The Great Flattening
Competence is the baseline. AI tools have pushed brand design toward the middle. Research and Cannes 2026 show the three things that still cut through.
Open a feed. Open a pitch deck. Open five SaaS homepages in a row. Somewhere around the third one you feel it: you have seen this before. The same soft gradients. The same rounded sans-serif. The same three-word value proposition sitting above a hero image that could belong to anyone. The same tasteful, muted palette that says: we are professional, we are considered, we are not trying too hard. Everything is competent. Almost nothing is memorable.
This is not a SaaS problem specifically. Open the beauty aisle. Open a DTC clothing brand. Open a fintech landing page. The feeling is the same across all of them. Call it the Great Flattening: the convergence of an entire market's visual output toward a shared, undifferentiated middle ground, where everything looks technically correct and nothing looks like anything in particular.
The Cannes Lions industry grappled with this directly in 2026. The research is now catching up with what practitioners have been sensing for two years. And the work that actually wins tells a story that runs directly counter to the direction most brands are heading.
What is causing the Great Flattening?
The Great Flattening is not a conspiracy. It is not laziness. It is the predictable result of two genuinely good things happening at the same time, pushing in the same direction.
The first is the democratisation of professional design tools. Figma, Adobe, Canva, and their successors have made clean, polished visual production accessible to teams of any size. A two-person startup can now produce work that would have required a full design department five years ago. The floor for visual quality rose across the entire industry at once. That is an unambiguously good development.
The second is the arrival of generative AI. Teams that once waited days for creative iterations can produce them in minutes. Volume became cheap. Speed became unlimited. That, too, is a genuine capability gain. The problem is not the tools. The problem is that both forces push visual output toward the same place.
Democratised design tools tend to converge toward established best practices, because those are what they are built to teach and reproduce. Generative AI tends to converge toward the statistical centre of its training data, because plausibility is what it is optimised to produce. When you combine unlimited volume with a systemic bias toward average, you do not get more diversity. You get a lot of output, and you get a lot of sameness.
The mechanism in plain terms:
More tools and more generation do not automatically produce more distinctiveness. They produce more of the mean. Distinctive work requires a brief that knows what it is trying to avoid — and that is still a human judgment.
The research confirms it
In January 2026, researchers Arend Hintze, Frida Proschinger Åström, and Jory Schossau published a study in the journal Patterns (Cell Press) that documented exactly what practitioners had been sensing. They built autonomous feedback loops between an image generation system and a language-vision model: image to caption to image, repeated across 700 trajectories with diverse starting prompts and seven temperature settings over 100 iterations. Every run converged. The output collapsed toward just 12 dominant visual motifs — scenes that were pleasant, polished, and devoid of any specific meaning. The researchers named the result visual elevator music.
The finding that matters:
The systems were not malfunctioning. They were doing exactly what they are designed to do: produce plausible output efficiently. The researchers' sharper point is that producing endless variations is not the same as producing innovation. A system can generate millions of images while exploring a tiny fraction of what is visually possible. Plausibility and distinctiveness are different objectives, and generative systems are only trained on one of them.
Bentzion Goldman, Design Director at Mother Design, described the year in branding with a precision that cuts the same way. Writing for Creative Boom in January 2026, Goldman noted that what stood out across the year's identity work was not how bad it was. It was the overwhelming okayness of it: a sea of identities with no technical fault and no spiritual centre. Mid-branding, at scale, across an entire industry.
The implication is not that AI has ruined design. It is that AI, used without strong human direction, tends to produce work that occupies the same narrow aesthetic territory as everything else trained on the same visual corpus. The tool reflects the corpus. The corpus reflects the mean. The mean looks like your competitors.
What Cannes Lions 2026 revealed
The advertising industry said the quiet part out loud this year. The defining theme at Cannes Lions 2026 was not a new format, a new platform, or a new media channel. It was what the industry called AI tension: a reckoning with what happens when generative tools produce creative at scale, and the cumulative output starts to rhyme.
What actually won Grand Prix awards told a different story from the entries. The Ordinary took the Health and Wellness Grand Prix for The Periodic Fable, created with Uncommon Creative Studio in London. The campaign built an entire dystopian classroom around the fake-science vocabulary the beauty industry uses to sell products: words like medical grade, pore-less, fat freezing, and eternal youth, each one presented as a lesson in a curriculum of misinformation. The campaign required a position that made an implicit enemy of most of the industry The Ordinary competes in. That is not a brief a model would volunteer.
SKF, a 150-year-old Swedish bearings manufacturer, won the Creative B2B Grand Prix for The Faroe Islands Space Program, created with NORD Stockholm. The campaign built a literal space program in partnership with Minesto and SEV to dramatise what frictionless engineering actually means in practice. A bearings company launched a space program to win a creativity award. The brief behind that campaign had to defend itself in every single meeting before it existed.
Heinz has demonstrated the same principle over a longer time horizon. In a 2021 experiment that became the cornerstone of the brand's Creative Effectiveness Grand Prix win at Cannes 2024, people in 18 countries across five continents were asked to draw ketchup without any brand name being mentioned. Ninety-seven percent drew a Heinz bottle. That recall is not the product of AI. It is the product of decades of consistent brief-led creative that built assets designed to survive memory rather than to look good in a pitch deck.
What these three share:
Each campaign required a specific point of view that excluded people, and the willingness to defend it through development. None could have emerged from a vague brief and a generation tool. The brief was the work. The execution followed.
What wins: a point of view
The first thing that still cuts through is a point of view. Not a positioning statement. Not a set of brand values articulated in a workshop. A belief specific enough that it pushes some people away.
The Ordinary's brand platform, The Truth Should Be Ordinary, is not a tagline. It is a position that makes an implicit enemy of every brand in the beauty industry that uses pseudo-scientific language to sell products. That friction is not an accident of execution. It is the source of the brand's distinctiveness. The campaign that won at Cannes was not a one-off. It was the natural output of a brief built on a position that excludes.
A point of view earns something that volume cannot buy: the ability to be disagreed with. Work that can be disagreed with can also be remembered. Work that is merely competent, pleasant, and inoffensive cannot. In a market where AI tools can produce unlimited competent output, a point of view is the one creative input that cannot be generated. It has to be decided. And decisions require someone willing to defend them.
The practical implication for SaaS brands is uncomfortable. Most brand positioning exercises are designed to include as many potential buyers as possible. The instinct is to be relevant to everyone. But relevance to everyone is another form of sameness. In a flattened market, the brands that cut through are the ones that chose a specific belief and made everything from it.
What wins: distinctiveness
The second thing that cuts through is distinctiveness. This is not the same as differentiation, though people use the words interchangeably. Differentiation is a category claim: we do X better than the competition in a way that matters to buyers. Distinctiveness is a memory asset: you know us when you see us, even without our name present.
The Heinz experiment makes this visible. Ketchup is not a category where one product is objectively better than another in a measurable way. Heinz is distinctive because decades of consistent, brief-led creative have built a set of assets — a specific red, a specific bottle shape, a specific tone of voice — that survive in memory. The product recall tested in 18 countries was not the result of performance marketing. It was the result of assets that were built to be remembered, not just noticed.
In a world where AI can produce unlimited visual output and every team has access to the same tools, distinctive assets have become more valuable than they have ever been, for precisely that reason. The output of AI tools looks like the training data. The training data looks like everything that already exists. Distinctive assets are the things that do not look like everything that already exists, because they were built specifically to avoid it.
Building distinctive assets requires the same thing a point of view requires: a decision about what to exclude. A brand that wants to own a specific colour cannot also use the entire palette. A brand that wants to own a specific tone cannot also be warm, playful, authoritative, and irreverent. Distinctiveness is built by subtraction, and subtraction requires conviction.
What wins: taste
The third thing that cuts through is taste. Taste is the human judgment that sits between the brief and the output — the trained, defended capacity to look at a hundred generated options and identify the one that serves the idea. It is not aesthetic preference. It is creative direction.
Taste has always mattered in creative work. What has changed is its position in the process. For most of the history of brand creative, generation was the constraint. Teams worked within time and budget limits that forced early decisions and made iteration expensive. Now generation is the cheapest part of the process. A team can produce more options in an afternoon than a previous generation of designers could produce in a month.
The new constraint is the judgment required to direct that generation toward something specific, and then to select, refine, and defend the output that serves the brief. The more AI tools you add to a creative process, the more your competitive advantage shifts to the humans running them — not for their ability to make things, but for their ability to know which things are worth making.
This is a structural shift in where creative value sits. Generation was once scarce. Judgment was abundant. Now generation is abundant. Judgment is the scarce input. The teams that understand this and invest in it accordingly — in briefs, in creative direction, in the willingness to reject options until the right one exists — will produce work that stands apart from the AI-generated mean. The teams that treat AI as a shortcut to the brief will produce competent work that disappears.
TheBullseye Perspective
We work with SaaS brands that are trying to stand out in categories where everything looks correct and almost nothing looks specific. The pattern we see most consistently is not a talent problem or a tool problem. It is a brief problem. Teams reach for AI tools before they have a brief worth executing, and the tools return competent output that reflects the average of their training data. The fix is not less AI. The fix is a brief with enough conviction to direct it. When the brief is right, the tools become an advantage. When the brief is vague, the tools accelerate sameness. Generation is now free. The brief is still the work.
Sources
Bentzion Goldman, Design Director, Mother Design. 'How being weird can save branding in 2026.' Creative Boom, January 2026. https://www.creativeboom.com/insight/how-being-weird-can-save-branding-in-2026/
Hintze A, Proschinger Astrom F, Schossau J. 'Autonomous language-image generation loops converge to generic visual motifs.' Patterns, Cell Press, January 2026. https://www.cell.com/patterns/fulltext/S2666-3899(25)00299-5
The Ordinary / Uncommon Creative Studio. 'The Periodic Fable.' Cannes Lions Health and Wellness Grand Prix 2026. https://www.adweek.com/creativity/the-ordinary-wins-big-among-8-grand-prix-winners-from-cannes-lions-2026-day-1/
SKF / NORD Stockholm. 'The Faroe Islands Space Program.' Cannes Lions Creative B2B Grand Prix 2026.
Rethink Toronto / Heinz. 'It Has To Be Heinz' (Draw Ketchup). Cannes Lions Creative Effectiveness Grand Prix 2024. https://adage.com/article/special-report-cannes-lions/rethink-torontos-heinz-ketchup-work-wins-creative-effectiveness-grand-prix/2566236/
Nitya Shukla Paharia
Creative Director & Head of Brand
Leading creative & design at TheBullseye, solving for clarity-first storytelling for SaaS and AI companies. Operating at the intersection of narrative, design, and video to translate complex products into high-conversion content across GTM, product marketing, and brand systems. Focused on building design that doesn’t just look good, but drives understanding and decision-making.
FAQs
AI design sameness is the visual convergence that occurs when brands use generative AI tools to produce creative work at scale, resulting in similar aesthetics across unrelated companies and industries. A January 2026 study published in Patterns (Cell Press) by Hintze, Proschinger Astrom, and Schossau found that autonomous language-image generation loops consistently collapse toward a narrow set of generic visual motifs. For brands, this means that AI-assisted design, without strong human judgment directing it, tends to produce competent, forgettable output that looks like every other team using the same tools trained on the same data.
The Great Flattening is the convergence of brand visual identities and creative output toward a shared, undifferentiated middle ground, driven by two forces happening simultaneously: the democratisation of professional design tools and the arrival of generative AI. Both forces raise the floor for quality while pushing output toward the statistical average. The result is a market where almost everything looks technically correct and almost nothing looks distinctive. The term captures the loss of the extremes — the weird, the specific, the brave — that occurs when average becomes the default.
Generative AI systems are trained to produce plausible output, not distinctive output. They reproduce the statistical centre of their training data, which is composed of existing visual material. Research by Hintze, Proschinger Astrom, and Schossau in Patterns (2026) demonstrated this directly: when AI image systems are fed their own output on a loop across 700 trajectories, the visuals rapidly converge toward 12 dominant commercially safe motifs, regardless of the starting prompt. The systems are functioning correctly. The problem is that plausibility and distinctiveness are different objectives, and most generative systems are only optimised for one.
The defining theme at Cannes Lions 2026 was AI tension: an industry-wide reckoning with what happens when generative tools produce creative at scale and the cumulative output starts to look the same. The Grand Prix winners ran in the opposite direction: The Ordinary won the Health and Wellness Grand Prix for a campaign that satirised the beauty industry's fake-science vocabulary; SKF won the Creative B2B Grand Prix for a campaign that built a literal space program; Heinz continues to demonstrate that decades of brief-led creative produces recall no AI output can match. Each winning campaign required a specific point of view and a brief willing to defend it.
Brand distinctiveness is the quality that makes a brand recognisable and memorable without its name present — a specific colour, shape, tone, or set of assets that survive in memory rather than just looking correct in a pitch deck. It is different from differentiation, which is a category claim. In a market where AI tools have made visual competence universal and free, distinctiveness has become the primary way a brand can earn recall and stand apart. The Heinz experiment — in which 97 percent of people in 18 countries drew a Heinz bottle when asked to draw ketchup without naming a brand — illustrates what distinctiveness built over time actually looks like.
SaaS brands can avoid looking generic by treating the brief, not the tool, as the primary creative investment. Generation — producing variations, iterations, and options — is now the cheapest part of the creative process. The scarce input is the conviction required to direct that generation toward something specific enough to exclude people, and the judgment required to select and defend the output that serves the brief. Practically, this means defining a genuine brand point of view before opening a generation tool, and using the tool to execute a brief rather than to produce a brief. Brands that use AI to shortcut the brief will produce output that looks like AI shortcutting a brief.
AI does not inherently hurt brand identity, but it amplifies whatever creative direction it is given. If that direction is vague or consensus-driven, AI will produce vague, consensus-driven output at scale and speed. If the direction is specific — built on a genuine point of view and a brief that knows what it is trying to exclude — AI becomes a tool for executing distinctive creative faster than was previously possible. The problem is not the tool. The problem is the brief. A team with a strong brief and AI tools will outproduce a team with a weak brief and AI tools every time.
Visual elevator music is a term coined by researchers Hintze, Proschinger Astrom, and Schossau in a study published in Patterns (Cell Press, January 2026). It describes the generic visual output that emerges when AI image generation systems are fed their own captions on a loop: images that are pleasant, polished, and completely devoid of specificity or meaning. The researchers identified just 12 dominant visual motifs that the systems consistently converged toward across 700 trajectories, regardless of the starting prompt. The term is useful for brand teams because it names what average-optimised AI output actually looks like at scale: technically acceptable, immediately forgettable.
