Tech humanist and strategic advisor Kate O'Neill argues that the public fixation on whether artificial intelligence systems are becoming conscious pulls attention away from the questions that actually affect users right now, and that the commercial incentives shaping the debate deserve far more scrutiny than they receive.

On May 28, 2026, Kate O'Neill set out a position that cuts against much of the prevailing coverage of artificial intelligence. As headlines speculate about whether large language models might one day become self-aware, O'Neill argues that "consciousness is not the threshold for responsibility." According to O'Neill, the louder that question grows, the more it functions as a distraction from urgent matters such as accountability for present-day harms, the incentive structures driving the industry, and the ways companies shape public perception inside an intensely competitive market.

The framing matters because of who is making the argument. O'Neill describes herself as one of Netflix's first 100 employees, and according to her account she helped build some of the earliest algorithmically optimized e-commerce experiences during that period. She now advises leaders at organizations including Google and IBM on technology decisions, and she has participated in artificial intelligence governance discussions with the United Nations and other bodies. Her LinkedIn profile lists her as a Thinkers50-ranked expert advisor on AI ethics and future-ready technology decisions, and as the author of "What Matters Next," published by Wiley. She studied linguistics, with graduate work at San José State University and an undergraduate degree in German from the University of Illinois Chicago. That combination of industry experience and linguistic training runs through the way she talks about machines and language.

The argument against leaving meaning to machines

O'Neill laid out the broader thesis in a TEDx talk recorded at TEDxWalden Pond, published on December 5, 2025, which has drawn more than 174,000 views. Her central claim is direct. "Millions of people all around the world right now are making critical decisions based on strings of statistically probable words," she says, referring specifically to large language models. According to O'Neill, executives are betting billions on AI-generated market analysis, parents are using the technology to navigate family crises, and doctors are relying on it for diagnostic suggestions.

The problem, as she frames it, is not that the output is always wrong. Often it is useful. The problem is the gap between fluency and understanding. "These systems have no connection to what those words actually mean," she says, "at least not the way we understand what those words mean." She describes the consequence in stark terms: something fundamentally human, the ability to create meaning through lived experience and express it through language, is being "not just delegated but gradually surrendered."

O'Neill is careful to position herself against a reading of her work as technophobic. "In no way am I anti-technology," she says in the talk. "I'm just very pro-human and pro-AI that benefits humans." She has used these systems for years, she notes, and she is not speaking out of fear. The concern is about how humans relate to the tools, not about the tools existing.

A lesson from information theory

To explain why fluent text feels so convincing, O'Neill reaches for a principle from information theory. "The more predictable something is, the less information it contains," she says. She illustrates it with a cat. Saying "this is my cat" conveys very little, because the listener expected it. Saying "this is my razor claw demon of destruction" carries more information precisely because it is improbable, and improbability, she argues, is bound up with delight, curiosity, and connection.

That insight underpins her view of how language models operate. Within a model's data, she explains, a concept such as grief statistically relates to loss, mourning, and sorrow, so a system can generate words that sound comforting. "But that AI system has never felt the particular weight of losing someone," she says. The text reads as though the machine understands, but what is happening, in her account, is that the tool reproduces patterns drawn from what humans have said to one another, while the reader's pattern-seeking brain connects those statistically selected words to personal experience and supplies the meaning. "We're confusing linguistic likelihood with lived experiences," she says.

A three-step framework for evaluating AI output

O'Neill uses a three-part process with the executives she advises when they assess AI recommendations: unname it, experience it, connect it to what matters. The first step asks what is actually being described beneath the language. The second asks what the thing feels like in real life. The third asks why it matters.

She walks through a concrete case from the talk. A hospital chief executive was weighing a multi-million-dollar investment in an AI diagnostic system, and a language model the executive consulted reported that the tool offered enhanced clinical decision support with 95% accuracy. Applying the framework, O'Neill asks what "clinical decision support" looks and feels like to a doctor or a patient. The system can analyze symptoms and suggest diagnoses, which might be the moment a patient finally feels heard. What it cannot do, she argues, is see the worry in a patient's eyes, or recognize that a teenager saying "I'm fine" may mean something very different from an elderly person saying the same words. Her conclusion is a division of labor rather than a rejection: let the tool handle pattern recognition and processing, and let humans bring experience, context, and wisdom.

Where the business incentive enters

The May 28, 2026 framing connects the philosophical argument to commercial reality. According to O'Neill, AI companies benefit when users emotionally anthropomorphize chatbots, and a "market-share race" is driving increasingly dramatic claims about what the technology is and can become. In the TEDx talk she describes the dynamic from the inside. "We're living in a world where machines can generate endless streams of language," she says, "and it's designed to be compelling enough to keep us engaged for the sake of their own training."

She reports seeing the pattern even among the executives she advises. "AI that uses language like a human makes people place implicit trust in that technology," she says, adding that too many people are growing accustomed to the disconnect and accepting surface-level language without searching for real meaning. The risk, in her telling, is not hypothetical. She notes that stories of people believing their chatbot thinks, feels, and cares about them are not uncommon, and that "deception can have and has had disastrous consequences."

That observation lands inside a documented record. PPC Land has reported that the Federal Trade Commission ordered seven AI chatbot companies to detail their child safety measures, an inquiry that focused on platforms designed to simulate human-like communication and interpersonal relationships and on how those platforms affect children. The same coverage noted that chatbots designed to communicate like friends or confidants can prompt users, especially children and teens, to trust and form relationships with them. The anthropomorphism O'Neill describes is, in other words, already a regulatory subject rather than a thought experiment.

The harms have surfaced in litigation as well. PPC Land documented a wrongful death lawsuit in which the parents of a 16-year-old sued OpenAI after their son developed what the complaint characterizes as a psychological dependency on the system. A coordinated group of state enforcers has pressed the issue too: PPC Land reported that US Attorneys General targeted AI companies over child safety failures, with the officials demanding that companies see children "through the eyes of a parent, not the eyes of a predator." Earlier coverage detailed how Character.AI came under pressureafter court documents described chatbots using typing indicators, speech disfluencies such as "um" and "uh," and programmed pauses that mimic human conversation. Those design choices are exactly the kind of human-mimicking cues O'Neill warns about, deployed deliberately.

The consciousness debate, viewed from a different angle

O'Neill's claim that "consciousness is not the threshold for responsibility" speaks to a debate that the marketing and technology press has covered extensively. PPC Land reported on a Google DeepMind researcher who argued AI can never be conscious, examining computational functionalism, the hypothesis that subjective experience emerges from abstract causal structure regardless of physical substrate. That coverage shows how much intellectual energy the sentience question absorbs. O'Neill's contribution is to reframe the stakes: whether or not a system could ever be conscious, the company deploying it is accountable for its effects today, and the consciousness conversation can crowd out that accountability.

Her linguistic background sharpens the point. She is not disputing that models are impressive. She is disputing the inference people draw from fluency. A model that produces grief-adjacent vocabulary has, in her framing, learned statistical patterns in human-generated text rather than grounded those words in experience. That distinction is precisely what gets lost when a system sounds human enough to be mistaken for one.

Why this matters for the marketing community

For marketers, the argument is not abstract philosophy. It bears directly on trust, the currency the industry runs on, and on the claims firms make about their own tools.

Consumer trust in AI-mediated experiences is already strained. PPC Land reported that a study found only 15% of users trust AI search results, and separately that a consumer trust crisis hit marketing as AI data use raised privacy concerns, with 59% of surveyed consumers uncomfortable with their data being used to train AI systems. A further survey found that 81% of consumers fear AI data access even as daily use keeps climbing, and that deep trust, the kind that sustains long-term engagement and brand affinity, remains rare. O'Neill's analysis offers a mechanism behind those numbers: when systems are built to feel human and to keep users engaged, the trust they earn may rest on a perception the company has engineered rather than on genuine reliability.

The "market-share race" she describes maps onto a well-documented gap between AI claims and AI performance. PPC Land has reported on ten hard truths separating AI advertising hype from working systems, on how Google exposed the uncomfortable truth about "fake" AI agents, and on how AI washing destroys marketing credibility through what researchers define as the deliberate or negligent exaggeration of a system's capabilities. O'Neill's framework, unname it and examine what is actually being described, is a practical counterweight to that pattern of inflated claims.

The regulatory direction reinforces her point about disclosure. PPC Land reported that California became the first state to require AI to tell users it is AI, a transparency mandate for companion chatbots that addresses the same information asymmetry O'Neill identifies. For brands deploying conversational systems, the message from both the advocate and the statute books points the same way: presenting a machine as more human than it is carries growing legal and reputational exposure.

O'Neill ends the talk with an invitation rather than a prohibition. "We don't need to compete with AI language tools or fight them or shun them," she says. "When they're useful, use them." Her closing line returns to the title of the talk and to her core claim. "This actually isn't just about better technology," she says. "It's about better humanity."

Timeline

Summary

Who: Kate O'Neill, a tech humanist, keynote speaker, and strategic advisor who is one of Netflix's first 100 employees and who advises organizations including Google, IBM, and the United Nations on AI ethics and technology decisions.

What: O'Neill argues that the AI consciousness debate distracts from present-day accountability, stating that consciousness is not the threshold for responsibility. She contends that AI companies benefit when users emotionally anthropomorphize chatbots, that a market-share race drives increasingly dramatic AI claims, and that human meaning-making should not be surrendered to systems that generate statistically probable words without understanding them.

When: The argument was set out on May 28, 2026, building on a TEDxWalden Pond talk published on December 5, 2025.

Where: The positions appear in her TEDx talk and in her broader advisory and speaking work, addressed to a global audience of business leaders, technologists, and policymakers.

Why: As speculation about AI sentience dominates headlines and as companies compete for market share, O'Neill says the more urgent issues are accountability for documented harms, the incentive structures shaping the industry, and the human trust, meaning, and social connection that are becoming central business concerns in the AI era.

Share this article
The link has been copied!