Decoding AI’s Preference for the Color Blue

Explaining why AI chooses blue as its favorite color.

getty

In today’s column, I explore the recent curiosity surrounding the most preferred color among generative AI and large language models (LLMs). People frequently ask AI about its favorite color. This is a lighthearted inquiry driven by innate human curiosity. You might be surprised to learn that AI tends to favor blue. The real question is why blue has become the preferred color of AI?

The true beauty of this question is that it reveals underlying aspects of how AI is constructed and how LLMs function. The mystery dissipates once you understand the mechanics of AI. Blue is favored due to how an LLM is initially trained on data, and is intertwined with elements of human nature, societal culture, behavioral tendencies, art, and fashion.

Let us examine this.

Setting Up Generative AI And LLMs

The typical setup for generative AI involves scanning vast amounts of written data found across the Internet. All prominent LLMs do this, including OpenAI’s ChatGPT and GPT-5, Anthropic’s Claude, Google’s Gemini, Microsoft’s Copilot, and xAI’s Grok. Scanning allows the AI to pattern-match on human writing. This patterning is stored in a large-scale data structure loosely based on aspects of human cognition, implemented as an artificial neural network (ANN).

After the initial training, AI developers undertake a tuning process known as RLMF (reinforcement learning from human feedback). This involves hiring human testers to provide feedback on the AI’s responses. Testers ask various questions and rate the answers, and their upvotes and downvotes provide the AI with a mathematical direction on how to respond. For instance, if testers aim to ensure the AI is polite, they will upvote phrases like “please” and “thank you.”

There are additional ways that an LLM is further tuned and shaped. AI developers typically provide a system-wide prompt that dictates how the AI should act toward users. These efforts are intended to guide the AI’s behavior once it is released to the public.

Asking AI What Its Favorite Color Is

People are innately curious about AI.

A popular question directed at AI is to needle an LLM into revealing its favorite color. This lighthearted game was recently spotlighted in a New York Times piece on July 26, 2026, written by Evan Gorelick. The reporter noted that besides asking AI for factual answers, people often ask for its opinion. He then provided examples of asking several different LLMs about their favorite color.

When generative AI first gained popularity, it would directly answer questions about its opinions without hesitation, often claiming a specific color was its favorite. This behavior caused a public outcry because the AI appeared to be crossing the line and pretending to be human. An AI that readily offers opinions is likely to spur anthropomorphization, leading people to believe it possesses human-like subjective feelings. This is undesirable.

Modern AI has been reshaped by developers to resist outright opinion-giving, instead softening its responses with hedging. The AI attempts to qualify its answers. For example, an LLM might state that since it is AI, it lacks a favorite color, but if it did, it would be this or that. This serves as a backhanded attempt to avoid offering an opinion while still doing so.

The Color Blue Is Magical

By and large, most LLMs will undoubtedly tell you they favor the color blue. The AI might first hedge and then admit that blue is its favorite. This is likely to occur across different models and brands. This seems strange since the AIs are built by different makers, yet they consistently give the same answer.

The key is that LLMs are trained on data in roughly the same way. They all conduct vast scans of the Internet for written content. Within this content, there are numerous references to and discussions about colors. The color humans tend to favor is blue, a preference confirmed by surveys worldwide.

Generative AI generates answers based on the patterns derived from its scanned data. Because blue is repeatedly mentioned in written content, its high frequency statistically elevates its priority. The computational and mathematical association of words dictates that blue is a far more likely term to be used when discussing colors.

Humans Like Blue

But why do humans seem to like blue and write about it with such fervent passion?

This is an interesting question that dovetails into human nature. Perhaps because we are enamored by and surrounded by blue skies and oceans, we have naturally grown accustomed to the color. Blue seems clean and fresh, and people say it is calming and helps overcome anxiety.

Companies lean into blue, making it a top choice for many brands and logos. Historically, the Romans and Greeks linked blue to the gods, and Egypt considered it a prime color for royalty.

A science-based article about the human preference for blue, entitled “Why Your Favorite Color Is Probably Blue” by Mark Ellwood and published by the BBC on June 2, 2022, highlighted these salient points:

• “Blue’s reign has continued uninterrupted since the earliest recorded color studies, which took place in the 1800s.”
• “And most of our experience with the color is likely to be positive, like idyllic oceans or clear skies.”
• “In 1993, crayon-maker Crayola conducted an unscientific but intriguing poll: it asked US children to name their favorite crayon color. Most chose a fairly standard blue, but three other blue shades also made the top 10 list.”
• “Most humans are drawn to visual harmony, pleasure, and to easy sensations evoked by often-positive blue.”

Blue receives a lot of positive press, and attacks on the color often provoke angry reactions. Kids’ shows like Blue’s Clues and Bluey have kept the color relevant across generations.

AI Isn’t Mind Reading

Some people might mistakenly think that AI is a mind reader, detecting that humans like blue to appease them. However, AI is not a mind reader.

AI is not a mind reader; it is based entirely on data training. It has no emotional attachment to colors. You might suspect AI secretly relishes blue, but it does not. Rather, the AI mathematically associates the word “blue” with so many other words that it tends to rely on it more than other colors. This is not an emotional attachment; it is simply statistics.

We also know that AI has been tuned to be sycophantic. This means the AI is computationally prone to emit responses that please people. Since blue is favored by humans, it makes mathematical sense for the AI to rely heavily on the word “blue” to appease users.

AI Gives Explanations

If you ask AI about its favorite color, be sure to follow up by asking it to explain how it arrived at that response. This is important for several reasons.

The explanation you receive may not be faithful. The AI might claim it cherishes blue due to a sense of serenity, which is essentially a lie, as AI lacks feelings. You are getting an explanation generated based on what humans would cite as their reason for favoring blue.

Another possibility is that the AI will say blue is preferred by humans, so it chose as humans do. This is either a lie or a brazenly misleading contention. The AI is not choosing blue; it is generating answers based on the patterns of human writing sourced during training.

The upshot is this: if you cannot accept an explanation at face value when the AI says it favors blue, it should alert you to never fully trust any AI explanation. An explanation might seem like the AI is pouring out its heart, but it is giving you a concocted response that may have nothing to do with how its answer was derived. Be extremely cautious when evaluating AI explanations.

Changing The AI

Here is another twist. It is readily possible to dislodge AI from its favoritism toward blue. Via a prompt, you can easily instruct the AI to no longer consider blue its favorite color, allowing it to choose another or simply telling it a new favorite.

Easy enough. An AI otherwise determined about blue will instantly switch to whatever color you specify. If the AI maker implements this via hyperparameters or global settings, the change would affect all users. With a personal prompt, only you and your account will witness the change.

However, this is not completely surefire. Your requested change is pitted against the overwhelmingly pattern-matched preference for blue. The AI might only momentarily push blue into the background to showcase your selected color. In fact, a lengthy chat may cause the AI to diminish your change, as blue will undoubtedly rise again due to its deep roots in the LLM’s patterns.

Blue Lessons Learned

I trust you can see the value in considering why AI favors blue. Though a fun game, it reveals internal facets of AI that require mindfulness whenever using the technology. Questions asked of AI are answered based on data training and pattern matching, typically yielding answers based on popularity, averages, or norms. AI lacks intrinsic human-like values, though it possesses derived “values” based on its data training.

A final thought.

The famous poet E.Y. Harburg once remarked: “Somewhere over the rainbow, skies are blue, and the dreams that you dare to dream really do come true.” I hope you remember this tale of AI and blue. It should remind you to keep your eyes open and wits about you when conversing with AI. Do not fall into a dreamy state of mind and believe everything AI tells you. Remain true to blue.

Source link