Blog · Review Essay · Last reviewed June 25, 2026

Lurking and the Person Who Became a User

Joanne McNeil's Lurking: How a Person Became a User is a history of online life told from the side of ordinary participants: searchers, posters, readers, profile-makers, pseudonymous observers, harassed targets, community members, and people slowly converted into measurable platform subjects.

The sharper definition is this: userhood is the administrative compression of a person into an account, trace, preference bundle, risk score, audience segment, prompt source, and memory object. McNeil's value is that she keeps asking what parts of personhood should survive that compression.

The AI-era test is a userhood ledger: what the system calls the person, what it records from visible and quiet behavior, what it infers, what it remembers, what it reuses, what it exposes to vendors or agents, and what the person can inspect, delete, refuse, or appeal.

The Book

Lurking: How a Person Became a User was first published by MCD/Farrar, Straus and Giroux in 2020. Macmillan's book page lists the current Picador edition at 304 pages, with the paperback on sale February 23, 2021. The publisher frames the book as a personal history of the internet from the point of view of the user, organized around concerns such as search, safety, privacy, identity, community, anonymity, and visibility.

That framing is the book's real contribution. Many internet histories follow inventors, founders, venture capital, standards bodies, platform wars, or famous collapses. McNeil starts with the people who had to live inside the resulting systems. Her subject is not the heroic web, the ruined web, or the inevitable web. It is the lived web: forums, blogs, search engines, Wikipedia, Friendster, MySpace, Facebook, Reddit, harassment, forgotten accounts, abandoned communities, and the strange endurance of watching without posting.

McNeil is a critic and writer whose work often concerns protocols, standards, network history, web culture, and infrastructure. Her author site identifies Lurking as her 2020 nonfiction book and notes her background in technology criticism, art writing, and institutions such as Eyebeam and the School for Poetic Computation. That mix matters because the book does not treat interfaces as neutral containers. It reads them as social environments with memory, pressure, and moral consequence.

Current Context

As of June 25, 2026, McNeil's user-centered internet history has become an AI-interface problem. People still search, scroll, post, watch, and maintain profiles, but they also prompt answer engines, talk to companions, accept summaries, delegate tasks to agents, generate images, receive personalized feeds, and leave interaction traces that can become model memory, safety telemetry, recommender input, ad targeting, fraud evidence, or training material. The person does not become a user only by posting. Quiet reading, hesitation, deletion, search, refusal, and repeated return can all become signals.

Current governance has partial handles for that transformation. The EU Digital Services Act regulates online-interface design, recommender explanations, and non-profiling recommender options for very large online platforms and search engines. The FTC's 2024 social-media and video-streaming staff report treated mass data collection, weak privacy controls, data retention, automated uses of personal information, and youth safeguards as consumer-protection problems. The EU AI Act's Article 50 adds transparency duties for certain direct AI interactions and synthetic outputs, while NIST's Generative AI Profile gives voluntary risk-management guidance for organizations building or deploying generative systems.

Companion systems make the stakes more personal. The FTC's September 2025 6(b) inquiry into AI chatbots acting as companions asked companies about engagement monetization, character development, safety testing, child and teen impacts, disclosures, age-rule enforcement, and use or sharing of personal information from conversations. Read through McNeil, that inquiry is not only about chatbots. It is about the next form of userhood: the person addressed as friend, confidant, customer, data subject, and retention opportunity at the same time.

Those instruments do not solve userhood. They show where the review object now sits: not only a platform's content policy, and not only a model's output, but the full relation by which a person is made legible, addressable, predictable, simulated, and governable through an interface.

Userhood as a Political Category

The word "user" looks small, almost administrative. In Lurking, it becomes a political category. A person online is invited to search, speak, watch, save, like, follow, block, report, rank, tag, and upload. Each action feels local. Together they create a body of behavior that platforms can sort, monetize, recommend against, expose, hide, or use to shape the next interface.

The stronger point is that userhood is not the same thing as agency. A user can click and still be governed by the menu. A user can consent and still have no practical alternative. A user can customize a feed while the ranking objective, data retention, ad market, moderation queue, and appeal channel remain outside reach. Calling someone a user can honor participation, but it can also hide the fact that the institution has already decided what forms of participation count.

This is where the book belongs beside work on surveillance, platform power, classification, and media theory. The internet did not simply give people new tools for self-expression. It changed what a person had to become in order to be present: an account, a profile, a metric source, a moderation object, a searchable trace, a social graph node, a content producer, a risk surface, a targetable audience, a training signal.

AI makes the category split further. The same person can be a user in the interface, a data subject in privacy law, a recipient of a platform service under the DSA, a deployer or affected person in an AI workflow, a principal delegating authority to an agent, and an unseen subject of someone else's prompt. Governance fails when it protects only the person at the keyboard. McNeil's frame pushes the review outward to everyone made usable by the system.

McNeil is careful about nostalgia. The early web was not pure sanctuary. It had harassment, exclusion, ugliness, status games, and scams. But older online life often left more room for pseudonymity, drift, small publics, temporary selves, and communities that were not yet fully optimized for growth. The later platform web made participation smoother and more compulsory while narrowing the range of acceptable personhood. The AI interface can narrow it again by treating the user as a context source before treating them as a person with boundaries.

The Intelligence of Lurking

Lurking is often treated as failure to participate. McNeil makes it legible as a form of attention. To lurk is to learn the room before speaking, to preserve privacy, to avoid harassment, to gather social knowledge, to resist the demand that every presence become content.

That matters because modern platforms tend to privilege visible participation. Posting, reacting, sharing, rating, and performing identity are easy to measure. Reading quietly is harder to monetize as public selfhood, even when it is the condition that makes a community meaningful. A platform that only recognizes action can mistake silence for absence and exposure for belonging.

The book's best insight is that online life is made by both speakers and witnesses. A forum, blog, subreddit, group chat, or comment section is not only a collection of statements. It is also a surrounding field of people learning norms, remembering conflicts, avoiding danger, and deciding whether the space is worth trusting. Lurkers are part of the social reality even when the interface cannot count them well.

The governance version is a right to low-pressure presence. A person should be able to read before speaking, search before identifying, leave without performance, and decline personalization without losing the room. This is not a demand for consequence-free anonymity. It is a demand that safety, moderation, and abuse prevention not automatically convert every observer into a public performer or every hesitation into product data.

A protected right to lurk is not a license for abuse. It is the difference between privacy-preserving presence and unaccountable harm. A well-governed community can limit harassment, fraud, impersonation, and coordinated manipulation while still allowing people to read, learn, recover, and decide slowly before making themselves visible. That distinction is easy to lose when every safety problem is answered with stronger identity demands and every engagement problem is answered with more pressure to post.

The AI-Age Reading

Read in 2026, Lurking looks less like a history of social media and more like a prehistory of AI-mediated userhood. The person who became a user is now being invited to become a prompt source, training signal, memory object, personalization profile, synthetic-relationship partner, and delegator of action to agents.

Search is the obvious bridge. McNeil's internet begins with people asking questions of systems that appear to answer. AI search and chat interfaces intensify that relationship. The old search box returned ranked paths. The new answer interface may summarize, remember, recommend, draft, decide, and act. The user is no longer just navigating the web. The user is being modeled by a system that may speak back as assistant, companion, tutor, analyst, recruiter, therapist-like listener, or workplace copilot.

The shift from person to user also helps explain the strange intimacy of AI products. A chatbot does not merely collect clicks. It collects uncertainty, longing, preference, embarrassment, intellectual dependency, anger, work context, family detail, and private rehearsal. The more conversational the interface becomes, the more userhood expands from behavior capture into cognitive and emotional capture.

Agents add a second shift: the user becomes a principal whose delegated authority can be spent by software. When an AI system can send, buy, book, publish, summarize, route, or delete, the old account trace becomes an action trace. The safety question is no longer only what the platform inferred. It is what the agent did with the user's permission, what memory shaped that act, what tools were exposed, and what receipt remains for later challenge.

There is also an affected-person problem. A user may ask an agent to summarize a coworker, screen an applicant, message a patient, classify a tenant, or search a child's record. The person at the prompt is not the only person converted into data. Userhood becomes contagious: one person's convenience can make another person's life more legible to a system they never chose.

This does not mean withdrawal is the answer. McNeil's book is not anti-internet. It is anti-amnesia. It asks readers to remember that platforms are built environments, not weather. They have owners, defaults, incentives, categories, retention policies, reporting flows, labor conditions, and cultural habits. AI systems inherit that entire history while adding generation, memory, agency, and synthetic presence.

Governance and Safety

The governance lesson is to protect the person before optimizing the user. A serious platform or AI review should ask what the system collects from visible action, quiet observation, search, dwell time, deletion, blocked drafts, prompts, uploaded files, saved memories, and agent tool use. It should also ask which of those traces become training data, personalization, ranking input, safety evidence, advertising, moderation evidence, law-enforcement response material, or a record exposed to vendors.

The DSA is useful here because it treats interface design, recommender systems, advertising, risk assessment, transparency, and researcher access as governance surfaces. The AI Act's transparency duties matter because AI-mediated userhood can hide automation behind fluent conversation or synthetic media. The FTC sources matter because deceptive design, fake social proof, broad surveillance, weak data minimization, and inadequate youth safeguards are not side issues. They are ways the user is made governable while being told the interface is merely convenient.

A userhood impact review should separate at least six roles: account holder, reader or lurker, content subject, data subject, affected third party, and delegated principal. Each role needs different protections. The account holder may need export and appeal. The lurker may need low-trace access. The content subject may need notice and takedown paths. The data subject may need deletion and purpose limits. The affected third party may need a way to contest an agentic action. The delegated principal needs receipts and revocation.

Useful controls are concrete: data minimization; visible and editable personalization records; memory controls; deletion and export; non-profiled or chronological routes where available; clear AI disclosure; synthetic-media labels where required; scoped agent permissions; receipts for external actions; privacy-preserving pseudonymity; abuse reporting that does not force public exposure; accessible appeal paths; and documented limits on training reuse. A product that cannot explain those controls is asking people to become users before it has earned the right to model them.

The safety standard is also cultural. A healthy online space should make room for witness, hesitation, pseudonymity, small publics, and repair. It should not treat every quiet reader as an engagement failure, every unposted thought as untapped data, every refusal as friction to remove, or every intimate disclosure as future personalization. The right to remain partly unmodeled is a practical design requirement, not a romantic complaint.

Where the Book Needs Friction

Lurking is deliberately personal and essayistic. That is one of its strengths, but also a limit. Readers looking for a comprehensive institutional history of the internet, a technical history of protocols, or a full political economy of platforms will need other books alongside it.

The book's user-centered method can also make infrastructure appear through experience rather than through direct analysis. We often feel design choices before we can name the systems that produced them. That is true to online life, but AI governance needs both sides: the phenomenology of being addressed by machines and the inspectable mechanics of data, models, ranking, moderation, procurement, and labor.

Finally, the book predates the mass deployment of generative AI. Its categories are still useful, but they need extension. A person is no longer only turned into a profile or audience segment. They may be turned into context for a model, raw material for synthetic output, a replicated voice, a memory in a companion system, or the hidden human residue inside an automated service.

What This Changes

The strongest reason to read Lurking now is that it restores the person inside the interface.

Modern systems keep asking people to accept compressed roles: user, account, customer, citizen, member, candidate, viewer, target, risk, engagement unit, prompt. Each role makes some actions easier and others nearly unthinkable. The interface then teaches the role back to the person until the category feels natural.

McNeil's discipline is to ask what becomes of privacy, ambiguity, witness, small publics, pseudonymity, memory, and refusal when every system wants legible participation. That question becomes sharper with AI. A humane AI interface cannot merely be useful, fluent, and personalized. It has to preserve the human right to remain partly unmodeled: to read without performing, ask without being absorbed, receive help without becoming raw material, and participate without being reduced to the term "user."

Source Discipline

This review separates book metadata, author context, published reception, regulatory claims, standards guidance, and interpretation. Macmillan, MCD/FSG, the Macmillan reading guide, and McNeil's author site support book facts and author background. Reviews from Kirkus, Publishers Weekly, Christian Science Monitor, and Esquire support reception context only. EUR-Lex, FTC, and NIST sources support the current governance claims checked for the June 25, 2026 review date.

The AI reading is an application of McNeil's userhood framework, not a claim that the book predicted current model architectures or that AI systems are conscious, divine, or AGI. The claim is narrower: generative interfaces intensify older platform patterns because they make user traces more conversational, intimate, synthetic, memorable, and actionable.

For source discipline, claims about userhood should name the surface. A feed, search box, chatbot, companion, marketplace, dating app, browser agent, moderation queue, and workplace copilot each makes a different kind of user. Evidence should identify the data collected, the retained record, the recommender or model involved, the downstream use, the user controls, and the appeal path.

Regulatory sources should not be flattened. The DSA is EU platform law, not a global internet bill of rights. The EU AI Act's Article 50 duties apply on their legal timeline and do not cover every manipulation of userhood. FTC staff reports and 6(b) studies document regulator findings or information-gathering, not a court judgment against every service. NIST guidance is voluntary risk-management material, not proof of compliance. Those limits are why the page treats governance as evidence discipline, not magic protection.

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