We taught 200 Caribbean children to spot AI slop. Brands should pay attention.

TLDR

Fluency is a credibility signal, and a generation is being trained to stop trusting it. The Genius Project 2026 put over 200 Caribbean young people aged 5 to 18 through a month of AI, machine learning, statistics and mathematics, awarded US$1 million in cash and prizes, and ran a parallel track teaching parents to recognise AI slop, deepfakes and misinformation. Orbital Brand Science was a sponsor, contributing in-kind support and family outreach. The commercial implication is direct: the surface qualities that automated brand content optimises for, smooth syntax and confident tone, are exactly what an increasingly literate audience is learning to discount.

Why a brand science publication is writing about a children's bootcamp

Orbital spends most of its time on signals: attention, emotion, memory, the measurable traces of how people actually respond to brands. We do not usually cover education programmes. We are covering this one because a specific module inside it changes an input we measure constantly.

Processing fluency is one of the most reliable findings in consumer neuroscience. Content that is easy to read is judged more truthful, more familiar and more trustworthy than content that is not, largely independent of whether it is actually true. Marketing has traded on that effect for a century. Generative AI industrialised it, because fluency is precisely what a language model produces by default, and it produces it at zero marginal cost.

What The Genius Project taught, to children and to their parents, is that fluency is now uninformative. That is a small sentence with a large downstream effect on how brand communication will be received over the next decade.

What the programme actually did

Over 200 Caribbean young people joined The Genius Project in 2026. The youngest was five, the oldest eighteen. Across one month they moved from using AI tools to building machine learning models. US$1 million in cash and prizes was awarded and no family paid tuition, which is only possible because sponsors funded it.

The curriculum is deliberately ordered. Tools first, because a young person needs a visible result quickly. Then the ground shifts: students go from prompting a model to understanding what a model is, which means training data, features, labels, and the difference between a system that learned a pattern and one that memorised an answer. That requires statistics, and statistics requires mathematics. For the older cohorts it meant Python, notebooks, and the very ordinary experience of code that runs correctly and returns the wrong number.

Teams then built for four problem areas: crime and community safety, poverty and access to services, sport, and the ethical questions raised by AI systems themselves. Every team had to state who their system could fail, what data it should never hold, and what they would say to a person the model got wrong. That was a build requirement rather than a lecture module, which is a distinction most corporate AI ethics training still has not made.

The module that matters commercially

Parents ran their own track alongside their children. The practical layer came first: account and privacy settings, what a chatbot retains, what should never be pasted into one, and how to recognise a website built to harvest information.

Then judgement. Telling a generated image from a photograph. Checking a claim before forwarding it. Recognising AI slop, the fluent and confident text that happens to be wrong, and understanding that the fluency is the trap rather than the reassurance.

Once an audience learns that smooth, confident and generic is what a machine produces for free, smooth confident generic stops functioning as a trust signal and starts functioning as a warning.

That inversion is the finding worth taking into a marketing meeting. Detection literacy does not stay in the domain where it is taught. A parent who has spent four weeks learning to spot a generated product review, a fabricated testimonial and a synthetic influencer does not switch that faculty off when they open a brand's email. The scepticism generalises, and it generalises to the cues rather than to the sources.

What this changes for brand work

Volume stops being an advantage. When generation is free, output volume signals only that generation was free. The brands that benefit from the shift will be the ones whose content carries evidence a machine cannot cheaply fake: a named person accountable for the claim, a number with a source, a specific detail that only someone who was present would include.

Specificity becomes the credibility carrier. Generic fluency was doing the trust work. As it depreciates, the load transfers to verifiable specifics. This is measurable in our own work: attention holds longer on concrete claims than on smooth abstractions, and the gap widens for audiences with higher stated scepticism about AI content.

Disclosure is cheaper than being caught. Brands using generative tools in creative production should say so, plainly, and describe the human role in the process. The audience being trained right now will detect it eventually. Voluntary disclosure costs a line of copy; discovered concealment costs the trust that took a decade to build.

Attribution and consent are becoming table stakes for creative work. The programme taught fourteen-year-olds to ask whose data trained a system before trusting its output. They will ask the same question about the music, imagery and voices in Caribbean advertising, and they will ask it publicly.

The completion number, published

Completion across all programme areas currently stands at roughly 15 percent. The programme publishes that figure rather than reporting enrolment alone, which is worth noting in a sector where impact reporting is usually a marketing exercise.

Source: The Genius Project programme data, August 2026, measured across all programme areas. The comparison bar is an indicative benchmark: completion in large open online courses is commonly reported in the mid single digits. These are not matched populations; the benchmark is here to give the 15 percent a sense of scale.

The drop-off has two identified causes: the transition from tools to mathematics, which is where any technical curriculum loses people, and infrastructure, meaning unreliable connections and nowhere quiet to work. Both are being addressed in the 2027 design with shorter mathematics modules, offline-capable materials, and a direct call to any participant who goes quiet for more than three days.

The hackathon

The month closed with a final hackathon. Teams presented to judges, defended their builds, and answered for their design choices. Congratulations to the winners, and to every team that presented at all. Defending a technical build in front of a panel of adults is difficult at thirty, and several of these presenters were not yet thirteen.

One observation from the room, offered as a note rather than a finding: the presentations that landed hardest were the ones that named a limitation early. That is the same pattern we see in message testing with adult audiences. Acknowledged weakness reads as credibility. It appears the instinct is present before anyone teaches it, and most professional communication trains it out.

Why we start early, and why brands should want it

The case for beginning at age five is not that five-year-olds should write code. They should not. A five-year-old sorts objects, counts them, spots a pattern, and learns that the machine's guess can be wrong. That last lesson is the foundation of everything that follows: a child who learns it at five verifies at twelve and does not forward the deepfake at twenty.

A brand should want that audience. Markets where nobody can tell real from synthetic are markets where trust collapses uniformly and no honest brand can price its credibility. An audience that can tell the difference is an audience that will pay for the difference. Detection literacy is not a threat to good brand work. It is what makes good brand work legible.

Who funded the 2026 programme

The Genius Project charges families nothing, which works only because organisations across the region contributed cash and in-kind support. The 2026 sponsors were StarApple AI (prize funding, instructors and curriculum), Maestro AI Labs (technical mentorship and lab time), the Caribbean AI Association (regional backing and CARICOM reach), 14West (hackathon and prize pool), AI Trinidad and Tobago (delivery across the twin islands), Orbital Brand Science (in-kind support and family outreach), and Adrian Dunkley personally.

To every sponsor, judge, volunteer instructor and parent who gave up a month of evenings: thank you. To the students: you did the hard part.

Families can register for the next cohort at beagenius.org. Tuition-free, from age five, virtual for participants outside Jamaica, no prior coding experience required. Organisations that want to sponsor a cohort can reach the programme through the same site.

Frequently asked questions

What is The Genius Project?

A Caribbean non-profit running tuition-free AI education for young people aged 5 to 18, founded by Adrian Dunkley. Students start with AI tools, then move into machine learning, statistics, mathematics, teamwork and problem solving, and build working solutions to real social problems. It runs in person in Jamaica and virtually across the region.

What is AI slop and why does it matter to brands?

AI slop is generated content that reads fluently and confidently while being generic, unverified, or wrong. It matters because processing fluency is treated by the brain as a credibility signal. Audiences learning to detect slop are learning to discount the exact surface qualities that a lot of automated brand content now optimises for.

What did Orbital Brand Science contribute to the 2026 programme?

In-kind support and family outreach, helping the programme reach households across the region. The Genius Project charges families nothing, so it runs entirely on sponsor cash and in-kind contributions.

How much was awarded and how many joined?

US$1 million in cash and prizes across one month, with over 200 participants aged 5 to 18. Completion across all programme areas currently stands at roughly 15 percent. No family paid tuition.

Why train the parents as well as the children?

A child who understands AI better than every adult around them is not in a safe position. Parents covered account and privacy settings, safe use of AI tools and websites, critical thinking, and how to recognise AI slop, deepfakes and misinformation. Reaching adults through their children works because the motivation is already present.

How can families or sponsors get involved?

Registration and sponsorship enquiries are at beagenius.org. The programme is tuition-free, takes students from age five, runs virtually for participants outside Jamaica, and requires no prior coding experience.