# Why people like AI — and why they object to it

## Research report

**Prepared:** 9 August 2026  
**Purpose:** Expand the analysis of the Facebook discussion supplied by the user, identify the full range of reasons people give for adopting or rejecting AI, and test the proposition that direct personal usefulness is the decisive route to acceptance.

## Executive conclusion

The central proposition is broadly supported, but it needs one qualification: **a successful first-hand experience is a powerful conversion mechanism, not a universal cure for distrust**.

People are increasingly willing to use AI because it gives them immediate capability, saves time, lowers the cost of expertise, helps them learn, improves communication, and makes difficult tasks easier to start. The most persuasive benefit is not that AI is technologically impressive. It is that a person can give it a problem they actually have and receive a useful result within minutes.

At the same time, use does not automatically produce confidence. The clearest current pattern is **practical acceptance combined with institutional distrust**. Global research finds high use and widespread belief in benefits alongside low willingness to trust systems, strong support for regulation, and concern about mistakes, privacy, manipulation, jobs and human relationships.

The public debate is therefore not accurately described as “pro-AI versus anti-AI.” It is better described as four overlapping positions:

1. **Utility adopters:** “It helps me do things I could not otherwise do.”
2. **Conditional users:** “It is useful, but only with checking, privacy and human control.”
3. **Social critics:** “The technology may work, but the economic and political system around it is dangerous.”
4. **Rejectionists:** “The costs to human capability, employment, environment or civilisation are unacceptable.”

The Facebook thread appears more pro-utility than the wider public because people who have strong practical experience are more likely to comment, and because the thread is already about a technology-literate topic. It is useful evidence of argument types, not a representative poll.

## What the strongest current evidence says

### Global adoption is high, but trust lags behind use

The 2025 University of Melbourne–KPMG global study surveyed more than 48,000 people in 47 countries. It reported that 66% use AI regularly and 83% believe AI will produce a wide range of benefits, but only 46% are willing to trust AI systems. Sixty-six percent also reported relying on AI output without evaluating its accuracy, and 56% said they had made mistakes at work because of AI. This is a crucial finding: **people can find AI useful before they find it trustworthy**. [KPMG / University of Melbourne](https://kpmg.com/xx/en/our-insights/ai-and-technology/trust-attitudes-and-use-of-ai.html)

The same study found that 70% believe AI regulation is needed. The public is not asking to choose between use and safety; it is asking for both.

The 2026 Stanford AI Index reports a similar dual movement: globally, the share saying AI products and services offer more benefits than drawbacks rose from 55% in 2024 to 59% in 2025, while the share saying AI makes them nervous rose to 52%. Optimism and anxiety are rising together. [Stanford AI Index 2026, public opinion](https://hai.stanford.edu/ai-index/2026-ai-index-report/public-opinion)

### Direct experience appears to matter

Pew’s 2025 international survey found that people who had heard a lot about AI were more likely in many countries to feel excited rather than concerned. That is consistent with an experience effect, although it does not prove that using AI causes optimism: people who are already curious or confident may be the ones who seek it out. [Pew Research Center, global attitudes](https://www.pewresearch.org/global/2025/10/15/concern-and-excitement-about-ai/)

Google and Ipsos’ 2026 “Our Life with AI” survey of 21,000 people in 21 countries found that 74% of users use AI to learn something new or understand a complex topic. The survey’s stated shift is from curiosity to utility. Students, teachers and parents reported especially high use, with learning, schoolwork, saving time and day-to-day assistance among the leading uses. This is a company-sponsored survey, so it should be treated as useful but not independent of commercial interests. [Google–Ipsos, Our Life with AI](https://blog.google/products-and-platforms/products/education/our-life-with-ai-2025/)

The safest version of the claim is therefore:

> People often become more favourable to AI after it solves a personally meaningful problem, but personal usefulness does not erase concerns about who controls the system, what happens to their data, or what wider effects follow from mass adoption.

## Why people like AI: the comprehensive list

### 1. It gives ordinary people new capability

This is the deepest attraction. AI lets a person make a spreadsheet, write code, analyse a document, produce a plan, draft a formal letter, interpret technical language or create a visual concept without first becoming a specialist.

The benefit is experienced as agency:

- “I can do this myself.”
- “I no longer need to wait for an expert.”
- “I can at least get to a competent first draft.”
- “I can participate in a conversation that used to be closed to me.”

For a small business, the gain may be access to capabilities it could never afford: marketing, bookkeeping analysis, customer correspondence, website work, research, proposal writing and basic automation. AI is not necessarily replacing an existing employee in this case; it is providing a capability that did not exist at all.

### 2. It saves time and removes drudgery

People value the compression of hours into minutes. Common examples are summarising long material, drafting routine correspondence, extracting information from documents, preparing meeting notes, formatting text, generating lists, and handling repetitive administrative work.

The public argument is often less “AI will make me rich” than “it gives me back time.” That time may be used for higher-value work, family, rest, creativity or simply getting through an overloaded day.

There is a critical distinction here. AI may save time for the worker while also causing an organisation to demand more work. The same capability can feel liberating at personal scale and exploitative at managerial scale.

### 3. It removes the blank-page problem

AI makes it easier to begin. It can provide a structure, checklist, outline, first draft, list of options or proposed sequence of actions.

This matters to people who procrastinate, have attention difficulties, feel anxious about writing, are unfamiliar with formal processes, or know what they want but cannot express it clearly. Often users do not want the machine to finish the job; they want enough momentum to get started.

### 4. It is an always-available tutor

AI allows people to ask basic questions repeatedly, request a simpler explanation, change the level of difficulty, ask for examples, and receive immediate clarification without embarrassment. This is different from a search engine: the system can reframe information interactively.

Important learning uses include mathematics, coding, languages, science, history, practical repairs, professional development, regulations, financial terminology and medical vocabulary.

The appeal is especially strong for adults who missed formal education, people learning in a second language, isolated learners, and people who cannot afford private tutoring.

### 5. It personalises information

AI can explain the same topic as:

- a beginner’s guide;
- a technical briefing;
- a worked example;
- an analogy;
- a checklist;
- a comparison of options; or
- a step-by-step procedure.

That adaptability is one reason it often feels more useful than conventional search. It turns information into an interaction shaped around the user’s actual circumstances.

### 6. It improves communication

AI helps people clarify, shorten, translate, re-tone and organise what they are trying to say. It is valuable for second-language writers, people with dyslexia, people with low writing confidence, people dealing with government or legal institutions, and anyone who has a difficult message to compose.

The user’s experience is often not “AI wrote this for me.” It is “AI helped me express what I meant.”

### 7. It makes expertise more accessible

People use AI to prepare before dealing with a doctor, lawyer, accountant, builder, insurer, employer or government agency. They ask it to explain a quotation, identify questions, compare terms, translate jargon or show where uncertainty remains.

Used responsibly, this does not replace the professional. It helps the customer become less passive and less intimidated.

### 8. It helps people think

AI can function as a thinking surface. Users ask it to test an argument, identify assumptions, propose counterarguments, compare strategies, expose weaknesses, map consequences and organise scattered ideas.

This is the strongest response to the claim that all AI use is “letting the machine think.” Passive acceptance of an answer can weaken thinking; active interrogation can extend it. The difference is the user’s behaviour.

### 9. It accelerates research and synthesis

People increasingly need to combine notes, reports, transcripts, spreadsheets, images, emails and web material. AI can help identify themes, contradictions, gaps, relationships and priorities across a volume of material that is difficult to hold in working memory.

This is particularly valuable for research, policy work, journalism, planning, archival projects and personal records.

### 10. It lowers the cost of experimentation

AI makes it cheap to generate and compare alternatives: business names, advertising copy, layouts, software approaches, lesson plans, story ideas, visual styles and project structures.

The key benefit is not that every output is good. It is that failed attempts become inexpensive, and the user can iterate quickly.

### 11. It lowers the cost of professional-quality first drafts

AI can move someone from “nothing” to “something reviewable.” That reduces the price of reaching a professional review stage in law, design, coding, translation, analysis, marketing and administration.

This is economically important for individuals, community groups and small businesses. It can also threaten professional livelihoods when organisations use first drafts as a substitute for paying specialists.

### 12. It increases independence

AI is available outside office hours, without an appointment and without needing to know exactly which professional to contact. That matters for urgent problems, isolated people, people with limited money and people who are reluctant to ask for help.

### 13. It is patient and non-judgmental

People may ask AI questions they would not ask another person: basic questions, embarrassing questions, relationship questions, career questions or emotionally difficult questions. The lack of visible social judgement is a real part of the product’s appeal.

It is also a risk: a system that feels understanding can create misplaced trust or emotional dependence.

### 14. It improves accessibility

Useful applications include speech-to-text, text-to-speech, translation, image description, simplified instructions, summarisation, form assistance and interface navigation. These can increase independence for people with visual, hearing, reading, cognitive, language or mobility barriers.

### 15. It supports health and science

People are attracted to AI’s potential in diagnosis, medical imaging, drug discovery, personalised treatment, clinical administration, public-health surveillance, weather forecasting, materials discovery, protein research, logistics and climate modelling.

The attraction is understandable: some problems involve more variables and more information than unaided human reasoning can comfortably process. The appropriate conclusion is potential, not automatic safety. WHO supports carefully governed health applications but warns that plausible-sounding errors, bias, privacy failures and disinformation can harm patients. [WHO guidance on AI for health](https://www.who.int/news/item/16-05-2023-who-calls-for-safe-and-ethical-ai-for-health)

### 16. It removes undesirable work

Supporters want AI to reduce monotonous administration, dangerous inspections, repetitive production, exposure to hazardous environments and physically damaging tasks. The optimistic argument is about improving the quality of work, not merely increasing output.

### 17. It provides entertainment, companionship and curiosity

Some people begin with jokes, role-play, image creation, historical simulations, games or unusual conversations. Curiosity can be the entry point to practical use.

Companionship is more controversial. It can reduce loneliness for some users, but a commercial system that imitates affection or agreement may manipulate vulnerable people. It is both a reason people like AI and a reason others oppose its expansion.

## What people are actually saying in favour

Across the Facebook discussion and the open-ended survey evidence, pro-AI arguments tend to cluster into a small number of recurring sentences:

- “It is a tool, like the internet, electricity or the printing press.”
- “Learn it or be left behind.”
- “It lets one person do the work of a team.”
- “I use it every day and it has saved me hours.”
- “It gives small businesses a chance.”
- “It helps me understand things I could not afford to have explained.”
- “It does not replace expertise; it makes expertise more productive.”
- “The answer is not to ban it but to regulate it.”
- “The benefits in medicine, science and accessibility are too large to reject.”
- “The problem is bad use and bad governance, not the underlying tool.”

Pew’s open-ended U.S. survey provides an unusually useful check on these anecdotes. Among people who rated AI’s societal benefits as high, 41% most commonly cited efficiency and freeing people’s time; 23% cited expanding human and technological abilities, including faster scientific and medical progress and broader access to information. [Pew, public responses in their own words](https://www.pewresearch.org/science/2025/09/17/americans-on-the-risks-benefits-of-ai-in-their-own-words/)

## Comprehensive list of objections

### Human capability and culture

- **Loss of critical thinking:** people may stop checking, remembering, researching or struggling productively.
- **Deskilling:** people may lose the ability to write, calculate, navigate, code, diagnose or troubleshoot when the system is unavailable.
- **Loss of curiosity and patience:** instant answers may reduce the motivation to investigate.
- **Erosion of creativity:** generated work may become statistically competent but less original.
- **Cultural homogenisation:** repeated model patterns may favour dominant styles and flatten difference.
- **Loss of authorship and dignity:** if machines can perform valued creative and intellectual tasks, human achievement may feel less meaningful.
- **Reduced human contact:** automated teachers, carers, therapists, doctors and customer-service agents may be cheaper but less humane.
- **Companionship and emotional manipulation:** personalised systems may flatter, persuade, reinforce delusions or cultivate dependency.
- **Children’s development:** young people may outsource foundational writing, reasoning and social learning before developing those capacities.

Pew’s 2025 U.S. survey found that 53% believed AI would worsen creative thinking, compared with 16% who thought it would improve it; 50% said it would worsen meaningful relationships, compared with 5% who thought it would improve them. In open-ended responses, the most common stated reason for seeing high risk was erosion of human abilities and connections. [Pew, human abilities and society](https://www.pewresearch.org/science/2025/09/17/how-americans-view-ai-and-its-impact-on-people-and-society/)

### Truth, trust and information

- **Hallucinations:** fluent, confident errors can be more dangerous than obvious ignorance.
- **Fabricated citations and evidence:** false sources, images, recordings and documents can look authentic.
- **Misinformation at scale:** production becomes cheaper and faster than human fact-checking.
- **Fraud and impersonation:** voice cloning, phishing, fake executives, romance scams and identity theft become easier.
- **The “liar’s dividend”:** once deepfakes are common, genuine evidence can be dismissed as fake.
- **AI slop:** feeds become crowded with generic articles, illustrations, comments, marketing and summaries.
- **Loss of provenance:** people may not know who made content, what data it used or whether it has been altered.

In Pew’s open-ended responses, 18% of those assigning high societal risk cited threats to accurate information, and 11% cited criminal or malicious use. [Pew, public responses](https://www.pewresearch.org/science/2025/09/17/americans-on-the-risks-benefits-of-ai-in-their-own-words/)

### Work, wages and inequality

- **Job displacement:** especially in routine administrative, customer-service, media, design and entry-level knowledge work.
- **Wage suppression:** employers may use AI to weaken bargaining power even when jobs remain.
- **Loss of entry-level routes:** if AI performs junior tasks, people may lose the apprenticeship path to expertise.
- **Work intensification:** faster tools can produce higher quotas, shorter deadlines and constant availability.
- **Unequal distribution:** owners of models, chips, data and platforms may capture gains while workers absorb losses.
- **Credential inflation:** if output becomes cheap, employers may demand more qualifications or unpaid tests.
- **Global labour arbitrage:** AI may further concentrate high-value work in firms and regions with capital and compute.
- **Transition failure:** even if new jobs eventually appear, displaced workers may not have the time, money or geography to move into them.

The evidence does not justify either “AI will destroy all jobs” or “new jobs will automatically compensate.” The World Economic Forum’s 2025 employer survey projected 170 million jobs created and 92 million displaced by 2030 across all major transformation trends, a net increase of 78 million, while also reporting that 41% of employers expect to reduce their workforce because of AI. The ILO’s 2025 update emphasises that exposure is task-specific and that transformation is more likely than complete occupational replacement in many cases. [WEF Future of Jobs 2025](https://www.weforum.org/publications/the-future-of-jobs-report-2025/in-full/2-jobs-outlook/) · [ILO Generative AI and jobs: 2025 update](https://www.ilo.org/publications/generative-ai-and-jobs-2025-update)

### Power, privacy and democratic control

- **Data privacy:** prompts, documents, images, health information and personal histories may be stored or reused.
- **Surveillance:** facial recognition, workplace monitoring, behavioural prediction and data matching can become cheaper.
- **Concentration of power:** a small number of companies control models, cloud infrastructure, chips and distribution.
- **Opaque decisions:** people may be assessed by systems they cannot inspect or challenge.
- **Bias and discrimination:** historical data can reproduce unequal treatment in hiring, lending, insurance, policing, welfare and healthcare.
- **Responsibility gaps:** developers, deployers, users and vendors may each blame the others when harm occurs.
- **Loss of consent:** AI is embedded into search, phones, education, workplaces, banking and public services, making refusal difficult.
- **Digital exclusion:** people without devices, broadband, literacy or money may be disadvantaged.
- **National dependence:** governments may rely on foreign firms for strategic infrastructure and decision support.
- **Regulatory lag:** institutions may be too slow, technically weak or captured by industry.

NIST’s generative-AI risk profile groups risks around governance, data, content provenance, pre-deployment testing, incident disclosure, privacy, information security, bias and human-AI configuration. The important implication is that risks are not one single “AI risk”; they arise at different stages of design, deployment and use. [NIST Generative AI Risk Management Profile](https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence)

### Environment and infrastructure

- electricity demand for training and inference;
- water and cooling requirements;
- land, grid and transmission pressure;
- mining and semiconductor production;
- rapid hardware replacement and e-waste;
- local communities bearing infrastructure costs;
- emissions depending on the energy source and location; and
- opportunity cost when scarce electricity or water could serve other needs.

The concern is real but needs accurate scale. The IEA estimated that data centres consumed about 415 TWh, or 1.5% of global electricity, in 2024 and projected roughly 945 TWh by 2030 in its base case—just under 3% of global electricity demand. That is material and fast-growing, but “AI consumes the world’s resources” is not a defensible description of the current global share. [IEA, Energy and AI](https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai)

### Safety and existential risk

- autonomous weapons and lethal target selection;
- cyberattacks and automated vulnerability discovery;
- systems acting beyond their operators’ understanding;
- manipulation of institutions, markets or public opinion;
- loss of control over highly capable systems;
- replication or persistence beyond intended boundaries; and
- catastrophic misuse by states, corporations or individuals.

These arguments differ in confidence. Fraud, misinformation, privacy leakage and biased decisions are already observable risks. Large-scale labour disruption and social dependency are plausible medium-term risks. loss of control over advanced systems is a more uncertain, longer-horizon risk. They should not be collapsed into one probability figure without specifying the system, timeframe, mechanism and meaning of “catastrophe.”

## What the Facebook discussion adds

The Facebook comments appear to show three important features.

First, **few participants deny that AI works**. The debate has moved beyond “this is just hype.” Even sceptics generally accept that the systems are capable.

Second, the pro-AI side is grounded in direct use: spreadsheets, coding, medicine, research, planning, marketing and business tasks. This is a different kind of argument from a prediction about the future. It is a testimony about a completed task.

Third, much of the anti-AI position is actually anti-concentration, anti-unaccountable deployment or anti-forced adoption. People worry about billionaires, corporations, surveillance, job losses, resource use and the disappearance of human judgement. Some are not asking for a ban; they are asking who gets to decide and who pays the price.

That means the earlier estimate of roughly 55–60% pragmatic/pro-utility, 20–30% conditional, and 15–20% strongly anti-AI may be a reasonable **impressionistic coding of the supplied thread**, but it must not be reported as a population estimate. The sample is self-selected, the post is framed around Elon Musk and AI risk, and commenters are not a random sample of the public.

## The key strategic implication: demonstrate personal usefulness

The most effective introduction is not a general presentation about artificial intelligence. It is a concrete problem supplied by the person themselves:

1. Ask what is frustrating, time-consuming or intimidating right now.
2. Use their own document, message, spreadsheet, photograph or question.
3. Produce a useful first result quickly.
4. Let them correct it and see that the system responds.
5. Show one relevant failure mode and how to check for it.
6. Leave them with a repeatable method rather than a spectacle.

The best first demonstrations are usually:

- explaining a document they already need to understand;
- drafting a difficult message;
- organising accumulated notes or photographs;
- solving a spreadsheet or planning problem;
- comparing a purchase or quotation;
- preparing questions for a professional;
- creating a practical personal plan; or
- turning an idea into a visible first draft.

The compelling proposition is not “AI is impressive.” It is:

> “AI gives you more control over your own work, decisions and ideas.”

The limitation is equally important: a good demonstration can produce adoption, but it should not be presented as proof that every use is safe, fair or socially desirable.

## Final assessment

The strongest pro-AI case is personal and immediate: capability, time, learning, independence and access. The strongest anti-AI case is systemic: loss of human capacity, unreliable information, economic displacement, concentrated power, environmental cost and weak accountability.

The two sides are not symmetrical. Pro-AI arguments usually describe a benefit someone can experience today. Anti-AI arguments often describe consequences that appear when millions of people and powerful institutions deploy the technology at scale. Both can be true.

The likely durable public position is therefore neither unconditional enthusiasm nor absolute rejection. It is **conditional adoption**: people will use AI when it solves a real problem, while continuing to demand human alternatives, privacy, transparency, verification, regulation and a fair distribution of the gains.

## Sources and methodological notes

- The Facebook thread is a qualitative, self-selected sample and should not be treated as representative public opinion.
- Survey results differ because they ask different questions: use, excitement, trust, expected benefits, experienced benefits and societal risk are not interchangeable measures.
- Company-sponsored surveys are useful for identifying reported use cases but should be read alongside independent polling.
- Projections about jobs and advanced AI are scenarios or expectations, not settled forecasts.
- The report uses “AI” broadly. Generative chatbots, recommendation systems, medical models, autonomous systems and industrial machine learning have different capabilities and risks.