Originally published here.

Almost 70 years ago, Marshall McLuhan first wrote that “the medium is the message”. He argued that each medium reshapes society more through its form than its content. The features of a medium, like the format of a television broadcast or the character limits of a social media platform shape our perceptions, behaviours, and interactions. Ultimately, they influence our individual and social experiences more than the actual words or images themselves.

Consider the following examples. In 1960, the first televised US presidential debate - Nixon vs. Kennedy - aired on both TV and radio. On TV, Kennedy’s composure and good looks shaped audience perceptions at least as much as his words. Radio listeners, lacking visual cues, focused their evaluation on the candidates’ arguments and delivery. TV’s visual focus changed how many viewers judged the exchange, illustrating McLuhan’s point that a message’s form can rival a message’s content. Looking at more recent mediums, Twitter’s character limit rewards punchy, emotive takes, quote-tweets and headlines. This format shifted discourse toward brevity and performative provocation, regardless of topic. And last but not least, TikTok’s feed and instant skip pushes creators to front-load a hook in the first seconds, reshaping music structure, and rewarding visual gags over exposition.

The most skilled people have long understood that beyond their message, it is their ability to perform on a given medium that fuels their influence.

Three years ago, when AI had its breakthrough moment with the advent of ChatGPT, there was a heated debate about whether AI itself marks a new medium. Most critics said no. It is software and infrastructure that automates and personalises production inside existing media, rather than a new channel with its own grammar or audience structure. Until today, sceptics argue that AI is not bigger than the Internet transformation, which entirely reshaped how we communicate and live (think social media or online shopping). They point out that few would say AI has changed their lives in the same way. From this view, the Internet is the medium, and AI is merely another tool that runs on it.

I believe this is wrong. A medium, at its simplest, is the channel through which content flows from sender to receiver. When following this definition, AI clearly marks a new medium, as content that would have previously been posted to online forums or search engines now get posted to LLM-powered chat windows instead.

But more importantly, the grammar around AI is also changing. The good test for an influential shift in our social structures are brands that become verbs (to uber) or entirely new words that become part of our daily vocabulary. Previous medium shifts came with their own special vocabulary. With the rise of television we began endless zapping, watched senseless talk-shows and Trash-TV and got hooked on cliffhangers. Since the Internet era, we email, google, binge-watch or Netflix and chill.

Only three years into generative AI, we vibe-code and some of us suffer from AI psychosis. The emdash is the linguistic phenomenon du jour, a marker of AI slop: low-effort, low-quality AI-generated content that clogs feeds. Companies are building “the Lovable of X.” Much of these words may only be familiar to the very online so far, but this will change rapidly.

Yet, the real and last test of what marks a new medium is whether it shifts audience structures: how society at large communicates (one-to-one, one-to-many, many-to-many) and creates (text, sound, image, video).

Consider past examples of medium shifts.

Each new medium builds on the last, typically expanding modes of creation and communication. The printing press expanded access to text. Radio added sound. Television added moving images. The Internet enabled user-generated content that could be shared between many people via social media platforms. What was previously beholden to expert editors more and more consumers were able to create for themselves. The rise of the creator economy began.

In comparison to previous media, the key characteristics of AI are its generative, interactive and private nature.

Mode of production - AI is generative: In traditional media, content creation was limited to professionals and then broadcast or distributed. Social media lowered the barrier to production by letting anyone create and share content, but what was produced still had to be written, filmed, or designed by humans. AI goes further by collapsing the act of production itself. It doesn’t simply publish what people make but instead makes with them in real time. Truly everyone is now a creator.

Mode of interaction - AI is interactive: Unlike books, radio, or television, AI adapts completely to what we ask of it. It is the most interactive form of content and content creation, only matched by gaming, yet vastly more expansive.

Mode of consumption - AI is private: Unlike traditional media that were one-to-many (TV, radio, newspapers), many-to-many (social platforms), or one-to-one (private chats), AI is the most intimate medium yet: one-to-machine. We shape it with natural language much as we shape our own thoughts and dreams.

Last but not least, taken together, the interactive and generative features of AI enable people to go one-to-machine-to-many: a person co-creates with a model, and then scales to a crowd.

Communication scholar John M. Culkin wrote in 1967 that “we shape our tools, and thereafter our tools shape us.” AI is poised to be the most consequential medium yet: it scales like mass media, feels as intimate as private thought, and is as interactive as games. Early signs of the medium shaping the message are everywhere.

AI is influencing how we copy-paste write. Bullet-point openings, three-part slogans, energetic verbs, and “not X but Y” reframes. What were previously individual, imperfect voices are merging into a homogeneous, AI-generated tone. A/B testing loops drive convergence on what performs well. 10 years into despising clickbait, we still have not dropped it because it drives eyeballs.

At the same time, AI avatars and content on auto-pilot are rising. Individual creators are building multi-million-pound businesses by automating content, producing output at unprecedented scale and detail. When a person is a configurable avatar, the message shifts from “who said it” to “which AI-generated version of you performs best.”

While firmly believing AI will unlock incredible creativity, I think this trend will inevitably also lower content quality - at least to the aesthetic mind. I learned this from several years of performance marketing at Instagram and Facebook: It’s not the most thoughtful, deep or even beautiful posts that perform best, aka drive sales. Instead, it’s bright colours and extreme calls-to-action. Marketers starting out with noble hopes for their campaigns often ended up in clickbait territory. When it comes to the AI medium being the message, we may see AI-induced “personality drift”: even the most nuanced thinkers and personalities inevitably developing into AI-powered hyper-engagement machines. If everyone can create on auto-pilot, is the risk that everyone sounds the same?

Earlier media prepared us for this moment. Each shift expanded what we could process, from text to sound to video, and increased both speed of access and the privacy of exploration. Just as adolescence is a buffer between childhood and adulthood - a space to practise freedom and responsibility - the Internet has been our rehearsal ground for the deeper privacy and powerful creation capabilities that AI now makes possible.

We are, however, confronting another shift before we have finished adapting to the last one. It took nearly five centuries to move from the printing press to radio. It took about 30 years to move from radio to television, then roughly 40 years from television to the Internet, and only about 10 years from the Internet to social media. The pace has collapsed from centuries to decades.

We have barely processed the consequences of social media, a medium that has reshaped human perception, behaviour and interaction, and we are already in the midst of another upheaval. And so the real question is not if AI is another medium (it most certainly is), but whether we are ready - or will end up like teenagers forced to become adults overnight.