The Terms Are Not a Contract

The Design Brief — How Terms of Service Emerged

The contract, as a legal instrument, was designed to record a genuine meeting of minds. Two parties, with roughly equivalent information and bargaining power, negotiated terms they both understood and to which they both genuinely agreed. The resulting document was the record of that agreement. It could be enforced precisely because it reflected what both parties actually intended.

Terms of service did not emerge from this tradition. They emerged from the mass-market consumer economy of the twentieth century, where producing individual negotiated agreements for millions of transactions was not operationally feasible. The standard-form contract — a pre-written document offered on a take-it-or-leave-it basis — became the dominant legal instrument of consumer commerce. The consumer could not negotiate; she could only accept or decline. Courts accepted this arrangement on the theory that the market would discipline abusive terms: consumers dissatisfied with one company's terms could choose a competitor's.

The digital economy eliminated the last structural constraint on standard-form abuse. When a platform operates in a market with two or three dominant competitors, when switching costs are high due to network effects and data lock-in, and when all major platforms offer substantially similar terms, the theoretical disciplinary power of consumer choice is purely formal. The consumer's choice is between accepting terms she cannot read and cannot negotiate, or forgoing the service entirely. For services that have become essential infrastructure — email, search, social communication — this is not a meaningful choice.

The document that emerged from this history is no longer a contract in any functional sense. It is a liability management instrument, designed by the institution, for the institution's protection, and presented to the consumer as though it were a mutual agreement. Understanding what it actually is requires examining how it grew to be unreadable — and why that growth was not accidental.

The Architecture of Length — How 14,000 Words Is Not an Accident

The average terms of service document for a major digital platform runs approximately 14,000 words. Some are substantially longer: Apple's iTunes terms of service were famously estimated at longer than Shakespeare's Hamlet. Amazon's conditions of use, PayPal's user agreement, Google's terms of service — each runs to thousands of words of dense legal prose, subdivided into dozens of sections and sub-sections, each incorporating by reference additional documents that are themselves thousands of words long.

This length is not the product of thoroughness or precision in the service of the user. It is the product of a specific legal strategy: the creation of constructive notice. A document that addresses every conceivable situation, however improbably, provides the institution with a legal record that the term existed and was available to be read. Length functions as a liability inoculation. Each additional clause is another potential claim foreclosed, another limitation on user rights extended, another obligation shifted from the company to the consumer.

The strategy is self-reinforcing. Regulatory and litigation environments that require specific disclosures produce longer documents, not clearer ones. When a data breach settlement requires a company to disclose data retention practices, the company's legal team adds a clause — it does not simplify the document to make the new disclosure legible. Each new legal requirement produces new text on top of existing text, and the resulting document grows in length while declining in comprehensibility.

The research on optimal comprehension confirms what common sense suggests: legal prose at this length and complexity cannot be meaningfully processed by consumers. A 2016 study by Bakos, Marotta-Wurgler, and Trossen found that fewer than one in a thousand retail software license agreements were opened, and a small fraction of those were read for more than a few seconds. For platform terms of service, the reading rate is lower still. The document's unreadability is not a failure of implementation. It is the feature that makes the document function as its authors intend.

The Legal Fiction — The Constructive Notice Doctrine

American contract law enforces terms of service through a doctrine called constructive notice. The doctrine holds that a party is deemed to have notice of a term if the term was made available to her, regardless of whether she actually read it. Applied to consumer agreements: if the terms were linked on the platform's website and the user clicked “I agree,” she is legally bound by every provision in those terms, including provisions she did not read, could not have understood if she had read them, and would not have accepted if she had understood them.

Constructive notice was developed in property law contexts where the relevant documents were filed in a public registry and the parties were presumed to have both the capacity and the incentive to search the registry before completing a transaction. Applied to consumer software agreements, the doctrine performs a function it was not designed for: it converts the fiction of implied knowledge into legal enforceability. The consumer did not know. The law deems her to have known. The institution proceeds on the basis of the deeming.

Courts have begun to grapple with the limits of this doctrine, particularly in the context of arbitration clauses buried within lengthy terms. The Supreme Court's decision in AT&T Mobility v. Concepcion (2011) upheld class-action waivers in arbitration clauses, a ruling that dramatically expanded the practical scope of terms of service enforcement. Subsequent rulings have been less uniform: some courts have declined to enforce terms that were insufficiently prominent or that the user had no meaningful opportunity to review. But the general doctrine of constructive notice for digital consumer agreements remains robust, and its effect is to make legally enforceable a document that no one reads.

The Carnegie Mellon study that produced the “0% of users read them” finding was methodologically precise: researchers measured actual reading behavior, not self-reported intentions. The finding is not that users intend to skip the terms — it is that the terms are universally skipped, including by users who intend to read them. This is the reading gap as a structural feature, not a behavioral anomaly. And the legal system enforces the gap as though it were a legitimate meeting of minds.

What Terms Actually Say — A Precise Accounting

A systematic content analysis of major platform terms of service reveals a consistent structure: obligations run primarily from the user to the platform; rights run primarily from the platform to itself. This structure is not disguised. It is stated plainly, in legal language designed to be technically accurate while practically impenetrable.

Standard provisions across major platform agreements include: a license granted by the user to the platform to use, reproduce, modify, adapt, distribute, and create derivative works from any content the user uploads or creates on the platform — typically worldwide, royalty-free, and sublicensable. The license is often irrevocable for content the user has shared with others. A unilateral right to modify the terms at any time, with or without notice, and to deem continued use of the platform as acceptance of the modified terms. A broad limitation of liability clause capping the platform's financial exposure to the user at either zero or the amount the user paid in the past twelve months — typically zero for free services. A mandatory arbitration clause requiring the user to resolve any dispute through binding individual arbitration, waiving the right to jury trial and class action. A choice-of-law provision selecting the jurisdiction most favorable to the platform, regardless of where the user is located.

Each of these provisions would be material to a user's decision whether to use the platform. Each is buried in a document no user reads. Each shifts risk, rights, and remedies in the institution's direction. Together, they constitute not a mutual agreement but a unilateral declaration of terms by the institution, ratified through the legal fiction of constructive notice.

The Liability Shield — What the Documents Protect

The core function of a modern terms of service document is not to record what service the platform provides. It is to limit what the platform is responsible for when things go wrong. Every major operational risk the platform faces — data breaches, algorithm-driven harm, content moderation failures, platform outages, financial losses caused by platform errors — is addressed in the terms through a limitation of liability clause, an indemnification clause, or both.

The limitation of liability clause is the document's load-bearing wall. In its standard form, it provides that the platform shall not be liable for any indirect, incidental, special, consequential, or punitive damages, regardless of the cause of action and regardless of whether the platform was advised of the possibility of such damages. It typically caps direct damages at the amount the user paid in the preceding twelve months. For free services, this means the user's maximum recovery is zero, regardless of what the platform does and what harm results.

The indemnification clause runs in the opposite direction: it requires the user to defend and hold harmless the platform from any claims arising from the user's activities on the platform, including claims arising from the platform's own content moderation decisions, algorithm-driven distribution of the user's content, or the platform's use of data the user provided. The user assumes legal exposure for harms that the platform's design choices cause, because the user accepted terms they did not read that shifted that exposure to them.

The result is an asymmetric risk allocation that would be unenforceable in a genuine negotiated contract between parties with roughly equivalent bargaining power. Courts have traditionally refused to enforce unconscionable contract terms — terms that are so one-sided as to be oppressive. The doctrine has had limited application to terms of service, partly because of the constructive notice presumption and partly because courts have been reluctant to second-guess the terms of mass-market agreements. But the liability shield function is the clearest evidence that terms of service are not agreements in any functional sense. They are instruments through which the institution shifts risk to the user while calling the shift an agreement.

Clickwrap vs. Genuine Consent — The Mechanisms and Their Limits

Contract law recognizes two primary forms of consumer agreement to terms of service: clickwrap and browsewrap. Clickwrap agreements require the user to click an explicit acknowledgment — typically an “I agree” or “I accept” button — before accessing the service. Browsewrap agreements rely on a notice somewhere on the website that continued use constitutes acceptance. Courts have generally enforced clickwrap agreements more readily than browsewrap, on the theory that the affirmative click creates a stronger record of assent.

The theory rests on a distinction that has no empirical support. Clicking “I agree” does not indicate that the user has read the terms. The research is unambiguous on this point: the click is a gate to be cleared, not a moment of deliberation. Users click “I agree” at the same rate regardless of what the terms say, including in laboratory studies where researchers have inserted provisions requiring users to assign their firstborn child to the company. The button is a formality. Its presence converts a legal fiction into a legal record, but it does not convert the fiction into a fact.

Genuine consent, as the doctrine has always required in other contexts, involves actual knowledge and voluntary agreement. Medical informed consent — the legal standard governing patient authorization for procedures — requires that the patient understand the nature of the procedure, its risks, its benefits, and the alternatives, and that the patient's agreement be free from coercion. Financial advice regulations in many jurisdictions require that advisors communicate in plain language that clients can understand, and that clients acknowledge comprehension of specific material risks. The constructive notice standard applied to consumer terms of service is the only major area of American law in which legally binding agreements are enforced against parties who demonstrably did not and could not have read them.

This is not a technical or operational gap. It is a policy choice. The legal system has chosen to enforce terms of service through constructive notice because the alternative — requiring genuine comprehension — would impose costs on platform operations and would likely invalidate significant portions of the terms currently in use. The policy choice benefits platforms and imposes costs on users. The user's click registers in the platform's legal record. The user's ignorance of what she clicked does not.

What Reform Would Require

The Legibility Standard developed in CR-005 applies directly to terms of service. Genuine consent to a consumer agreement would require three things that current terms of service uniformly fail to provide: readability at an eighth-grade level; comprehension verification for material terms; and clear presentation of alternatives, including the option to decline and the consequences of declining.

Regulatory reform proposals have clustered around three approaches. The first approach — mandatory plain-language requirements — would require terms of service to be drafted at a specified reading level for consumer-facing platforms. The EU's Digital Services Act moves in this direction for some categories of disclosure, and the Consumer Financial Protection Bureau has issued guidance on plain-language requirements for financial disclosures. Extension of plain-language requirements to consumer platform terms represents a significant but technically achievable step.

The second approach — content regulation of material terms — would prohibit specific categories of provisions that are currently standard: mandatory arbitration for consumer claims below a threshold amount, class-action waivers in consumer contexts, unlimited data licensing rights in excess of what is operationally necessary for the service, and limitation of liability provisions that eliminate all recovery for foreseeable harm. This approach does not require users to read anything — it ensures that what the terms contain cannot be weaponized against users regardless of whether they read it.

The third approach — comprehension verification for specific provisions — would require that users demonstrate understanding of key material terms before those terms become enforceable. This is operationally more complex but addresses the core problem: the terms that harm users most are precisely the terms that are most buried, most technical, and most difficult to understand. A comprehension verification requirement for arbitration waivers, class-action prohibitions, and broad data licensing provisions would substantially reduce the scope of harm achievable through the current constructive notice system. The Consent Record returns to these questions in CR-005, which develops the full Legibility Standard across all four domains.

Sources

• Yannis Bakos, Florencia Marotta-Wurgler, and David Trossen. “Does Anyone Read the Fine Print? Consumer Attention to Standard-Form Contracts.” Journal of Legal Studies 43, no. 1 (2014): 1–35.
• Lorrie Faith Cranor. “Necessary but Not Sufficient: Standardized Mechanisms for Privacy Notice and Choice.” Journal on Telecommunications & High Technology Law 10 (2012): 273–307.
• Aleecia McDonald and Lorrie Faith Cranor. “The Cost of Reading Privacy Policies.” I/S: A Journal of Law and Policy for the Information Society 4, no. 3 (2008): 543–568.
• Margaret Jane Radin. Boilerplate: The Fine Print, Vanishing Rights, and the Rule of Law. Princeton University Press, 2013.
• Omri Ben-Shahar and Carl E. Schneider. More Than You Wanted to Know: The Failure of Mandated Disclosure. Princeton University Press, 2014.
• AT&T Mobility LLC v. Concepcion, 563 U.S. 333 (2011). Upholds class-action waivers in arbitration clauses within consumer agreements.
• Mohr v. Williams, 95 Minn. 261 (1905). On consent as a prerequisite for lawful medical intervention.
• Federal Trade Commission. Bringing Dark Patterns to Light. FTC Staff Report, September 2022.
• Electronic Privacy Information Center (EPIC). Consumer Agreements: The Unread Contract. Policy Brief, 2021.
• EU Digital Services Act (Regulation (EU) 2022/2065). Article 14 requirements on transparency in terms and conditions for very large online platforms.