When Did AI Art Become Popular?

From 1973 to 2021: The Long Road Nobody Noticed

It felt like AI art appeared overnight — one year your feed was normal, the next it was full of impossible portraits and astronaut cats. But the technology behind that "sudden" explosion had been quietly developing for nearly five decades. Popularity didn't arrive when AI art was invented; it arrived when the tools finally reached everyone.

The story starts in 1973, when Harold Cohen's AARON program began producing original artwork through hand-coded rules. Cohen became an inspiration to many of the contemporary AI artists of the 21st century, yet AARON remained a curiosity confined to galleries and academia. For forty years, machine-made images required programmers, not prompts.

The first real technical shift came in 2014, when Ian Goodfellow and colleagues introduced generative adversarial networks. GANs were the first systems that could learn an aesthetic directly from image datasets rather than follow explicit rules, and deep learning more broadly caused a significant shift in the world of AI art throughout the 2010s. The ingredients were finally in place — but almost nobody outside the field noticed.

Then came the moment mainstream institutions couldn't ignore. In 2018, the AI-generated portrait Edmond de Belamy sold at Christie's for US$432,500 — roughly 45 times its estimate of US$7,000–10,000. The sale marked the first time the traditional art world publicly acknowledged that machines could make something collectors would pay serious money for.

Still, recognition isn't the same as popularity. Between 2018 and 2021, the pieces sat waiting:

• DALL-E 1 arrived in 2021, hinting at what text-to-image generation could do
• Latent diffusion research was published in December 2021, slashing the compute needed to generate images
• Stable Diffusion's training data — roughly five billion images indexed by the nonprofit LAION — already existed, waiting to be unleashed
• Free and low-cost access models were proving that consumer demand for image tools was real

The lesson of this long quiet road is that capability alone never creates popularity. AARON ran for decades, GANs matured for years, and Christie's made headlines — yet none of it changed everyday behavior. What changed in 2022 was access: text-to-image tools became widely available to the public, allowing users to generate imagery with little effort. As the EFF put it, AI art creation became accessible to anyone with a weird idea and an internet connection.

That pattern repeats across the AI landscape today. Powerful techniques sit unused until someone wraps them in an interface ordinary people can actually use — the same principle driving the agent skills ecosystem, where instruction packages turn raw model capability into practical workflows. SkillzBench is building its free directory and benchmarking tools precisely because the gap between "the technology exists" and "the technology is trustworthy and usable" is where most of the real value lives.

The road from 1973 to 2021 proves the point: AI art didn't become popular when it became possible. It became popular when it became easy.

2022: The Year AI Art Went Mainstream

In 2022, AI art stopped being a research curiosity and became something anyone with a weird idea and an internet connection could do. Three text-to-image models — DALL-E 2, Midjourney, and Stable Diffusion — launched to the public that year, and image generation became as simple as typing a prompt.

The convergence of several factors explains why this moment stuck. According to Wikipedia's history of AI art, text-to-image models "became widely available to the public, allowing users to quickly generate imagery with little effort" — a decisive break from earlier systems that required technical expertise.

• Radical accessibility: generation required no coding, no GPUs of your own, and no artistic training — just a text box.
• Massive training data: Stable Diffusion learned from the LAION-5B dataset of roughly 5–6 billion images indexed by the nonprofit LAION, per EFF's analysis.
• Free and low-cost entry: Stable Diffusion shipped as free open source, and later free tiers from tools like Bing Image Creator lowered the barrier further.
• Viral social moments: headshot avatars remixed into superheroic and anime styles spread across Twitter in 2022, putting AI imagery in front of millions.

The scale behind these tools matters. A model trained on billions of images could produce convincing output for almost any prompt, and open-source release meant anyone could run, modify, or build on it. Zapier's overview of the category notes that what started with DALL-E 2 grew into a movement with AI-generated work appearing on magazine covers, in Hollywood productions, and across social media (Zapier).

The economics reinforced the momentum. MarketsandMarkets attributes the market's growth — from USD 8.7 billion in 2024 toward a projected USD 60.8 billion by 2030 — to generative AI's ability to "reduce the cost and time of media production at scale" (MarketsandMarkets). When a technology is cheap, fast, and open to everyone, adoption stops being a question of if and becomes one of how quickly.

That speed also created the fragmentation professionals now navigate: dozens of generators, many "essentially just different user interface skins for the same art-generating algorithms," embedded into everyday tools like Canva, Photoshop, and Bing. For teams building automated visual workflows on top of these models, the challenge is knowing which tools and agent skills actually perform — which is precisely the problem SkillzBench is building toward, with a free directory that benchmarks and security-scans agent skills so builders can equip their agents with confidence. The waitlist is open now for those who want to be first in line.

Corporate Adoption and the Legitimacy Fight

Nothing announces a technology's arrival quite like a Fortune 500 company handing it to consumers. In March 2023, Coca-Cola launched "Create Real Magic," the first platform combining GPT-4 and DALL-E for consumer-facing art creation, inviting digital artists to remix the brand's iconography. CMO Manolo Arroyo framed it as a speed revolution, saying AI would "increase the velocity from weeks to days" in creative content production.

That same year, AI image generation quietly stopped being a standalone product and became a feature embedded in tools people already used. As Zapier's analysis of the category puts it, AI image generation is "increasingly a feature and not a product," integrated into Canva, Adobe Photoshop, Bing, and X/Grok. Corporate partnerships followed: Microsoft–Adobe integration arrived in Microsoft 365 in March 2024, AWS launched its Titan Image Generator in late 2023, and Google partnered with WPP in April 2024. The market reflected this momentum, growing from $5.1 billion in 2023 to $8.7 billion in 2024, with MarketsandMarkets projecting $60.8 billion by 2030 at a 38.2% CAGR.

Yet mainstream adoption arrived alongside a fierce legitimacy fight. When Jason M. Allen's Midjourney-created "Théâtre D'opéra Spatial" won the 2023 Colorado State Fair art competition, the debate over whether AI output counts as art reached a boiling point. A 2023 survey found 76% of respondents said AI-generated content should not be considered art, against 24% who said it should. In August 2023, a US district court ruled that AI art was ineligible for copyright protection because it failed to meet the human authorship requirement.

The artist community's objections were more specific than aesthetic. According to aggregated survey data:

• 74% of artists consider AI scraping of their work unethical
• 89% worry that copyright laws are outdated
• 54.6% fear a decrease in income
• 70% of US adults believe artists should be compensated when AI uses their work

The courts became the battleground for those concerns. Stable Diffusion faces two lawsuits — an artist class action and a suit from Getty Images — while Midjourney is currently being sued by Universal and Disney. These legal challenges are less a sign of failure than a signal of arrival: industries only litigate technologies that matter.

For businesses navigating this contested landscape, the practical takeaway is that AI image generation is now a professional creative tool, with all three leading platforms permitting commercial use. And as image generation embeds itself inside everyday software, agent skills can automate the surrounding workflow — sourcing, editing, and publishing visuals. SkillzBench is building a free directory of vetted agent skills plus a free MCP connector so teams can browse, benchmark, and safely equip the skills they choose; the waitlist is open now.

What the Numbers Say: Market Growth and Real Adoption

Numbers rarely lie about momentum, and the AI art market's numbers are staggering. According to MarketsandMarkets, the AI image and video generator market grew from $5.1 billion in 2023 to $8.7 billion in 2024, and is projected to reach $60.8 billion by 2030 — a compound annual growth rate of 38.2%.

That projection deserves one caveat. AIPRM's statistics aggregation puts the global AI image market at just $0.9 billion by 2030, up 254% from $0.26 billion in 2022. The gap reflects differing market definitions — one counts the broader generator ecosystem, the other a narrower slice — so treat any headline figure with healthy skepticism.

Adoption data tells a similar story of scale. Midjourney holds roughly 26.8% market share and DALL-E about 24.35%, meaning these two platforms alone command around half the market. Stable Diffusion-based models have generated more than 12.5 billion images — roughly twelve times more than the next platform — while Adobe Firefly, Midjourney, and DALL-E 2 have each crossed the one-billion mark.

The productivity case explains why businesses keep adopting these tools:

• Generative AI tools can enhance artists' performance by 25%, according to PNAS Nexus research cited by MarketsandMarkets.
• The same research found a 50% boost in the likelihood of positive peer feedback for artists using these tools.
• A survey of 480 marketing professionals found that perceived usefulness is the critical mediating factor driving adoption of AI image generators.
• MarketsandMarkets attributes much of the market's growth to AI's ability to reduce the cost and time of media production at scale.

Corporate behavior confirms the trend is structural, not a fad. Coca-Cola launched "Create Real Magic" in March 2023, the first platform combining GPT-4 and DALL-E for consumer-facing art creation. Microsoft integrated Adobe's tools into Microsoft 365, AWS shipped its Titan Image Generator, and AI image generation has become "a feature and not a product" — embedded in Canva, Photoshop, Bing, and X/Grok rather than confined to standalone apps.

For professionals, the practical takeaway is that image generation is now infrastructure. And just as image generation became a feature inside everyday tools, agent skills are doing the same for AI workflows — SkillzBench is building a free directory and MCP connector so you can browse, verify, and benchmark which skills actually lift your agents' performance, with the waitlist open now.

Putting AI Images to Work: From Novelty to Workflow

Knowing the history is one thing; putting AI images to work in your own workflow is another. The good news is that the same accessibility that made AI art mainstream in 2022 now works in your favor — generation is as simple as typing a prompt, and free options abound, from Bing Image Creator to Canva's free tier.

The first step is picking a tool that matches your actual need. As one comparison of leading platforms puts it: "For maximum privacy, Stable Diffusion running locally is unmatched. For quality and convenience, Midjourney leads. For integration with AI assistants and text handling, DALL-E 3 excels." All three leading platforms permit commercial use, so professionals can generate assets for client work, marketing, and product design without licensing ambiguity.

A few practical guidelines:

• Match the tool to the task. Midjourney holds roughly 26.8% market share among AI art tools, and its reputation for visual quality makes it a strong default for brand imagery and concepting.
• Start with free tiers. Leonardo.Ai offers 150 free tokens per day, and Stable Diffusion is free and open source — enough to test prompts and styles before committing budget.
• Favor integration where it saves time. AI image generation is increasingly "a feature and not a product," embedded in Canva, Adobe Photoshop, and Microsoft 365 — using it inside tools you already pay for keeps the workflow tight.

Treat the generated image as one step, not the finish line. A market analysis cites PNAS Nexus research finding that generative AI tools can enhance artists' performance by 25% and boost the likelihood of positive peer feedback by 50% — gains that compound when generation feeds into resizing, captioning, publishing, and iteration rather than stopping at a single output. That's where agent skills come in: instruction packages that let AI agents handle the surrounding steps of a visual-content workflow automatically. SkillzBench is building a free, independent directory of these skills — with security scanning, sandboxed benchmarking, and a free MCP connector to load them into your agent — and the waitlist is open now for readers who want to browse and test free agent skills at launch.

The category is moving fast — one industry observer notes that "by the time you read this, there might be even more great apps available." Choose a tool for the job, start free, and build the rest of the workflow around it.