Blog · Review Essay · Last reviewed June 25, 2026

To Save Everything, Click Here and the Politics of Solutionism

Evgeny Morozov's To Save Everything, Click Here is a useful irritant for the AI age because it refuses the clean story that social problems become simpler once software can measure, optimize, gamify, or automate them. Its central warning is not that technology is useless. It is that institutions can use technical cleverness to avoid political judgment.

For this review, technological solutionism means the move that turns a contested social, civic, or institutional problem into a technical optimization task before authority, affected people, alternatives, appeal, and refusal have been examined. The danger is not the tool. The danger is the narrowed problem frame that makes the tool appear inevitable.

The Book

To Save Everything, Click Here: The Folly of Technological Solutionism was published by PublicAffairs in 2013. Hachette's current listing gives the on-sale date as March 5, 2013, with 432 pages, and frames the book around smart technologies, big data, quantified behavior, gamification, civic behavior, and the danger of recasting moral and political dilemmas as problems of technical efficiency.

Morozov was already known for The Net Delusion, a critique of naive internet-freedom politics. In the Carnegie Council event around this book, Joanne Myers placed the new argument in continuity with that earlier work: the same platforms sold as tools of liberation could also serve surveillance, propaganda, and control. To Save Everything, Click Here widens the target from foreign-policy cyber-utopianism to a broader habit of treating networked technology as the default answer to public life.

The book's key terms are intentionally abrasive. "Internet-centrism" names the mistake of treating "the Internet" as a single coherent force with a destiny of its own. "Solutionism" names the urge to redescribe complex social, civic, and ethical questions as technical problems awaiting cleaner interfaces, better incentives, and more data.

Current Context

As of June 25, 2026, Morozov's critique is no longer only a cultural warning about apps and smart gadgets. AI governance has made the problem operational: public agencies, schools, employers, platforms, and vendors increasingly need records showing why an automated intervention is lawful, necessary, proportionate, tested, monitored, and contestable. The question is not whether software can do something useful. It is whether the system's problem definition deserves institutional authority.

The EU AI Act makes part of this visible. Article 5 prohibits several AI practices that sit near the line between helpful intervention and coercive optimization, including certain manipulative or deceptive techniques, vulnerability exploitation that causes or is reasonably likely to cause significant harm, social scoring, and some biometric and emotion-inference uses. Article 27 adds a fundamental-rights impact assessment duty for specified deployers of high-risk AI systems. Those are legal thresholds, not Morozov's politics, but they confirm the basic governance grammar: some "solutions" require a prior account of purpose, affected people, risks, oversight, and complaint paths.

U.S. federal AI policy is narrower but still relevant. OMB M-25-21 requires covered agencies to complete AI impact assessments before deploying high-impact AI uses, maintain minimum risk-management practices, monitor systems through the lifecycle, and offer timely human review or appeal where appropriate. OMB M-25-22 turns adoption into a procurement question by requiring acquisition evidence around testing, monitoring, data use, portability, and vendor lock-in. Canada's public-sector Algorithmic Impact Assessment tool is another useful anchor: it uses 65 risk questions and 41 mitigation questions to determine the impact level of an automated decision system. Together with NIST's AI Risk Management Framework and ISO/IEC 42005 on AI system impact assessment, the current context supports Morozov's demand that "fixing" be treated as accountable institutional action, not a product mood.

Interface design has also become a governance object. The FTC's dark-patterns staff report treats deceptive or manipulative interface practices as consumer-protection concerns, including tactics that make people spend money or share data they did not intend to share. The European Commission's Digital Services Act materials frame large online platforms as accountability subjects rather than neutral interface providers. Morozov's older language of solutionism now meets a practical regulatory question: when does design assistance become hidden steering?

Solutionism as Institutional Evasion

The strongest part of Morozov's argument is that solutionism often begins before a product is built. It begins when an institution decides what kind of problem it is willing to recognize. If corruption becomes a dashboard problem, obesity becomes a tracking problem, education becomes a ranking problem, crime becomes a prediction problem, or citizenship becomes an engagement problem, then the political field has already been narrowed.

This is why the book remains relevant beyond its 2013 examples. The danger is not merely that a given app fails. The danger is that the app's frame becomes the institution's frame. A messy disagreement becomes a parameter. A conflict over values becomes an optimization target. A public question becomes a user-experience problem.

That move is attractive because it promises action without collective argument. The interface can be shipped while the moral dispute remains unresolved. The sensor can gather data while the legitimacy of surveillance goes undebated. The nudge can alter behavior while nobody has to admit that power is being exercised.

In AI terms, solutionism is a classification error about the problem itself. A hiring tool may be framed as reducing bias while leaving job requirements, credential inflation, disability accommodation, and worker bargaining power untouched. A benefits chatbot may be framed as access while leaving underfunding, eligibility complexity, and appeal burdens untouched. A wellness model may be framed as prevention while turning structural stress into individual compliance data. The technical system does not merely solve within a frame; it helps choose the frame.

A source-disciplined review of a proposed system should therefore begin with the non-technical alternatives. More staff, simpler rules, better pay, public records, accessibility repair, legal aid, collective bargaining, local discretion, or a real appeal path may solve the problem more honestly than an optimization layer. If the only options under consideration are vendor products, the solutionist narrowing has already happened.

The Politics of Friction

Morozov is especially valuable on friction. The standard technology story treats friction as waste: delay, ambiguity, hypocrisy, inefficiency, awkwardness, discretion, and human mess. The book asks whether some of that friction is protective. A society without friction may be easier to administer, but it may also be easier to steer.

Privacy is one example. A person may technically consent to tracking, but once tracking becomes the normal way to prove health, productivity, innocence, or civic virtue, refusal becomes suspicious. The Guardian's contemporary review highlighted this point through the social pressure created when some people publicize self-tracking and others decline. Individual choice becomes a collective coercion channel.

Democracy also depends on forms of productive inefficiency. Debate is slow. Due process is slow. Ambiguity preserves room for judgment. Hypocrisy can be a vice, but forced transparency can also destroy the private space in which people change their minds, dissent safely, or keep institutions from owning their whole identity.

The governance mistake is to treat all friction as a latency problem. Some friction is evidence. A complaint, delay, exception, conflicting account, unclear category, or human override may reveal that the system's representation is too thin. Seeing Like a State makes the same point from another angle: simplification becomes dangerous when the map outranks the people it maps.

That does not mean every inconvenience is sacred. A humiliating form, inaccessible website, needless appointment, duplicate proof requirement, or opaque appeal queue can be bad friction. The useful distinction is whether the friction protects agency, context, and accountability, or merely protects institutional convenience. Solutionism often gets this wrong by abolishing the visible inconvenience while preserving the power relation underneath.

The AI-Age Reading

AI makes Morozov's critique sharper because it gives solutionism a fluent interface. Earlier solutionism often arrived as apps, sensors, dashboards, badges, quantified-self devices, recommender systems, and social platforms. AI can wrap the same administrative ambition in conversation: a tutor, coach, caseworker, compliance assistant, therapist-like companion, workplace copilot, HR screener, civic chatbot, or benefits guide.

The risk is not only error. It is problem laundering. An underfunded public service can become an AI triage workflow. A punitive workplace can become a productivity assistant. A surveillance regime can become safety analytics. A content platform can become personalized civic information. A school can turn assessment crisis into detector deployment. The model makes the intervention feel modern while the institution avoids the deeper question: what is this system allowed to decide about people?

Large language models also intensify the temptation to confuse explanation with accountability. A system can produce a calm paragraph explaining a decision, but the explanation may not reveal the policy choice, the training data, the procurement incentive, the risk threshold, the appeal path, or the person responsible. A polite rationale is not due process.

The book also helps analyze AI companions and persuasive agents. If behavior can be shaped through continuous feedback, memory, personalization, scoring, and intimacy, then "helpfulness" can become a political category. The question is not simply whether the system helps the user achieve a goal. It is who chose the goal, who benefits from the change in behavior, and what kinds of refusal remain available.

Agentic systems add another layer. A chatbot can reframe a problem; an agent can execute the reframing by filing a form, routing a case, escalating an alert, purchasing a service, changing a record, or issuing a notice. When a solutionist frame becomes delegated action, governance has to inspect tool permissions, credentials, logs, rollback paths, and the human authority that can stop the workflow.

Governance and Safety

The safety implication is that solutionism should be reviewed before deployment, not after the dashboard is built. A serious review asks: what problem is being defined, who defined it, which alternatives were rejected, what metrics will stand in for success, what power changes hands, which people will be measured or nudged, and what forms of appeal, refusal, and rollback exist?

The useful artifact is a problem-frame record. It should name the proposed system, affected population, institutional purpose, legal authority, non-technical alternatives, data sources, target metric, expected benefit, foreseeable harms, human oversight role, complaint path, procurement owner, vendor dependencies, and stop condition. This connects Morozov's critique to algorithmic impact assessments, AI procurement, AI governance, and algorithmic recourse.

Procurement is where solutionism often hardens into fact. Once a contract, integration, staff training plan, and vendor workflow exist, the question shifts from "Should this be automated?" to "How do we manage the automated system we already bought?" Buyers should require documentation, audit rights, model-change notice, data-use limits, export rights, human fallback, incident reporting, and the right to suspend or terminate when the promised fix creates harm.

For public services, schools, workplaces, health settings, and civic platforms, the core control is contestability. Affected people need notice that a system shaped the outcome, enough information to challenge the relevant record or category, and a human path with authority to change the result. Without that, AI solutionism turns institutions into smooth one-way interfaces: easy to use, hard to answer.

Where the Book Needs Friction

To Save Everything, Click Here can be exhausting. Morozov's polemical style is part of the book's force, but it sometimes makes the target too broad. Kirkus called it a useful corrective while noting the heavy hand. That is fair. Some technical interventions do reduce harm, widen access, expose corruption, assist disabled users, improve coordination, or make public services less humiliating.

The better reading is not anti-technology. It is anti-enchantment. A technological intervention should have to answer institutional questions: What problem definition does it smuggle in? What behavior does it normalize? What power does it hide? What evidence would prove it harmful? Who can refuse it? Who can appeal? Who maintains it? Who profits if ambiguity disappears?

That test preserves the book's best insight without turning critique into reflex. Software is not automatically domination. But neither is efficiency automatically care.

The missing operational test is accessibility. A technical fix may be justified when it gives disabled users access, reduces bureaucratic humiliation, catches errors, translates public information, shortens harmful delays, or makes an institution more answerable. The question is whether the system expands agency and recourse, or simply makes the existing command structure more efficient.

What This Changes

The book belongs in this catalog because it gives a vocabulary for a recurring pattern: a system promises to make reality legible, then treats the legible version as the only version worth governing. In the AI era, that pattern appears in scoring systems, synthetic tutors, wellness apps, work dashboards, recommender feeds, risk models, agentic workflows, and chat interfaces that translate institutional power into friendly text.

The practical lesson is to inspect the problem frame before admiring the solution. If a deployment begins by making people easier to rank, monitor, route, predict, discipline, or persuade, then technical success may deepen the original harm. A good system should preserve appeal, context, refusal, plural values, and accountable human judgment.

Morozov's title still works because the fantasy remains alive: every broken institution will be saved by a click, a model, a metric, a sensor, a prompt, or an agent. The harder work is less glamorous. It means deciding which problems should remain political, which frictions should remain human, and which efficiencies should be refused because they make the wrong world easier to run.

Source Discipline

This review separates book metadata, Morozov's argument, secondary criticism, current governance context, and this site's interpretation. Hachette/PublicAffairs and Carnegie Council support the book and author context. Kirkus and The Guardian support reception context. EU AI Act, OMB, Canada, NIST, ISO, FTC, and Digital Services Act sources support the current legal, policy, standards, and interface-governance claims checked on June 25, 2026.

Those current sources are not treated as endorsements of Morozov. They are evidence that problem definition, manipulation, impact assessment, procurement, interface design, and lifecycle risk management have become governance questions. A law, OMB memorandum, standard, staff report, and book review have different force; the page keeps those categories separate.

This page makes no claim that any AI system is conscious, divine, or AGI. The claim is institutional: a system can produce power by narrowing the problem, collecting data, optimizing a proxy, and making refusal harder, without needing an inner life.

Sources

Book links are paid affiliate links. As an Amazon Associate I earn from qualifying purchases.

Buy on Amazon Browse Books

Return to Blog · Return to Books