A terms-of-service page is not a contract. It becomes one only when a court is satisfied that the user was given reasonable notice of the terms and did something that signaled agreement. Companies spend weeks drafting limitation-of-liability language and about ten minutes on the checkout screen that decides whether any of it binds anyone.
The pattern in U.S. decisions is consistent enough to design around. Clickwrap agreements — where the user must take an affirmative action, such as ticking a box or clicking a button labeled to convey agreement, next to a visible link to the terms — are generally enforced. Browsewrap, where terms sit behind a footer link and the site simply asserts that continued use means acceptance, often is not. Everything interesting happens in the space between those two.
The assent spectrum, from strongest to weakest
Think of online agreement formats as a spectrum of how much the user had to do. The more affirmative the act, and the closer the notice sits to that act, the more likely enforcement becomes.
| Format | How it works | Typical judicial treatment |
|---|---|---|
| Clickwrap | Separate checkbox or "I agree" button, terms linked adjacent | Generally enforced when the link is visible and the action unambiguous |
| Scroll-wrap | User must scroll through the terms before proceeding | Strongest position; assent is hard to dispute |
| Sign-in-wrap | Button says "Create account"; nearby text says doing so accepts the terms | Enforceable if the notice is conspicuous and adjacent; frequently litigated |
| Browsewrap | Footer link only; continued use asserted as acceptance | Often unenforceable absent proof of actual knowledge |
| In-app pop-up on update | Existing users prompted to accept revised terms | Works if acceptance is required to continue, fails if dismissible |
The doctrinal frame is ordinary contract law, applied to a screen. Electronic form is not the obstacle — federal and state electronic-records statutes settled decades ago that a record or signature cannot be denied effect merely because it is electronic, a topic covered in the guide to electronic signatures and online contract formation. The obstacle is notice. A term buried where no reasonable user would look is treated as never presented at all.
Designing a flow that survives a motion
Product and legal need to look at the same screenshot. These are the details that repeatedly separate enforced agreements from unenforced ones.
- The reference to the terms sits immediately above or beside the action button, not in a footer or on a prior screen.
- The hyperlink looks like a hyperlink — underlined, contrasting color, and labeled "Terms of Service" rather than a vague word like "details."
- The button text and the notice text describe the same act: "By clicking Create Account, you agree to the Terms of Service and Privacy Policy."
- Font size and contrast are comparable to surrounding text. Grey 8-point type under a bright button is a recurring loser.
- The flow works the same on mobile, where the notice is often pushed below the fold by a virtual keyboard.
- Nothing competes for attention at the moment of assent — no promotional overlay, no autoplaying element.
- The screen is captured and versioned, with a timestamp, every time it changes.
Watch out: Interfaces engineered to obscure a commitment attract more than contract risk. The Federal Trade Commission treats manipulative design and unclear disclosure as deception or unfairness in its own right; its business guidance on privacy and security is a reasonable starting point for how regulators read a screen.
Changing terms without making them illusory
Nearly every set of online terms reserves the right to change them. The weak version — "we may modify these terms at any time; changes are effective when posted; your continued use constitutes acceptance" — is exactly the version courts have criticized. Where a company can rewrite its obligations at will, without notice, some courts have held the promise illusory and the modified terms unenforceable, and a few have used that reasoning to strike a newly added arbitration provision.
The reliable structure has four parts: advance notice through a channel the user actually receives (email or in-product prompt, not a silent website edit), a stated effective date that gives the user time to react, a mechanism for the user to reject by closing the account, and an express statement that changes do not apply retroactively to disputes that have already arisen. For paid services, add that pricing and term changes take effect only at renewal.
- Is the change material? Fee structures, dispute resolution, data use, and liability limits are material. Typo fixes are not. Reserve the heavier notice process for the changes that matter.
- How will the user learn of it? Pick a channel you can later prove was used — a mail log, an in-app acknowledgment record.
- What does the user do to accept? A required click on next login is far stronger than silence plus continued use.
- What happens if they refuse? Give a real exit: cancel without penalty, with a prorated refund where money was prepaid.
- Are existing disputes carved out? Say so in the clause. Retroactive reach is where modification clauses most often fail.
The clauses that get read hardest
Winning on assent is only the first step. Individual provisions face their own review, generally under state unconscionability doctrine and consumer-protection statutes. Arbitration and class-waiver provisions draw the most attention; they are usually enforceable when the process is fair and costs are not loaded onto the consumer, and the mechanics are set out in the guide to arbitration clauses and class-action waivers. Forum-selection and choice-of-law clauses are commonly upheld, but not when they make relief practically unavailable.
Liability caps and warranty disclaimers survive most challenges in commercial settings and face more resistance in consumer ones, especially where a statute grants a non-waivable right. Auto-renewal terms now sit under specific state statutes requiring clear disclosure, acknowledged consent, and an easy cancellation path — an area regulators have pursued actively. Fee-shifting clauses that make a losing consumer pay the company's legal costs are frequently struck.
Terms also need to line up with your privacy disclosures. Where a state privacy statute requires specific notices and rights, a terms page that contradicts your privacy notice creates a documented inconsistency; the California Attorney General's CCPA materials illustrate the level of specificity now expected. Businesses that sell to other businesses should keep the consumer-facing terms and the negotiated agreements separate, since SaaS and software contracts allocate risk on entirely different assumptions.
Proving assent three years later
Most terms-of-service fights are decided on a declaration from an engineer describing what a user saw. If your company cannot reconstruct that, you are arguing from a live site that has been redesigned twice since.
Practical step: Keep a versioned archive of every terms document with effective dates, plus dated screenshots or rendered captures of each assent screen, plus per-user records of which version was accepted, when, and from what device. Store them so they outlive the product team that built the flow.
Retention is a governance question, not just a legal one. The NIST Privacy Framework is a useful structure for deciding which records to keep, how long, and under what controls — helpful because assent logs themselves contain personal data and cannot simply be kept forever without a reason.
Quick answers
Do we need a checkbox, or is a notice near the button enough?
A notice adjacent to the button can be enough, and many large platforms rely on it. A separate checkbox is stronger because it removes the argument that the user never noticed the notice. The practical rule: use adjacent notice for low-stakes flows, and require a distinct affirmative act wherever the terms include arbitration, significant fees, or broad data-use rights.
Can we make terms apply to people who never created an account?
Rarely, and only with careful design. A visitor who reads a page has taken no action from which agreement can be inferred, which is why pure browsewrap fails so often. If you need terms to reach non-registered users — for example to govern scraping or API access — put an assent step in front of the relevant functionality rather than relying on a footer link.
We are adding arbitration to existing terms. What is the safest route?
Notify existing users directly, require an affirmative click to continue using the service, give a genuine opt-out window with instructions that do not require a phone call, and state that the change does not apply to disputes already pending or already accrued. Keep the acceptance records per user. Silent posting is the approach most likely to be challenged.
Does a longer, more detailed set of terms help?
Not usually. Length increases the chance a court finds a specific clause surprising or oppressive, and it makes conspicuousness harder to establish. Shorter terms with plainly worded key provisions — and genuinely important items called out on the assent screen itself — tend to fare better than fifteen thousand words nobody claims to have read.
A sensible order of operations
Start with the screen, not the document. Capture every place a user commits to something, and check each against the notice checklist above. Then audit the modification clause for advance notice, an exit, and a no-retroactivity statement. Next, review the high-scrutiny clauses — arbitration, caps, auto-renewal, fee-shifting — against the consumer statutes of the states where your users live. Finally, build the archive: versioned terms, dated screen captures, per-user acceptance logs. For related duties across contracts, data, and intellectual property, the technology, privacy, and IP pathway collects the neighboring guides, including how website accessibility claims can reach the same interfaces from a different direction.