Start free trial
Trending New Genres Authors
EnglishEnglish
EspañolSpanish
简体中文Chinese
繁體中文Chinese (Traditional)
FrançaisFrench
DeutschGerman
日本語Japanese
PortuguêsPortuguese
ItalianoItalian
한국어Korean
РусскийRussian
NederlandsDutch
العربيةArabic
PolskiPolish
हिन्दीHindi
Tiếng ViệtVietnamese
SvenskaSwedish
ΕλληνικάGreek
TürkçeTurkish
ไทยThai
ČeštinaCzech
RomânăRomanian
MagyarHungarian
УкраїнськаUkrainian
IndonesiaIndonesian
DanskDanish
SuomiFinnish
БългарскиBulgarian
עבריתHebrew
NorskNorwegian
HrvatskiCroatian
CatalàCatalan
SlovenčinaSlovak
LietuviųLithuanian
SlovenščinaSlovenian
СрпскиSerbian
EestiEstonian
LatviešuLatvian
فارسیPersian
മലയാളംMalayalam
தமிழ்Tamil
اردوUrdu
Searching...
SoBrief
The AI Con

The AI Con

How to Fight Big Tech’s Hype and Create the Future We Want
Behind every AI promise is a system, a workforce, and someone left to absorb its failures.
by Emily M. Bender 2025 288 pages
3.69
1k+ ratings
Listen
Immersive
V2.3
Amazon Kindle Audible
Try Full Access for 3 Days
Unlock listening & more!
Continue

Key Takeaways

Treat “AI” as a sales pitch, then inspect what it automates.

A marketable umbrella. The authors use “AI” for varied technologies sold as if they shared one humanlike intelligence. “AI hype” is the inflated promise that creates urgency and fear of missing out, helping attract sales and investment. The label groups together automated decisions, classification, recommendations, translation, and generated media, obscuring what each system actually does.

Look beneath the label. At a US Senate forum, leaders discussed p(doom), the supposed probability that machines might cause human extinction. The authors redirect attention to present harms: Detroit police arrested Robert Williams after a facial recognition search matched him to an image in a database of 49 million photographs. They do not reject all automation, noting that well-tested tools such as spell-checkers can be useful.

Analysis

The key move is taxonomic: replacing a glamorous umbrella term with a description of the task makes it easier to ask who chose the system, who benefits, and who bears its mistakes. That is especially useful in public debate, where one impressive chatbot demo can lend undeserved credibility to very different systems. The caution is that a broad label can still help people discuss shared infrastructure and incentives. Precision works best not by banning the word, but by unpacking it whenever a consequential decision is at stake.

Fluent text is not a mind, and confidence is not evidence.

Prediction, not intention. A large language model, or LLM, learns patterns from huge collections of text and repeatedly predicts a likely next token to extend a prompt. Neural networks let these systems scale that pattern matching, but fluent output does not establish that they understand the world. Reinforcement learning from human feedback, or RLHF, uses human ratings to favor responses judged more suitable. It polishes output, but does not create comprehension.

Our minds complete it. Joseph Weizenbaum’s 1960s chatbot ELIZA followed simple rules to imitate a therapist, yet users and professionals still imagined a thoughtful mind behind its replies. People instinctively interpret language as evidence of someone’s communicative intent. The authors call LLMs “stochastic parrots ”: systems that reproduce language patterns without understanding or communicative purpose.

Analysis

The distinction between linguistic fluency and mental understanding is easy to miss because conversation is one of the main ways people encounter other minds. The ELIZA example shows that this reflex predates modern chatbots and is not simply a failure of technical knowledge. Still, a tool need not possess intentions to be useful: pattern prediction can assist narrowly defined tasks. The relevant question is therefore not whether a system is secretly conscious, but whether its output is reliable enough for a particular use and whether a person remains responsible for checking it.

Automation can cheapen workers before it eliminates their jobs.

A familiar labor bargain. The Luddites were skilled British textile workers, not enemies of invention. They resisted factory owners using new machines to replace artisans with lower-paid workers and impose punishing conditions. The same conflict appeared in Hollywood, where writers and actors struck against proposals to use generated scripts and digital replicas. Writers won contract protections against being required to use AI, along with disclosure rules and pay gains.

People power the machines. “Ghost work” means hidden human labor that labels training data, filters harmful outputs, or intervenes when systems fail. ImageNet, a major image dataset, took two and a half years and nearly 50,000 workers across 167 countries to label 14 million images. Kenyan contractors who filtered harmful material for OpenAI reportedly earned less than two dollars a day. Automation can threaten one job while relying on another worker’s invisible labor.

Analysis

The historical comparison shifts attention from whether a machine can perform a task to who controls the resulting productivity gains. Even when jobs remain, automation can reduce worker discretion, increase monitoring, and push people into less secure roles. The hidden workforce also complicates the claim that machines simply replace human labor. A useful test is to follow the labor through the whole system, from data preparation to error correction, rather than counting only the visible jobs affected.

Do not let austerity disguise service cuts as technological progress.

Austerity gets an interface. Governments and companies often sell automation as a way to widen access, while the underlying problem is too few qualified people and too little funding. Pittsburgh’s Allegheny Family Screening Tool assigns child welfare cases a risk score using records from public agencies. Such a score can be used to justify surveillance and family separation instead of connecting families to support. Black and Indigenous families already face disproportionate exposure to these systems.

A plausible voice can fail. After the National Eating Disorders Association replaced hotline workers with a chatbot called Tessa, users received advice about weight loss and calorie counting that could worsen disordered eating. The bot was withdrawn, and the hotline was taken offline. The central question is not merely whether a system responds, but whether human care, accountability, and meaningful ways to challenge harmful decisions remain.

Analysis

The strongest insight is that a technical substitute does not repair an institutional shortage. A chatbot may answer more people per hour, yet access is not the same as care, especially when a person needs judgment, trust, or help contesting a decision. This distinction also sharpens policy analysis: the relevant comparison is not an imperfect tool against an ideal human service, but the proposed tool against better-funded human support. Where that support is withheld, automation can make neglect look like modernization.

Test systems on real tasks, real people, and costly failure cases.

A score is not suitability. Evaluation measures performance on a selected task and dataset; it does not automatically prove that a system is safe in another setting. The choice of whose cases appear, and which errors count, shapes what the result means. A voice system tested on casual Midwestern English cannot simply be presumed reliable for distressed callers using other dialects.

Match evidence to stakes. A study of Epic’s sepsis alert system examined 38,455 hospitalizations. The system missed 1,709 of 2,552 sepsis cases, while triggering alerts for 6,971 patients. The “Clever Hans effect,” named for a horse that seemed to do arithmetic by reading human cues, describes systems that succeed for reasons other than the ability being tested. Passing a bar exam likewise does not establish competence to practice law.

Analysis

Evaluation is not just a technical checkpoint. It is a decision about which people and failures count, and whether the measured task resembles the real one. A high average score can conceal a dangerous concentration of errors in a specific group or setting. The Clever Hans analogy captures a broader lesson: apparent success can reflect shortcuts in a test rather than the capacity people care about. For high-stakes uses, evaluation should include the consequences of mistakes and the possibility of refusing or reversing a decision.

Machine-made media borrows human craft while eroding its ecosystems.

Artifacts are not authorship. Image generators rely on artists’ work collected without meaningful permission, then produce images that may compete for the same commissions. Artist Karla Ortiz argues that creators should receive credit, consent over use, and compensation. In a Society of Authors survey, 26 percent of respondents said they had lost work to generative tools, and 37 percent reported lost income.

Pollution travels. Meta’s scientific text generator Galactica was withdrawn after three days amid fabricated and bigoted output. Synthetic product reviews and fake bylines also appeared in established publications. These cases harm more than individual readers: they crowd submission channels, weaken trust, and divert attention from human art, research, and reporting. Creative work depends on labor, provenance, and communities, not just on producing a plausible artifact.

Analysis

The ecosystem frame explains why an individual output can matter even when a reader spots its flaws. Floods of cheap material can overwhelm editors, search results, and the channels through which new creators find an audience. The case is strongest where generated content competes directly with the people whose work supplied its training material. At the same time, not every automated media task has the same stakes. The useful distinction is between bounded assistance that respects provenance and substitution built on uncredited extraction.

Track present harms and data-center costs instead of debating robot doom.

Two camps, shared premise. Doomers warn that autonomous superintelligence could end humanity; Boosters promise abundance and solutions to vast problems. The authors argue that both treat such a future as inevitable, differing mainly over whether it will be utopian or catastrophic. This focus can push present harms, including wrongful arrests, surveillance, labor exploitation, and state violence, out of view.

The material bill is measurable. Data centers require chips, electricity, water, and mineral extraction. One estimate discussed in the book puts global data-center electricity demand at 1,580 terawatt-hours by 2034, comparable to India’s use. Major technology companies also reported rising emissions during the expansion of AI computing. The authors contend that climate damage is a real, measurable risk, not a science-fiction scenario.

Analysis

The critique usefully asks whose safety counts when public attention is allocated between speculative catastrophe and documented injury. A future risk can deserve attention without displacing accountability for systems already affecting people. The contrast is not a demand to ignore uncertainty, but to avoid treating dramatic uncertainty as inherently more important than visible harm. The environmental argument makes the same point materially: computing is not weightless, and efficiency gains do not settle the question if total use keeps expanding.

Demand proof of inputs, outcomes, and appeal rights before deployment.

Ask for the chain of proof. Before adoption, identify what goes into a system, what comes out, what task it performs, who benefits, who bears errors, and what recourse exists. A photograph cannot establish that someone is a criminal: that category depends on behavior, circumstances, social norms, and law. Accuracy claims should be checked against independent audits and against the setting where the system will actually be used.

Rules already matter. A zero trust AI governance approach means regulators should not rely on companies to police themselves. It calls for enforcing existing laws, prohibiting unacceptable uses, and requiring companies to demonstrate that products are not harmful throughout their lifecycle. Documentation of training data and disclosure when automation is used make responsibility easier to trace and harmful decisions easier to challenge.

Analysis

These questions turn a vague product pitch into a practical audit of power and evidence. They also separate two issues that are often blurred: whether a system predicts accurately and whether its use is legitimate. A highly accurate tool can still serve an unjust policy, while an imperfect tool may be tolerable in a low-stakes, reversible setting. The decisive governance questions include who can inspect the evidence, who can appeal, and whether a human institution has enough capacity to make that appeal meaningful.

Organize and refuse when technology is imposed without community control.

Refusal can change terms. Collective action has won concrete limits: writers negotiated protections against compelled AI use, while artists use Glaze to make their work harder to imitate and Nightshade to interfere with models trained on treated images. Data workers have also organized against precarious platforms. These efforts do not reject every tool; they contest who controls it, whose labor counts, and who sets the conditions of use.

Build with affected people. The authors favor narrowly scoped systems, data consent and minimization, worker protections, public libraries, and community-led projects such as Te Hiku Media’s language technology under Māori guardianship. Personal opt-outs matter, but collective bargaining and enforceable rules reach further. The goal is not a world without technology, but one where communities can choose useful tools and decline harmful ones.

Analysis

The book’s closing emphasis on refusal turns critique into a question of agency: who gets to say no, and who can shape what is built instead? Union contracts and community control offer more durable leverage than individual consumer choices alone. Yet the ability to refuse is uneven, especially for people subject to government or workplace systems they cannot easily escape. That makes collective protections essential. Participation must include the people most affected, not just those who own the infrastructure or fund its development.

Analysis

The book’s strongest contribution is a change in the unit of analysis. Instead of asking whether an abstract entity called “AI” is intelligent, it asks what particular system is being used for, how it was built, and how its consequences are distributed. This reframing links language technology, labor relations, public services, science, and climate costs without pretending they are technically identical. Its recurring cases show how a promise of convenience or efficiency can become a justification for shifting risk onto workers, marginalized communities, and the public.

The argument is most persuasive when it connects a specific failure to the institution deploying the tool. A mistaken prediction is not merely a defective calculation if it influences an arrest, a care decision, or a student’s grades. The relevant questions include who selected the data, who chose the threshold, who benefits from automation, and whether a harmed person can obtain a remedy. This approach also explains why a benchmark score cannot stand in for social legitimacy.

The book’s sharpest rhetoric sometimes risks making unlike systems sound interchangeable. A recommendation feed, a speech recognizer, and a child welfare risk score differ in purpose and danger. The authors’ own insistence on narrow scope offers the best answer: criticize each system at the level of its task, setting, and governance. Nor does a lack of consciousness prove a tool has no utility. It proves that fluency alone cannot support claims of human judgment, empathy, or responsibility.

The result is a useful standard for technological claims: demand evidence about capability, but also ask whether the application is desirable, accountable, and worth its social and material costs. That standard resists both awe and fatalism. It leaves room for useful tools while rejecting the idea that a profitable product deserves automatic adoption or that its future is inevitable.

Summary in 30 Seconds
Language models generate likely text rather than demonstrated understanding, so fluent answers can still be false. Automated systems can reduce pay, hide human labor, and shift risk onto marginalized communities. The Epic sepsis tool missed 1,709 of 2,552 cases while flagging 6,971 others. AI infrastructure also consumes vast resources, with projected data-center electricity demand reaching 1,580 terawatt-hours by 2034. Deployment requires independent testing, disclosure, appeal rights, and collective consent.
Contains spoilers

Last updated:

Report Issue

Review Summary

3.69 out of 5
Average of 1k+ ratings from Goodreads and Amazon.

The AI Con receives mixed reviews, with ratings ranging from 1 to 5 stars. Many readers appreciate the authors' critical perspective on AI hype and their exploration of potential negative impacts. The book is praised for its informative content and accessible writing style. However, some criticize it for being overly skeptical, politically biased, or lacking in technical depth. Readers generally find the book thought-provoking, even if they don't agree with all of its arguments. Some reviewers note that the book's tone can be snarky or repetitive at times.

Your rating:
4.34
310 ratings
Want to read the full book?

Glossary

AI hype

Inflated technology promises

The authors use this term for exaggerated claims that a technology will transform society and must be adopted quickly. Such claims create fear of missing out, attract investment, and encourage people to view unlike automation systems as a single humanlike intelligence. The resulting attention can obscure who profits, who is harmed, and what a system actually does.

large language model (LLM)

Text pattern predictor

A statistical model trained on large collections of text to predict likely next tokens in a sequence. By extending prompts one token at a time, it can produce fluent passages that resemble human writing. The authors stress that modeling patterns in text does not itself provide access to meaning, communicative intent, or a humanlike understanding of the world.

stochastic parrots

Pattern-repeating language models

A term coined by Emily Bender and coauthors to describe large language models that generate plausible language by modeling patterns in their training data. The phrase emphasizes that fluency does not establish understanding or communicative intent. The authors also use it to resist metaphors that treat a text-generating system as a thinking, feeling participant in conversation.

reinforcement learning from human feedback (RLHF)

Human-rated output tuning

A technique in which human workers rate model responses, and those ratings are used to make preferred outputs more likely. In the book’s account, RLHF helps reduce unwanted or harmful responses and makes a system seem more polished. It does not give the model subjective experience or understanding, and the rating work may expose workers to disturbing material.

ghost work

Hidden human data labor

Human work that is essential to the apparent operation of automated systems but is obscured from users. It can include labeling images, rating model responses, filtering harmful content, and intervening when automated services fail. The work is often outsourced through multiple contractors, making labor conditions and the people performing the work difficult for consumers to see.

Clever Hans effect

Success for the wrong reason

A failure of evaluation in which a system appears to demonstrate a target ability but actually relies on unrelated clues. The term refers to a horse said to perform arithmetic by responding to subtle signals from people nearby. In machine learning, a model can score well on a benchmark by exploiting patterns in the test data instead of performing the intended reasoning task.

TESCREAL

A linked ideology cluster

An acronym for Transhumanism, Extropianism, Singularitarianism, Cosmism, Rationalism, Effective Altruism, and Longtermism. The authors use it for overlapping belief systems that influence some AI boosters and Doomers. These ideologies can combine visions of human transformation, space colonization, and future abundance with arguments that present-day suffering may be outweighed by benefits to hypothetical future populations.

zero trust AI governance

Regulation without self-trust

A governance approach that does not assume companies can adequately regulate their own systems. It emphasizes enforcing existing laws, establishing clear prohibitions for unacceptable uses, and placing the burden on companies to demonstrate that their products are safe throughout development and deployment. The framework treats accountability as a regulatory requirement, not a voluntary promise.

FAQ

What is The AI Con: How to Fight Big Tech’s Hype and Create the Future We Want by Emily M. Bender about?

  • Critical examination of AI hype: The book argues that much of what is marketed as artificial intelligence is a "con," designed to benefit tech companies and investors while obscuring real social and ethical issues.
  • Focus on real-world harms: Instead of speculative threats like rogue superintelligence, the book highlights actual harms such as labor exploitation, environmental degradation, surveillance, and misinformation.
  • Interdisciplinary approach: Drawing from history, sociology, labor studies, and environmental science, the book provides a comprehensive critique of AI’s societal impact.
  • Strategies for resistance: The authors offer practical advice for individuals, workers, and policymakers to resist AI hype and advocate for technology that serves communities, not just profits.

Why should I read The AI Con by Emily M. Bender?

  • Unveils the AI hype machine: The book exposes how AI hype manipulates public perception, policymaking, and investment, often distracting from urgent issues like climate change and social inequality.
  • Empowers critical thinking: Readers are equipped with frameworks and questions to critically evaluate AI claims and avoid being misled by marketing or sensationalism.
  • Highlights marginalized perspectives: The book centers the experiences of workers, climate refugees, and marginalized communities most affected by AI’s negative impacts.
  • Offers actionable solutions: It provides guidance on leveraging existing laws, regulatory frameworks, and collective action to hold AI companies accountable.

What are the key takeaways from The AI Con by Emily M. Bender?

  • AI is a marketing term: The book argues that "AI" is often used to sell the illusion of human-like intelligence, masking the reality of automation and statistical pattern matching.
  • Real harms over hype: It emphasizes that current AI technologies cause tangible harms—like job loss, bias, and environmental damage—while hype distracts from these issues.
  • Critical literacy is essential: Readers are encouraged to question AI claims, understand the limitations of technology, and advocate for ethical policies.
  • Human-centered solutions: The authors stress the importance of investing in human expertise, labor rights, and social services rather than relying on automated systems.

How does The AI Con by Emily M. Bender define and critique "AI hype"?

  • Deliberate strategy: AI hype is described as a calculated effort by companies and investors to inflate expectations and obscure the limitations and harms of AI technologies.
  • FOMO and hype cycles: The book explains how fear of missing out (FOMO) drives rapid adoption and investment, fueling a cycle that prioritizes profit over ethics.
  • Misleading metaphors: Anthropomorphizing AI with terms like "superhuman" or "hallucination" misleads the public about what AI can actually do.
  • Distraction from real issues: Both utopian and dystopian narratives about AI distract from addressing present-day harms and reinforce the interests of powerful actors.

What does The AI Con by Emily M. Bender say about the definition and reality of "AI"?

  • AI as a marketing label: The book asserts that "AI" does not refer to a coherent set of technologies but is used to sell the illusion of machine intelligence.
  • Diverse technologies grouped together: It explains that AI includes a range of automation tools—like decision-making systems, recommendation engines, and generative models—that are fundamentally different but lumped under one term.
  • Focus on automation: The authors argue for using "automation" instead of "AI" to clarify that these systems lack understanding or consciousness and simply perform statistical tasks.
  • Obscures human labor: The term "AI" often hides the significant human effort required to make these systems function safely and effectively.

How do large language models (LLMs) like ChatGPT actually work, according to The AI Con by Emily M. Bender?

  • Statistical prediction, not understanding: LLMs generate text by predicting the next word based on patterns in massive datasets, without any comprehension or intent.
  • Neural networks and embeddings: These models use neural networks and word embeddings to capture co-occurrence patterns, producing plausible but not necessarily meaningful text.
  • Human labor behind the scenes: LLMs depend on human annotators for rating outputs and filtering harmful content, often under poor working conditions.
  • Illusion of intelligence: The apparent fluency of LLMs can mislead users into attributing understanding or agency where none exists.

What are the main harms of AI discussed in The AI Con by Emily M. Bender?

  • Labor exploitation and job precarity: AI systems often replace stable jobs with precarious gig work or crowd labor, disproportionately affecting marginalized workers.
  • Environmental impact: Training and running large AI models consume vast amounts of energy and water, contributing to climate change and environmental degradation.
  • Surveillance and social control: AI enables pervasive surveillance, especially targeting marginalized groups, and reinforces systemic inequalities through automated policing and risk assessment.
  • Misinformation and trust erosion: AI-generated content floods information ecosystems, making it harder to discern authentic knowledge and undermining public trust.

How does The AI Con by Emily M. Bender address the environmental impact of AI?

  • Hidden resource consumption: The book reveals that AI’s energy and water use is often obscured by terms like "cloud computing," but involves intensive mining, manufacturing, and cooling.
  • Growing carbon footprint: Increasing model sizes and user bases multiply AI’s environmental impact, with data centers’ energy use projected to rival that of entire countries.
  • Corporate pledges fall short: Companies’ climate promises often fail to offset the increased emissions from AI investments.
  • Disproportionate burden: Environmental harms from AI disproportionately affect marginalized communities and climate refugees, while tech elites remain insulated.

What does The AI Con by Emily M. Bender say about AI’s impact on labor and workers?

  • Automation replaces stable jobs: AI is used to replace well-paying, stable jobs with precarious gig or crowdwork, often under exploitative conditions.
  • Invisible human labor: Much of AI’s apparent competence depends on low-paid workers performing data labeling, content moderation, and filtering.
  • Worker resistance: Unions and labor movements, such as the Writers Guild of America, are actively resisting AI-driven job displacement and setting standards for worker protections.
  • Need for stronger laws: The book calls for enforcing and expanding labor protections, including laws like the PRO Act, to safeguard workers in the AI era.

How does The AI Con by Emily M. Bender critique the concept of artificial general intelligence (AGI)?

  • Lack of clear definition: The book points out that neither AI Doomers nor Boosters have a justifiable, operational definition of "general intelligence," making AGI claims speculative.
  • Historical context: It references Alan Turing’s work to show that foundational AI questions are too vague without precise definitions.
  • Quasi-spiritual beliefs: The term AGI has taken on a quasi-spiritual significance among some researchers, lacking empirical grounding.
  • Distraction from real harms: Focusing on AGI diverts attention from the tangible harms caused by current AI technologies.

What strategies and advice does The AI Con by Emily M. Bender recommend for resisting AI hype and harms?

  • Ask pointed questions: The book encourages asking what is being automated, how systems are evaluated, who benefits or is harmed, and what labor and data practices are involved.
  • Everyday resistance: Individuals can resist by refusing to use AI tools, mocking synthetic content, and supporting labor unions fighting AI-driven job loss.
  • Support information literacy: Backing libraries and promoting information literacy helps counter the spread of AI-generated misinformation.
  • Advocate for regulation: The authors call for enforcing existing laws, demanding transparency, and prioritizing human and civil rights over unchecked innovation.

What are the best quotes from The AI Con by Emily M. Bender and what do they mean?

  • "Artificial intelligence, if we’re being frank, is a con: a bill of goods you are being sold to line someone’s pockets." This quote encapsulates the book’s central thesis that AI hype primarily serves commercial interests rather than the public good.
  • "Always read the footnotes." The authors urge readers to look beyond surface claims and examine the evidence and assumptions underlying AI technologies.
  • "AI is a marketing term. It doesn’t refer to a coherent set of technologies." This challenges the common perception of AI as a unified field, emphasizing its use as a buzzword.
  • "Science is a human, social endeavor." The book reminds readers that technological progress and scientific discovery are collective, human-driven processes, not the result of autonomous machines.

About the Author

Emily M. Bender is a professor of linguistics at the University of Washington, specializing in computational linguistics and natural language processing. She is known for her critical stance on the limitations and ethical implications of AI language models. Bender has published numerous academic papers on the subject and is a prominent voice in discussions about AI ethics and responsible development. Her co-author, Alex Hanna, is a former research scientist at Google's Ethical AI team. Together, they bring expertise from both academic and industry perspectives to examine the societal impacts of AI technologies and challenge prevailing narratives about artificial intelligence.

Download PDF

To save this The AI Con summary for later, download the free PDF. You can print it out, or read offline at your convenience.
Download PDF
File size: 0.46 MB     Pages: 20

Download EPUB

To read this The AI Con summary on your e-reader device or app, download the free EPUB. The .epub digital book format is ideal for reading ebooks on phones, tablets, and e-readers.
Download EPUB
File size: 1.48 MB     Pages: 27
Want to read the full book?
Follow
Listen
Home
Swipe
Library
Get App
Try Full Access for 3 Days
Listen, bookmark, and more
Compare Features Free Pro
📖 Read Summaries
Read unlimited summaries. Free users get 3 per month
🎧 Listen to Summaries
Listen to unlimited summaries in 40 languages
—
❤️ Unlimited Bookmarks
Free users are limited to 4
—
📜 Unlimited History
Free users are limited to 4
—
📥 Unlimited Downloads
Free users are limited to 1
—
Risk-Free Timeline
Today: Get Instant Access
Listen to full summaries of 26,000+ books. That's 12,000+ hours of audio!
Day 2: Trial Reminder
We'll send you a notification that your trial is ending soon.
Day 3: Your subscription begins
You'll be charged on Nov 25,
cancel anytime before.
Consume 2.8× More Books
2.8× more books Listening Reading
Our users love us
1M+ readers
Trustpilot Rating
TrustPilot
4.6 Excellent
This site is a total game-changer. I've been flying through book summaries like never before. Highly, highly recommend.
— Dave G
Worth my money and time, and really well made. I've never seen this quality of summaries on other websites. Very helpful!
— Em
Highly recommended!! Fantastic service. Perfect for those that want a little more than a teaser but not all the intricate details of a full audio book.
— Greg M
Save 62%
Yearly
$119.88 $44.99/year/yr
$3.75/mo
Monthly
$9.99/mo
Start a 3-Day Free Trial
3 days free, then $44.99/year. Cancel anytime.
Unlock a world of fiction & nonfiction books
26,000+ books for the price of 2 books
Read any book in 10 minutes
Discover new books like Tinder
Request any book if it's not summarized
Read more books than anyone you know
#1 app for book lovers
Lifelike & immersive summaries
30-day money-back guarantee
Download summaries in EPUBs or PDFs
Cancel anytime in a few clicks

Why one summary?

Scanner
Find a barcode to scan

⭐
⭐
⭐
We have a special gift for you
Open
38% off
DISCOUNT FOR YOU
⭐
⭐
⭐
$79.99
$49.99/year
Only $4.16 per month
Continue
2 taps to start, super easy to cancel
Settings
General
Widget
Loading...