Human minds do not pass intelligence to each other consistently. So marketing teams converge on a shared fiction about what is working. They group hallucinate. And they build entire content programs on top of it.
Dennis Wakabayashi · The Global Voice of CX · 12 min read
I have been inside hundreds of marketing organizations. Large ones. Well-funded ones. Teams with genuine talent and real commitment to their craft. And I keep seeing the same thing.
The team meets. They share what they think is working. Someone references a piece that performed well three months ago. Someone else builds on that. A third person adds a layer of interpretation. By the end of the meeting, the group has arrived at a shared understanding of what their audience responds to.
That shared understanding is almost always partially wrong.
Not because the individuals are wrong. Because human minds do not pass intelligence to each other consistently. What one person observed, another interpreted differently. What the data actually showed got filtered through memory, narrative, and the natural human desire to make sense of things. By the time the insight travels from one mind to three minds to a content calendar, it has drifted. Sometimes a little. Sometimes completely.
The team is not lying. They are not lazy. They are doing what human minds do when they try to transfer intelligence to each other. They are group hallucinating.
Dennis Wakabayashi · © 2026This is the group hallucination problem. It is not unique to marketing. It is the reason most brands, governments, content creators, and investors fail to realize sustainability of any kind. When the people making decisions are working from a collectively constructed approximation of reality rather than reality itself, the results do not hold. They cannot. Strategy built on a shared fiction produces outcomes that drift, stall, and eventually collapse, regardless of the talent, the budget, or the conviction behind them. In marketing, it shows up as a content program that does not compound. But the mechanism is the same everywhere. It is not a talent problem.
I want to be precise about what I mean by group hallucination. I am not describing confirmation bias, though that is part of it. I am not describing groupthink, though that is related. I am describing something more fundamental: the structural inability of human minds to transfer observed intelligence to each other without distortion. We compress. We interpret. We fill gaps with pattern. Every transfer loses fidelity. In a marketing team making dozens of decisions a week, those losses accumulate into a shared reality that diverges measurably from the actual one. The same process plays out in a government ministry, an investment committee, and a media organization. The domain changes. The mechanism does not.
Intelligence about what an audience actually responds to lives in the world. It lives in how ideas move through professional networks. It lives in what gets shared, cited, acted on, and ignored. It lives in the algorithms that moderate, portray, and distribute human ideas at global scale.
For that intelligence to inform a content decision, it has to travel from the world into a human mind, then from that mind into a meeting, then from the meeting into a shared team understanding, then from that understanding into a brief, then from the brief into a piece of content. Each step is a transfer. Each transfer loses fidelity.
The team does not know it has drifted. That is what makes group hallucination different from a simple mistake. A mistake is visible. A drift in shared reality is invisible from inside the group. The team is aligned. The briefs are coherent. The content is confidently produced. And it is built on a foundation that has quietly separated from the truth.
The most dangerous content programs are not the chaotic ones. The most dangerous ones are the coherent ones built on a shared hallucination. They produce confidently, consistently, and at scale. In the wrong direction.
This is why strong creative talent does not solve the problem. Talented people group hallucinate just as reliably as anyone else. The instincts are sharper. The execution is better. But if the shared understanding the team is executing against has drifted from reality, better execution just takes you further in the wrong direction faster.
The algorithms that moderate, portray, and distribute human ideas at global scale are not hallucinating. They are pattern recognizing at a scale no human team can match. They know what actually moves. What actually spreads. What actually builds trust over time. ATLAS² integrates with those signals mathematically. It does not ask the team what they think is working. It reads what is actually working from the systems that govern how ideas travel.
Below are seven of the forces ATLAS² measures. Each one is a place where the team’s shared understanding commonly drifts from reality.
Audience signal origin
Teams consistently hallucinate their audience’s language. They use the words their category uses, not the words their audience uses. The difference is measurable. Content that starts from verified audience discourse outperforms content that starts from internal assumptions by a factor that compounds across a program.
The team cannot see this drift from inside the meeting. It feels like audience knowledge. It is actually category fluency. The two feel identical from the inside. They perform very differently in the world.
ATLAS² scans public audience discourse before ideation begins. The origin of every idea is verified against what the audience actually says, not what the team believes they say. Assumption distortion is scored and flagged before a brief is written.
Idea novelty in market context
Every team believes their angles are fresher than they are. This is not arrogance. It is a natural consequence of the transfer problem. The signal about what is already saturating the market lives in the global discourse. The team’s awareness of that saturation is filtered through what they personally read, what their tools surface, and what their category peers are visibly producing. That is a small sample of a very large field.
What feels original inside the room is often well-worn in the market. The team group hallucinates novelty because they are comparing their idea to what they remember seeing, not to what is actually circulating.
ATLAS² assesses idea novelty against the actual state of market discourse before production begins. Saturation is measured, not assumed. Ideas with low birth rate and high decay rate are identified at the concept stage, not after budget has been spent.
Content propagation in professional networks
When a team reads a piece in a meeting and feels that it resonates, they are detecting something real. But what they are detecting is resonance with the people in the room. That is a sample size of four to twelve people who share significant context, vocabulary, and professional perspective. It is not a reliable proxy for how the piece will perform with the audience it is actually trying to reach.
Resonance in the world is a propagation rate. Either a piece moves beyond its initial audience or it does not. The team’s felt sense of resonance and the piece’s actual propagation rate are weakly correlated at best.
ATLAS² estimates propagation coefficient before publication using audience belief state data. The goal is a reproduction number above one, consistently, across the program. Not occasionally on a breakout piece. Consistently, as a production standard.
Implicit promise and trust distance
Every piece of content makes promises. The tone makes a promise. The headline makes a promise. The framing makes a promise. The audience reads those promises instantly and measures everything that follows against them. When the delivery does not match the implication, trust erodes. Quietly. With a 60 to 90 day lag before it shows in the numbers.
Teams hallucinate alignment between their implicit promises and their delivery because they wrote both. They know what they meant. The audience only knows what they received. The distance between those two is where trust bleeds out slowly, invisibly, over quarters.
ATLAS² scores the distance between what a piece implies and what it delivers across functional, emotional, and temporal dimensions. Pieces that score high on implicit promise distance are revised before publication, before the trust damage has time to accumulate.
Information density and reader attention
CMOs and marketing leaders read at speed. They are making a continuous decision about whether the next sentence is worth the attention. When the density drops, when a paragraph restates what was just said, when a transition adds words without adding information, they leave. Not dramatically. They just stop reading.
Teams hallucinate density because they are reading their own content with the full context of what they meant to say. That context fills in the gaps. The reader does not have that context. What feels like a well-developed idea to the writer often feels like padding to the reader.
ATLAS² measures information density section by section before publication. High-redundancy passages are flagged. The standard is to keep density close to the reader’s cognitive channel capacity throughout, not just in the opening. Padding is identified before it reaches the audience.
First-contact trust and temporal relevance
Teams hallucinate hook quality by evaluating it against their own curiosity about the topic. But the reader’s decision to keep going is not about the quality of the hook in isolation. It is about whether the hook meets them where they are, right now, in this professional moment. Temporal relevance is the variable. Strong writing is the baseline. Teams consistently nail the writing and miss the timing.
A perfectly written hook for a problem the audience is not currently experiencing performs like a weak one. Timing is not a creative instinct. It is a data question.
ATLAS² scores hooks against current audience belief state data. Temporal relevance is verified against what the audience is actively experiencing right now, not assumed from the content calendar. A hook that scores below threshold is rebuilt at the concept stage.
Credibility structure across the content program
Teams evaluate credibility piece by piece. They ask whether this piece is credible. That is the wrong question. The right question is whether the program is building trust velocity over time. A single credible piece surrounded by inconsistent ones does not build trust. It creates confusion about who you actually are.
The group hallucination here is the assumption that piece-level quality aggregates into program-level trust. It does not. Trust is built by consistency of signal over time, not by the average quality of individual pieces. The algorithms that distribute content know this. They reward consistent trust signals and penalize inconsistency in ways that compound invisibly over quarters.
ATLAS² tracks trust velocity as a program-level metric across the full content history. The goal is a consistently rising trust rate, not isolated high-credibility pieces. Credibility signals are scored structurally before every piece is cleared, with the program trajectory as the frame of reference, not the individual piece.
I want to say this directly because I think it is the insight most marketing organizations are not ready to hear.
The problem is not that your team is not talented enough to perceive what is working. The problem is that human perception was never a reliable data source for this question to begin with. We are not built to track how ideas move through global networks. We are not built to pattern recognize across millions of simultaneous signals. We are not built to hold the full state of market discourse in our heads and update it continuously as it evolves.
This is not a marketing problem. Governments make policy on it. Investors allocate capital on it. Content creators build audiences on it. Brands spend decades on it. The shared understanding that drives the decision was constructed by humans transferring intelligence to each other, compressing it, interpreting it, filling gaps with pattern. By the time it reaches the decision, it is a plausible approximation of reality. Not reality itself. And strategy built on a plausible approximation of reality produces results that do not sustain.
We compensate by talking to each other more. Better meetings. More alignment. Larger research budgets. All of which produce more refined versions of the same shared fiction. The process designed to produce collective intelligence is the process through which the hallucination enters and deepens.
The algorithms that moderate, portray, and distribute human ideas around the world are not working from a team’s shared understanding. They are working from the actual signal. What moves. What spreads. What builds trust with an audience over time. That signal is mathematical. It is pattern-based. It is consistent. And it is entirely accessible to a system designed to read it.
ATLAS² does not try to fix the human transfer problem.
It bypasses it entirely.
It pattern recognizes and mathematically integrates with the algorithms that moderate, portray, and distribute human ideas at global scale. The ground truth does not come from the team’s collective memory. It comes from the system where the audience actually lives.
For a CMO, that means a content program that compounds because it is built on what is actually true, not on what the team collectively believes to be true.
For any organization making decisions that need to sustain, it means something larger. It means access to a shared reality that human transfer alone has never been able to produce.
That is not a measurement improvement. That is a different foundation for how organizations know what they know.
