarely what decides the case

Operators who think about their arbitration clause at all tend to focus on the words inside it: mandatory versus optional, which arbitration provider to name, whether disputes get resolved in the operator's home state. Those choices matter once a dispute is already headed to arbitration. They are not what a court checks first.

The first question a judge asks is whether the specific person suing actually agreed to the clause, and courts have drawn a sharp line between two ways that agreement gets formed online. A "browsewrap" agreement is one where the terms sit behind a link, usually in the footer, and using the site is treated as agreement to whatever is in them; nobody has to click anything to accept. A "clickwrap" agreement, or something close to it, requires an actual action, a checkbox or a button under text that says what clicking it means.

The difference is not academic. In Nguyen v. Barnes & Noble, a federal appeals court held that a footer link to terms of use, with no checkbox and no required click, did not bind a website user to the arbitration clause buried in those terms, because there was no evidence he had actual notice of it and nothing about the purchase process required him to do anything that counted as agreeing. The retailer's arbitration clause was real, clearly drafted, and completely unenforceable against that specific customer, because the site never made him do anything that proved he had seen it.

Contrast that with Meyer v. Uber Technologies. Uber's registration screen did not use a separate checkbox either, but it displayed text stating that creating an account meant agreeing to the terms of service, with a hyperlink to them, directly under the button the user had to press to finish signing up. A different federal appeals court held that design gave reasonably conspicuous notice and that clicking the button was an unambiguous act of agreement, enforceable even though nothing was technically "checked." The clause itself was similar in both cases. What differed was whether the site could show the user did something specific that counted as saying yes.

For an advertiser signup flow, this means the location and design of the acceptance step matters more than any sentence inside the arbitration clause. A link at the bottom of the page is close to worthless as proof. A checkbox, or a button directly under a clear statement of what agreeing means, is what a court is actually looking for.

What a 2025 ruling shows about the evidence that actually wins

Contract-formation logic like Nguyen and Meyer explains the general rule. A ruling from January 2025, in the long-running In re Google Digital Advertising Antitrust Litigation, shows what it looks like when a business actually has to prove it in court, on facts close to what a classifieds operator will eventually face: an arbitration clause added after advertisers had already signed up, not one baked into a first-time signup flow.

Google amended its advertising terms of service in September 2017 to add an arbitration clause to an advertiser base that already existed. When two advertisers later sued and Google moved to compel arbitration, the court initially denied that request in 2024 because the facts on the record were too thin. Google came back after conducting fact discovery specifically on how the amendment had been rolled out, and in January 2025 the court granted the motion for those two advertisers.

What changed between the loss and the win was not the clause. It was the evidence. Google's notice process had used an email, a public blog post, and an interstitial alert shown to advertisers when they logged into their accounts, each pointing to the new terms; the amended terms let advertisers opt out of arbitration within thirty days through an online form; and Google's business records showed, for each of the two advertisers, the exact date they had accepted the new terms and that neither had used the opt-out form. The court treated that combination, a real notice campaign plus a provable acceptance record plus proof of no opt-out, as enough to make the amendment binding.

The lesson an operator should take from that ruling is not about arbitration specifically. It is that a business can build the most defensible notice process imaginable and still lose in court if it cannot produce a record proving that process actually happened, for that specific advertiser, on that specific date. Google won not because its lawyers wrote a clever clause in 2017, but because eight years later it could still pull up who saw the new terms, when, and whether they opted out.

Building a signup and renewal flow that produces that record

The practical translation for a classifieds directory is a short list of changes to how advertisers sign up and renew, not to what the arbitration clause says. At the point an advertiser creates an account or purchases a listing, the acceptance step needs to be an action, a checkbox or a button directly under a plain sentence describing what agreeing means, not a link sitting quietly at the bottom of the page.

That action needs to be logged: the timestamp, the specific version of the terms in force at that moment, and which action the advertiser actually took. A terms-of-service page that only shows the current text is not enough on its own; if the clause changes later, the operator needs to be able to show what an advertiser agreed to on the date they agreed to it, which means keeping the old version tied to the record of that acceptance, not just the newest one on the live site.

The harder case is an existing advertiser base, the same situation Google was in. Adding or changing an arbitration clause after advertisers already have active accounts cannot rely on silence counting as agreement, and a single email is a weaker record than the multi-channel notice that worked in the Google case. An interstitial shown at the next login, with a link to the specific change and, ideally, a genuine window to opt out, gives the operator both a stronger legal position and a paper trail if the amendment is ever tested. The same discipline that matters for canceling a subscription without an argument over the refund applies here: a clear, logged, unambiguous action beats a well-worded policy nobody can prove anyone read.

Most of this evidence never actually goes in front of a judge. Its real value shows up earlier, when an advertiser's lawyer looks at a demand letter response that includes a timestamped acceptance record, a version of the terms tied to that date, and proof no opt-out was used, and decides a private arbitration is a better use of a client's money than a contested motion to compel that the operator is positioned to win.

What the clause cannot do, and what it is actually worth

An arbitration clause governs disputes between the operator and an advertiser. It has no effect on a subpoena, a CyberTipline report, or any other obligation the operator owes to a court or a government agency; those run on entirely separate rules. It also cannot be used to route every category of claim into arbitration by contract alone. US federal law now carves sexual assault and sexual harassment claims out of mandatory pre-dispute arbitration regardless of what a contract says, so an operator should not assume a blanket clause reaches every possible dispute a platform could face.

Even within ordinary contract disputes, a properly formed arbitration clause is not automatically bulletproof. Courts can still strike down a specific clause under ordinary state-law unconscionability doctrine, most often when the arbitration cost the clause imposes is so disproportionate to the size of the claim that using it becomes impractical for the person on the other side. That risk is highest exactly where a classifieds advertiser base actually sits: a mix of incorporated companies and individual, one-person accounts, and courts read a one-sided cost clause against an individual advertiser far less forgivingly than the same clause against a company. A clause that makes the advertiser front the arbitrator's fees, rather than splitting them the way most commercial arbitration providers default to, is building in exactly the defect a court is most likely to use to throw the whole thing out.

What a properly formed clause does deliver, when it is built correctly, is substantial: the Supreme Court's Concepcion decision confirmed that a class-action waiver paired with an arbitration clause is generally enforceable under federal law, which is the piece that actually matters at scale for a directory with thousands of advertisers under the same terms. Without it, a single disputed decision to pull a verified badge or a single banning policy applied consistently across a customer base is exactly the kind of pattern a plaintiff's lawyer looks for to build a class claim; with a clause built on real, provable acceptance, that same decision stays a series of separate, private, individually resolved disputes.

None of this requires new legal language. It requires an honest look at the signup screen: is acceptance a click or a link, is it logged with a timestamp and a version, and if the terms ever change for advertisers who already have accounts, is there a real notice process behind it or just an updated page nobody was ever shown. Fix those three things and the clause that was already sitting in the terms of service starts being worth something.