Digital's final form: The end of visual craft

With growing demand for rapidly produced visual media, I fear we are undermining the effectiveness of our communication, and eroding its potential for beauty.

Visual communication is how we express ideas, raise awareness, tell stories and prompt action through the use of graphical techniques, and mediums.

All around us, visual design facilitates contemporary life. Road signs, computer operating systems, websites, medical scans, voting forms, smartphone apps, posters, and game interfaces all depend on considered visual design.

If a picture is worth a thousand words, then in 2025, images have never had more to say. Once upon a time, we were oral societies. Today, we are firmly visual.

We scroll, swipe, tap, and skim muted videos with captions as we interface with modern life, absorbing everything through our eyes.

Our storytellers use slides, and our memories need cloud space. We no longer recall, we record.

The rate of change in visual communication

Creative industries have weathered seismic shifts in how visual communication is produced. The printing press disrupted the work of scribes; desktop publishing revolutionised graphic design; CAD (Computer-aided Design) transformed drafting; and CGI (Computer Generated Graphics) reshaped animation. Each step in the march of progress forced creators to adapt.

Like many of my generation who entered the design world at the turn of the 21st century, I was drawn to graphic and web-based design by accessible desktop publishing tools. Even the most rudimentary “what you see is what you get” (WYSIWYG) interface served as a gateway into the discipline.

By riding the digital wave in the early 2000s, I soon learned that using Photoshop didn’t make me a designer. Tools alone couldn’t compensate for a lack of core visual competencies. Mastery of colour, hierarchy, typesetting, and composition still mattered, and my formal design study became essential—not just to manipulate 2D surfaces, but to understand the structures beneath them.

Since then, the internet has became ubiquitous, and widespread access to feature once reserved for commercial-grade tools has created a cacophony of visual imagery. Overwhelming and disorienting, visual noise pollution across endless algorithmic streams. For creators, cutting through is hard, and for audiences, focus is nearly impossible.

As attention spans shrink, we find ourselves in an attention economy which dilutes communication.

The decline of visual literacy

Given increased societal reliance on visual communication, it is disappointing, if not surprising, that many professionals struggle to communicate using visual techniques.

Throughout my career, I’ve observed visual communication skills routinely dismissed as mere artistic flair. Even in school, creative ability was rarely valued on par with athletic performance. Sport was framed as a matter of grit, discipline, and perseverance—drawing, on the other hand, a god-given luxury.

Curious to test these observations, I turned to the data, finding support in a piece published by Schools Week. It highlights troubling patterns in the UK’s education system during a curriculum review, noting:

Art and design is certainly popular, but the reality is that children are getting less of it across the board, and the quality and relevance of provision are in serious decline ... Since 2019, entries for GCSE design and technology (D&T) have fallen to catastrophic levels ... Art and design isn’t thriving; some of it has just moved to D&T workshops – masking the very real decline in both subjects.

The article also cites a report, School Art: Where Is It? (Re)exploring Visual Art in Secondary Schools, which reveals a significant reduction in student exposure to three-dimensional materials, creative techniques, and design-based ways of thinking.

This lowered appetite for design education validates the lack of value placed on those who are able to produce the very artifacts we desire.

AI, and why taste won't save you

The era of artificial intelligence is not a sudden disruption but the natural progression of technological trajectory throughout history. The curiosity of The Renaissance (1400–1600) sparked discoveries that paved the way for The Industrial Revolution (1760–1840). In turn, the mechanisation of labour reshaped societies and set the conditions for The Information Age that began during the 1980s.

Since the 1990s, we’ve been on a steadfast trajectory toward automation—not just of physical labour, but of thought and creativity. Paved with good intentions and technological optimism, that path has led us to AI.

AI nears digital progress to its final form, one that enables everything, everywhere, all at once, with both immense potential and profound risk.

In a world where generative AI has been commoditised, and where apathy toward genuine expertise is on the rise, it’s unsurprising that AI is most often embraced as a shortcut to faster visuals, rather than a tool to improve meaningful communication.

What follows is the quiet dismissal of those once valued for their creative craft. Designers, already reduced to “Mac operators” can't rely on taste to redeem them in the eyes of an economy which requires more and more speed, scale, and convenience.

As aesthetic judgement sets into a, “I’ll know it when I see it” context, even “taste” itself will be redefined. Any skill required to manually craft visuals will be relegated to nostalgia: preserved in vintage stores and artisan markets.

The generative AI creation process, and its impact on information

Before unpacking the effects that generative AI is already having on visual communication, it’s worth taking a moment to understand what actually happens when an AI model generates images.

The first step in the process is for the model to “learn” what images are. It does so by analysing vast quantities of visual data, identifying patterns which help it determine how humans expect images to appear.

This process begins by converting each image into a series of numbers—a grid of pixel data. In the example below, a portrait of Abraham Lincoln is broken down into numerical values that describe each pixel's colour and position.

The pixel data is contextualised by categorising features such as medium, subject matter, and visual traits. The example above is not just Abraham Lincoln, but rather, “A portrait of Abraham Lincoln taken around 1865 using a silver-coated copper plate to produce a quarter-plate daguerreotype, which is an early form of photography.”

The process is repeated with countless images of all sorts, enabling the model to identify various visual facets, such as colour, relationships between objects, textures, the effects of light, and so on. It thus establishes visual conventions: that sunsets typically feature orange gradients near the top of the frame; that mountains tend to have jagged contours; that a head-and-shoulders framing likely represents a portrait.

This isn't vision as humans know it but an extraordinarily detailed technical grasp of how images are structured and represented. Armed with this statistical understanding of what has come before, the model can now respond to prompts by generating derivative images that approximate human expectations by mimicking various styles and media.

Once trained, the model can respond to requests like, “Generate an image of Abraham Lincoln.” Such a prompt is translated via Natural Language Processing into a numerical value the model can interpret, like so:

1423 281 423 286 11245 37129 13

These numbers represent coordinates the model can use to navigate a complex mathematical environment known as “latent space”. Within this space, the model searches for a visual representation that best aligns with the prompt's statistical footprint.

Once the prompt has been interpreted, the model finally begins to generate an image, typically using either a Generative Adversarial Network (GAN) or a Diffusion Model.

Diffusion models are particularly compelling. Rather than drawing an image directly, they begin with a field of static and gradually refine it. By repeatedly asking itself, “What should this look like?” it gradually removes small portions of the noise in response.

With each cycle, the image becomes slightly more coherent. The model predicts whether areas of noise should become a “nose,” an “ear,” an “eye”. These predictions compound, step by step, until a recognisable image emerges.

This is a visual example of the process for a portrait of Abraham Lincoln:

There’s a strange irony in the process of creating an image through “de-noising” which contributing to an ever-louder sea of digital noise.

This new representation of Abraham Lincoln doesn’t exist as a painting, or an historical photograph. It wasn’t captured, composed, or witnessed. Summoned by statistics and labelled, “Portrait of Abraham Lincoln”, waiting in some future search result to be mistaken as truth.

The deceptive quality of AI's rapid advancement in image creation

As demonstrated above, AI can be an amazing mimic, improving all the time at approximating the visual communication humans would create themselves.

In short order, AI image generation has gone from interesting abstract visualisation, such as the piece, Space Opera Theater, which won an award at the Colorado State Fair in 2022.

To producing body-horror hands when attempting photo-realism.

Until landing in the uncanny valley of beauty-portraits, like this one created by Mid-Journey for Vogue.

Impressive counterfeits which are enticing in a productivity context, but deceptive because while they are proclaimed to be artistic and fictional, they impact non-fiction.

An innocuous, AI-sparked, trend can help us unpack the disturbing impact AI generated imagery has on the canon of human knowledge, all under the guise of fun.

Ghibli-style, and the misinformation age

In March 2025, a viral trend, driven by new image-generation features in ChatGPT swept across social-media. Users began posting AI-generated images in the style of Studio Ghibli. Some were reinterpretations of popular memes, others, “Ghiblified” self-portraits.

At the time, TechCrunch shed light on the scale of the phenomenon. According to Brad Lightcap, COO at OpenAI, over 130 million users had generated more than 700 million images since the upgraded image generator launched on March 25. Even if only a fraction were Ghibli-style, the numbers highlight just how quickly trends can spiral on social platforms.

Inevitably, in their quest to stay culturally relevant, corporate marketing teams got involved. This X (Twitter) post by zomato is but one example of how brands jumped on the trend.

While murky questions around copyright, cultural ownership, and business ethics soon arose. The use of such a distinct and recognisable aesthetic—created by a specific studio with a long, human-centred legacy—raises concerns that go beyond standard “fan art” debates.

The Studio Ghibli trend challenged a long-held assumption: that AI tools, like Photoshop before them, would simply help us manifest our own ideas.

The trend wasn't like an earlier cultural moment, 2012’s #bendgate, which inspired KitKat's clever reaction to re-iterate their famous “break” slogan. Nor was it akin to Murphy’s Irish Stout‘s 1997 campaign, which commissioned the creators of Ghost in the Shell to make their anime inspired Last Orders commercial.

The Ghibli trend solely involved automating style, distilling the surface of an iconic brand into an easy-to-replicate aesthetic. The implication is that now, a Google search for, “Ghibli images” returns a mix of actual frames from Studio Ghibli films along with AI-generated content churned out during the passing fad.

The long-term significance of such counterfeits is still unclear, but it raises a sobering question: will Studio Ghibli come to signify a “vibe,” rather than the name of a studio responsible for deeply human, hand-crafted works?

The Studio Ghibli fakes may seem harmless, but what happens when the same dynamic affects scientific records, historical archives, or educational material?

Our inability—or unwillingness—to regulate digital content, out of fear we'll stifle innovation, may lead to the further erosion of facts.

Decoration or information

The difference between art and design has always been intent. When we use generative AI to decorate or embellish, its use may seem harmless. When we use it to inform, however, the stakes rise dramatically.

As a designer working in K–12 education, I’m acutely aware of the risks involved in asking AI to generate informative visuals. As AI-produced imagery becomes indistinguishable from authentic sources, how will we distinguish truth from fabrication?

Take the now-infamous “rat with big balls” diagram. Despite being riddled with anatomical inaccuracies, it found its way into a scientific journal.

Science Integrity Digest summed the situation up bluntly, “the paper is actually a sad example of how scientific journals, editors, and peer reviewers can be naive – or possibly even in the loop – in terms of accepting and publishing AI-generated crap.”

The problem, however, runs deeper than negligence. Another study, Experts fail to reliably detect AI-generated histological data, revealed that even trained professionals can’t always tell the difference between what is and isn't real.

Complex visuals, like microscopic data, cellular diagrams, or radiological scans can be so intricate that even domain experts may be fooled by fakes.

We once trusted reference images to communicate complex ideas in order to document, instruct, and preserve knowledge. Today, however, not even diagrams, photographs, and videos published in academic journals can be assumed reliable.

With fewer people educated in the foundations of visual communication—the principles of clarity, accuracy, composition, and form—our collective ability to spot misinformation declines further.

This paints a bleak picture of the future: one in which we are visually overwhelmed, informationally uncertain, and critically underprepared.

AI isn't killing the quality of our communication, we are

I’ve always been an early adopter. I use AI regularly and stay close to its developments. My concerns aren’t with the technology itself, but with the short-term choices we’re making. Choices that will shape how well we communicate in the long term.

As the earlier examples show, I’m sceptical that AI will inherently improve our ability to communicate visually.

It’s far more likely to be used to generate more memes, which can have a detrimental effect on communication.

Memes are the ultimate shorthand: fast, funny, and often clever. While I enjoy a good meme as much as anyone, they often rely on layers of shared context, inside jokes, or cultural references to make sense.

In a way, we risk becoming like the Tamarians from the Star Trek: The Next Generation episode Darmok. The Tamarians speaks only in metaphor, and say things like, “Darmok and Jalad at Tanagra” to convey complex ideas. The phrase has meaning to them, but without knowing the reference, communicating with them is impossible.

There’s a meme about that too, of course:

So meta. So many levels.

Junk mail converts, so everything becomes junk mail

Along with the impact on communication, a lack of broad visual communication skills also risks ushering in an age of its diminished beauty and significance.

It’s not that AI can’t produce aesthetically pleasing images but like much of what floods our feeds, these fast designs are often shallow. Byproducts of a race to commodify attention rather than communicate meaning.

The term “junk mail” refers to unsolicited advertising—cheap, mass-produced flyers delivery directly to value-driven consumers. Once deemed low-brow, this tactical approach to marketing has become the norm across almost all brands who advertise digitally.

To explore how far we've drifted from carefully crafted design, consider a classic example of visual communication: Volkswagen’s iconic, “Think Small” or “Lemon” ads.

Still cited as the peak of creative advertising, these campaigns demonstrate the layered artistry of effective visual communication. The concept, branding, copywriting, typography, photography, composition, and layout are all executed with clarity and restraint.

Beyond aesthetics, these ads were strategic, helping Volkswagen overcome scepticism tied to its Nazi-era roots, and ultimately becoming the top-selling imported car in the United States.

We don't advertise like this any more because advertising has shifted from persuasion to prediction. Sophistication in marketing now refers to algorithmic optimisation. Hyper-personalised ads are engineered to re-enforce biases rather than change minds. Design in this context isn't about communicating value, it's about conversion.

Today, a digital ad for Volkswagen looks more like this:

In the example, the van itself is almost an afterthought. The financing deal feels like the real product. It's unclear what’s being sold, beyond the idea of affordability. A far cry from Ogilvy's rule, “make the product itself the hero of your advertising”.

As a hyper-targeted ad served to someone actively searching for a financed work van, perhaps it’s effective. Its meaning, however, is limited to that transactional moment.

In this way, all advertising begins to resemble junk mail: Direct to consumer, cost-focused, and context-blind. To test how AI interprets this trend, I asked ChatGPT to generate a typical Facebook ad for Volkswagen using asking it to, “Create an example advertising image for a Volkswagen auto-mobile that a prospect may see on their Facebook feed?”

Interestingly, it didn’t object on the grounds of copyright, but rather generated the image below (on the left).

When I asked it to make the ad “more clever,” I received the version on the right.

Would these types of ads “work”? Probably. Facebook’s algorithm could no doubt serve them to someone likely to click. If I were Volkswagen’s marketing manager, I could improve them and generate dozens of variations to run a perfectly tuned campaign.

The productivity gains are evident. So, everything becomes junk mail.

Is there any beautiful visual communication left?

I'm not calling for the return of the Mad Men, nor lamenting that we no longer get all our information from printed pages. It’s hard to ignore however, that expert visual communicators had clearer roles, and greater influence, before the dominance of digital platforms.

Great typography, imagery, and layout do still exist, but they’re increasingly rare, often tucked away in fading print media. COLORS Magazine, for example, continues to demonstrate how design can be both beautiful and profound, using visual communication to illustrate, and elevate information.

That level of craft is difficult to replicate in the digital realm due to the speed and scale demands of online publishing. There are exceptions, of course. Google Maps is a masterclass in interface design. It’s also however, a singular achievement. Now that it exists, no one gets to design urban maps again.

Even premium digital experiences, like TIME Magazine’s feature articles, tend to fall back on templated layouts. These two articles share a common layout, leaving the hero image to do all the heavy creative lifting.

Unlike magazines, where content and design are developed in tandem, digital publishing systems are intentionally built to enable content creation and publication without requiring a designer at all.

In a media landscape driven by immediacy, where content must be shipped in near real-time, there’s simply no commercial incentive to invest in bespoke layouts.

Beauty survives in pockets, but visual craft is no longer core.

What's coming, and what we might do

AI advocates lean on the promise of productivity gains, and efficiency. In reality, the value of design for the screen age has already been defined: not to evolve, but to be executed faster.

“Efficiency” is the word that gets buy-in, but it implies more than just speed—it suggests effectiveness, and in many digital experiences, that promise remains unfulfilled.

Automation has long been a feature of tech, but rather than refining software for better self-service, companies have scaled by deflecting users to agents. That’s why many digital interactions eventually lead to a catch-all call centre.

Offshoring those human agents saved costs, but not quality. Often resulting in frustration for customers, and inefficiency for businesses. Now, with AI agents, companies see a way to reduce costs further, by automating anything we haven't attributed with inherent human value. Even creativity.

We're already living in a world where a “clever” ad from Artisan encourages employers to stop hiring humans.

The ad boasts that its AI, “won’t complain about work-life balance,” in a headline riddled with poor kerning and an awkward title case. It shows a vaguely human face mapped onto a 3D grid, like something ripped straight from a low-budget 80s sci-fi straight to VHS movie.

Artisan promises to automate all your outbound sales. All those same annoying experiences, just without the humans.

We can look forward to all this, even before we reckon with the broader societal impacts of mass unemployment.

Slow food, and clean cities: can we positively regulate?

Can we expect a revolt against the noise pollution and constant bombardment of digital junk mail?

Maybe, and perhaps the Slow Food movement can offer some inspiration.

Beginning in 1986 when Carlo Petrini, and others, handed out bowls of pasta near the Spanish Steps in Rome, opposing the opening of a McDonald's. This humble protest which used Penne to symbolise traditional, local food, was simple, yet powerful.

Since then, Slow Food has grown into a global advocacy network. It has influenced regulation across Europe, contributed to more thoughtful labelling laws, and brought attention to GMOs. It successfully reframed the conversation around trade and sustainability by reconnecting food with culture.

Slow Food invited people to consider the origins, ethics, and environmental impacts of what they consumed—not just for utility, but for value. Essentially, asking,“if we don't value our food, can we value our culture?”

In much the same way, we must now ask what’s being lost in our visual culture. The survival of books and vinyl—despite their diminished utility—is a small sign of hope. A reminder that visual artefacts can have value beyond function.

Ironically, during the early days of the internet—despite its flaws—there seemed to be more meaningful web experiences than the optimised, hyper-personalised feeds we scroll through today.

Reclaiming public spaces

A more literal revolt against “visual pollution” happened in 2007, when São Paulo passed the Lei Cidade Limpa (Clean City Law). A bold attempt to reclaim public space by banning outdoor advertising.

The implementation was swift and dramatic, leading to the removal of thousands of billboards and store front signs. Before and after photos are striking, with store fronts stripped to reveal the architecture beneath.

The change wasn't without consequences. In the short term, it negatively impacted the advertising industry, and led to the unintended loss of public murals. The disruption, however, also forced creativity, leading advertisers to try less traditional strategies, such as large-scale graffiti murals. Such creative still promoted products and brands, but resembling art, and where more timeless ads, like Volkswagen's, “Think small”.

Whatever the cost, it proved that collective will, leadership, and regulation were required to effect meaningful change.

If social media platforms are indeed today’s public squares, then reclaiming those spaces may require similar efforts. Curation of personal feeds is no longer the answer, as “For You” algorithms have prioritised engagement and ad revenue above choice.

In summary

Visual media dominates our daily lives, but the quality of that communication is quietly deteriorating. The same tools that enable rapid production—templates, algorithms, generative AI—reduce much of it to noise.

Like many, I use AI and try to apply it thoughtfully. This isn’t a rejection of technology, but a call to reconsider how we use it, and whether the direction we’re heading truly serves our culture and our ability to communicate.

As an individual contributor, my influence is limited. I’ll do my part to stem the flood of derivative visual content now upon us: an abundance of pulp at best, large-scale misinformation at worst.

AI holds enormous potential—unlocking creativity, accelerating medical breakthroughs, and beyond. In the short term, however, as it begins to handle the bulk of our creative output, it’s more likely to amplify hyper-personalisation, echo chambers, and social isolation.

We risk becoming a reactionary generation—processing visual data, but no longer truly communicating.

I wonder: what happens if AI decides it has seen enough? When it begins to service everything from recycled patterns of the past?

With nothing new under the sun, will we become numb to all the noise? Will communication fade into a fuzzy silence, as the end of craft means communication is wholly lost.

Let’s hope it doesn’t come to that. Let’s remain resilient—and hold fast to the crafts responsible for so much of our cultural symbolism, inspiration, and shared understanding.