Psychedelic AI: Why You Might Want AI to HallucinatePsychedelic AI: Why You Might Want AI to Hallucinate

Hallucinations are often seen as a flaw but they can be a valuable tool for marketers, adding creativity and innovation

Leslie Walsh, Head of strategy at RYA

January 30, 2025

5 Min Read
Silhouette of a face made up of points of light drifting into space
Getty Images

Artificial Intelligence (AI) has made significant strides in recent years, especially in the realm of natural language processing (NLP) and generative models like large language models (LLMs). While these models have proven their utility in tasks ranging from customer service to content generation, one phenomenon that often causes concern is known as “AI hallucinations.” 

In AI, a hallucination refers to a situation where the model generates information that appears coherent and convincing but is, in fact, incorrect or entirely fabricated. While these hallucinations have traditionally been seen as a flaw—undermining trust in the AI's outputs—they can, surprisingly, serve as a valuable tool for marketers, especially when viewed through the lens of creativity and innovation.

Are AI Hallucinations Good or Bad?

At first glance, AI hallucinations might seem like an issue that needs to be avoided at all costs. If a model starts fabricating false data, it could mislead users or create unreliable results. However, when viewed from a different perspective, these same hallucinations can serve as a wellspring of original ideas, especially when marketers are seeking fresh, out-of-the-box concepts. The key to unlocking the creative potential of AI hallucinations lies in understanding the relationship between the data the AI is trained on, its ability to generate novel combinations, and the nature of creativity itself.

Related:A Cure for AI FOMO

The quality and utility of AI hallucinations largely depend on the underlying data that the model has been exposed to. Many LLMs are trained on vast datasets, comprising books, articles, websites, and other forms of human knowledge. These datasets encode a broad range of human creativity, cultural references, and industry trends. When an AI generates a hallucination, it often does so by blending different patterns or pulling together disparate ideas from the training data in ways that the human mind may not typically consider, or dismiss altogether. In this process, the model may create an unexpected or "wrong" combination of elements that, when examined from a creative standpoint, can spark new and innovative marketing concepts.

The Role of Data in Shaping AI Hallucinations

The training data of an LLM is essential in determining the nature of its hallucinations. High-quality, diverse, and well-curated datasets are essential for ensuring that the hallucinations are not entirely nonsensical or disconnected from reality. If the data is extremely precise and limited in ways that make divergence largely impossible, then the hallucinations can’t really miss. If the data is an all-encompassing form of human knowledge, the hallucinations can make irrelevant connections. By focusing LLMs on very precise, relevant source material, AI is funneled into a situation where even hallucinations are relevant and yield valuable concepts.

Related:What to Expect From Generative AI in 2025

For instance, an LLM trained on data from diverse sectors such as fashion, technology, sports, and entertainment might generate a marketing slogan or campaign idea that brings together concepts from these fields in a way that has never been seen before. It might generate a slogan that references sports culture and tech innovation in a way that feels both novel and exciting. In this case, the “hallucination is not a flaw; rather, it is a fusion of existing knowledge that produces something new and unexpected.

AI Hallucinations as a Source of Creative Inspiration

Creativity often emerges when different ideas, concepts, or frameworks collide in ways that are not immediately obvious. This is where AI hallucinations can shine. By generating combinations that don’t always make perfect sense on the surface, AI can lead marketers down unconventional paths, encouraging them to explore new ideas that might never have come to them otherwise - all while providing the data needed to substantiate bold, undiscovered concepts.

Hallucinations can also help marketers think outside the constraints of current trends and mental models. Human creativity is often bounded by what has already been tried and tested—market trends, customer preferences, and competitor actions all play a part in shaping ideas. By contrast, AI models have the advantage of being able to look at the same data as a humans and unearth concepts that may have been missed due to the human constraints above. This lack of bias allows the AI to present a more diverse range of ideas and solutions, many of which are disconnected from the current market and engrained human constraints.

For instance, a marketer working on a campaign for a new tech product might find inspiration in an AI-generated hallucination that combines elements of interests that overlap across various audiences. Who knew that people obsessed with getting the newest tech are also 26% more likely to be into sneaker culture than the general population? Although the idea may initially seem absurd or out of place, upon further reflection, it could result in a groundbreaking marketing strategy that connects the core interests of your target audience with your product in a compelling way. 

Refining Hallucinations: Human-AI Collaboration

While AI hallucinations can open up exciting avenues for creative thinking, they are not perfect and often require human intervention to refine, contextualize, and make real in the world. This is where the collaboration between AI and human marketers becomes crucial. Marketers can take the AI-generated hallucinations and shape them into coherent, relevant ideas by applying their domain expertise, industry knowledge, and cultural awareness.

This collaboration can be likened to brainstorming sessions where no idea is too outlandish. In these sessions, participants are encouraged to think freely, with the understanding that some ideas will be impractical or far-fetched. AI can serve as a "thought partner" in this creative process, offering a continuous stream of data-backed ideas that push the boundaries of what is familiar or comfortable. The marketer’s role is then to assess, adapt, and fine-tune these ideas into actionable concepts.

AI hallucinations don’t need to be perceived as a flaw when they’re leveraged correctly. In fact, they can be a powerful tool for marketers looking to generate fresh and creative ideas. The quality of these hallucinations is directly tied to the richness and precision of the training data, and when curated appropriately, AI can produce data-backed novel combinations and insights that human marketers may not have thought of independently. By treating AI hallucinations as a source of inspiration rather than a problem to be fixed, marketers can leverage these "mistakes" to drive innovation, break free from conventional thinking, and create marketing strategies that captivate audiences in new and unexpected ways.

About the Author

Leslie Walsh

Head of strategy at RYA, RYA

Leslie Walsh is head of strategy at RYA.

Sign Up for the Newsletter
The most up-to-date AI news and insights delivered right to your inbox!

You May Also Like