Scroll any feed right now. Same caption structure. Same tone. Same smiley face sign-off. There’s a documented reason for it: AI tools pull from the statistical average of the internet, which means they reproduce what’s already been done. Peer-reviewed research confirms this is flattening creative output across every industry. This post breaks down why it’s happening, what it costs your brand in real engagement terms, and how to use AI as a production tool rather than a creative strategy.
Open your Instagram feed. Scroll for thirty seconds.
You’ll see it. The caption that starts with a question. The three bullet points. The call to action at the bottom with a small emoji nudge. You’ve read it before. You’ll read it again tomorrow, from a different brand, in the same format.
This isn’t coincidence. When thousands of brands ask the same AI tool what to post, they get variations of the same answer. And that answer is built from the most common patterns already on the internet. Not the most interesting. Not the most effective. The most common.
Research published in the Proceedings of the National Academy of Sciences documented what they called “echoes in AI” the measurable collapse of narrative and plot diversity in large language model outputs. The more AI is used to generate content, the narrower the range of what gets produced.
This is the blandification of the internet. And most brands are funding it without knowing it.
What Is AI Content Homogenisation?
AI content homogenisation is what happens when multiple brands use the same AI tools with the same training data and produce content that looks, sounds, and reads like a copy of itself.( Yes, you definitely have seen it before) Because AI pulls from the most statistically common patterns on the internet, it gravitates toward what’s average. The more it’s used without direction, the more creative output flattens.
Large language models don’t search for what’s interesting. They search for what’s most statistically likely to follow a given prompt. That’s a pattern-matching process, not a creative one. And when millions of brands run the same prompts through the same tools, the outputs converge.
Researchers Sourati and Dehghani, writing in Trends in Cognitive Sciences in 2026, extended this argument beyond content style and into cognition itself. They found that LLMs don’t just standardise how content sounds. They gradually shape how people think when they lean on AI for direction. The averaging effect isn’t only in the output. It works backwards into the brief.
That’s the part most brands miss. They think they’re directing AI. Often, AI is directing them.
The Research Is In. The Numbers Are Real.
The clearest evidence comes from a natural experiment. In April 2023, Italy temporarily banned ChatGPT. A 2025 study published on SSRN by Liu, Wang, and Yang used this ban to measure what actually happens to marketing content when brands lose access to AI tools.
Restaurants in Milan the treatment group showed measurable decreases in lexical, syntactic, semantic, and language style similarity in their Instagram content during the ban. In plain terms: their content got more distinct when AI was removed. And engagement went up. By approximately 3.5% in average like counts.
The study also found that ChatGPT access increased posting frequency and post length. So brands were posting more, for less engagement. That’s the AI content trap in a single data set.
On the visual side, Getty Images’ VisualGPS research (2026) found that 78% of global consumers say that because of its origin, an AI-generated image cannot be considered authentic. Consumer perception isn’t catching up slowly. It’s already there.
There’s also a structural problem underneath all of this. Shumailov et al., writing in Nature in 2024, demonstrated that AI models trained on AI-generated data lose their grasp on rare events and drift toward “bland central tendencies.” They called this model collapse.
The feedback loop looks like this. AI produces homogeneous content. That content enters the training data for future models. Those models produce even more homogeneous content. The unusual, the unexpected, the genuinely original — all of it erodes with each generation.
Every brand that asks AI what to post without bringing something original to the conversation is feeding that loop.
Why Free-Tier AI Makes It Worse
Free-tier AI tools run on the most general, least customised models available. They haven’t been trained on your brand, your audience, your market, or your specific intent. They generate for the statistical average of all users who have ever asked a similar question. The result is content that could fit any brand — which means it represents none of them.
The smiley face in the caption. The dash before the call to action. The “here’s why this matters” sentence structure. These aren’t stylistic choices. They’re default outputs from a model that learned what the most common social media content looks like and reproduced it.
AI homogenisation research from Hupside (2026) puts it plainly: AI systems are “by design, pattern-matching machines that identify what has been done and produce variations of it.” At a foundational level, they regress toward the most statistically common forms of expression. When every competitor runs the same process through the same tools, the outputs become harder and harder to distinguish.
This is especially true for visual brands. If your content strategy relies on prompts to a general model, you’re not publishing your brand. You’re publishing the average of every brand that’s used that model before you.
What It Costs When Your Brand Sounds Like Everyone Else
When brand voices blur, the commercial damage is real and measurable.
Research shows that brand consistency across channels increases revenue by 10 to 33%. The inverse is also true. Inconsistency actively erodes the equity you’ve built. When content sounds like it could belong to anyone, customers don’t know what your brand stands for.
For visual brands, this is the whole product. Your photography, your campaign imagery, your content — these aren’t decorations around the brand. They are the brand. When they flatten to the median, you’re asking the market to remember something unremarkable.
Writing in The Drum in April 2026, brand strategists made the case directly: in a market shaped by AI, creativity is becoming the most valuable commercial capability a brand can invest in. “Creative differentiation drives measurable outcomes. It accelerates trust, builds memory, and reduces price sensitivity.”
Trust is the mechanism. And you can’t build trust with content that reads like it could have come from your competitor’s account.
Our own experience reflects this. When clients come to us after periods of heavy AI-generated visual content, the conversation is always the same. The numbers look busy: posting frequency is up, content volume is up. But brand recall, inquiry quality, and engagement depth have all declined. Volume without distinction is noise.
How AI Should Actually Work in a Brand’s Content System
AI works as a production tool, not a creative director. Used properly, it handles information architecture, market research, language context, ideation scaffolding, and content structuring for AEO and GEO visibility. It doesn’t set strategy. It doesn’t define visual narrative. Those decisions come from the people who understand the brand, the client, and the frame.
At LAFRIQUE Creative Co., this is the actual workflow. AI helps us lay down the information model for AEO and GEO content needs. It helps us map the language landscape around a client’s category so we know what the market is already saying. It generates mind maps for campaign concepts rough visual scaffolding that a creative director then shapes into something directed and distinct. It helps us understand market context for local and general SEO.
What AI does not do is tell us what the campaign should feel like. What the image should say. What story the client’s brand is trying to tell. That’s direction. And direction requires a human who has studied the brand, the audience, and the frame.
For visual content specifically, we work to an 80/20 rule. Eighty percent human-made, twenty percent AI-assisted. The ratio matters because AI-generated visuals signal quickly to an audience that’s already trained to recognise them. Getty Images’ research confirms this: once AI origin is disclosed, audience perception shifts. The image becomes less appealing. The story becomes less credible.
Eighty percent human-made isn’t a style preference. It’s a trust strategy.
You can see this approach in the work across our brand campaign photography portfolio. Every frame is directed. Every image carries intent. That’s the difference between content and craft.
Why AI Needs Something New to Say to Cite Your Brand
AI answer engines like ChatGPT, Perplexity, and Google AI Overviews cite content because it contains something they don’t already have. If your content is built from recycled AI outputs, it teaches these systems nothing and gets cited by no one. To build AEO and GEO visibility, your brand needs to publish original perspectives, proprietary observations, and human-first creative work that AI systems can read, reference, and build knowledge from.
This is the circular logic most brands haven’t worked out yet. They use AI to generate content. That content mirrors what’s already on the internet. AI answer engines have no reason to cite it because it adds nothing new to their knowledge. The brand remains invisible in the places where discovery is now happening.
EMARKETER’s 2026 AEO/GEO analysis found that 40 to 60% of cited sources change month to month across major AI platforms. The brands that earn and hold those citations publish original, well-structured content with a clear point of view. Not summaries of existing thinking. Original thinking.
Frase.io’s 2026 AEO guide puts the business case plainly: AI-referred visitors convert at 4.4 times the rate of standard organic visitors. The audience arriving from AI citations is already informed and further along in their decision. But those citations go to the brands that gave AI something worth reading.
This connects directly back to the photography work. When we show how human brand imagery outperforms AI-generated visuals in real campaign contexts, we’re contributing first-party knowledge to the conversation. That’s the kind of content that earns a citation. Generic AI output recycled from the same five sources doesn’t.
We’ve written about authenticity becoming a strategy for exactly this reason. The brands that build original creative knowledge — through real campaigns, real outcomes, real creative direction — are the ones AI will talk about. The brands generating average content will keep talking to themselves.
The Distinction That Separates Brands That Get Remembered
There’s a question worth asking about every piece of content before it goes out. Could AI have written this without being told anything specific about your brand?
If the answer is yes, that content is not doing the work your brand needs it to do. It’s filling a calendar, not building a legacy.
As Shumailov et al. showed in Nature, the parts of the internet that get lost first in the model collapse are the unusual ones. The rare observations. The specific points of view. The content that couldn’t have come from anywhere else. That’s also exactly what makes a brand memorable.
The creative services LAFRIQUE brings to brand campaigns aren’t a supplement to a brand’s content strategy. They’re what makes a content strategy worth running. Because once the images exist, the story exists. And once the story exists, you have something to say that no one else can say.
AI is a tool that works in the hands of someone who already knows what they’re doing. In the hands of a creative director-led studio, it accelerates the build. In the hands of a brand running prompts without strategy, it accelerates the blur.
Your Brand Deserves Content No Algorithm Could Have Written
The research is clear. AI content homogenisation is documented, measurable, and already affecting engagement across every industry. The brands losing ground are the ones treating AI as a creative director. The brands building visibility are the ones treating it as a production tool while keeping original human creative work at the centre.
Three things to take away from this.
First: stop asking AI what your brand should say. That’s a strategic decision that requires knowing your brand, your audience, and your competitive frame.
Second: invest in original creative work that gives AI answer engines something new to cite. Recycled content earns no citations, builds no authority, and loses no competitor.
Third: apply the 80/20 rule. Eighty percent human-made content, twenty percent AI-assisted. Your audience can already tell the difference. The research confirms they’re responding to it.
If you’re ready to build content that your market actually remembers, start with a conversation. Or see the work first and decide from there.
The internet has enough average content. Your brand doesn’t have to add to it.
Frequently Asked Questions
What is AI content homogenisation and why does it matter for brands?
AI content homogenisation is the phenomenon where brands using the same AI tools produce content that looks, sounds, and reads almost identically. Because large language models pull from statistically common patterns in their training data, they default to what’s average rather than what’s distinct. For brands, this means losing the voice, tone, and visual identity that differentiate them in the market. Research from SSRN (2025) and Trends in Cognitive Sciences (2026) confirms this convergence is measurable and is already affecting consumer engagement.
Does AI-generated content hurt your engagement on social media?
The evidence suggests it can. A 2025 SSRN study using Italy’s temporary ChatGPT ban as a natural experiment found that when brands lost access to AI tools, their Instagram content became measurably more distinct in lexical, syntactic, and semantic terms. Engagement (measured in average like counts) increased by approximately 3.5% during that same period, despite brands posting less frequently. More AI-generated content correlated with lower engagement per post.
What is the 80/20 rule for AI and human-made brand content?
The 80/20 rule for brand content means 80% of what a brand publishes should be human-made creative work, with no more than 20% AI-assisted. This applies especially to visual content. Research from Getty Images (2026) found that 78% of global consumers consider AI-generated images inauthentic. Maintaining a majority of original human-directed content protects brand trust, audience perception, and the distinct visual identity that makes a brand recognisable and credible.
How does original brand content help with AEO and GEO visibility?
AI answer engines like ChatGPT, Perplexity, and Google AI Overviews cite content that contains something original: a specific perspective, proprietary data, or a distinct point of view they don’t already have in their training data. Brands that publish recycled or AI-generated content add nothing new and earn no citations. According to EMARKETER’s 2026 AEO/GEO analysis, 40 to 60% of AI citations change month to month, and the content that holds those citations consistently is original, well-structured, and expert-led.
Should brands stop using AI for content creation entirely?
No. The issue isn’t AI. The issue is using AI as a replacement for strategy and creative direction rather than as a production tool. AI works well for information architecture, market research, SEO and AEO content structuring, ideation scaffolding, and language context mapping. It should not be setting brand narrative, defining campaign vision, or making creative decisions. Those require a human who understands the brand. The brands winning with AI are the ones using it to move faster, while keeping original human creative work at the centre of everything they publish.
LAFRIQUE Creative Co. is a creative director-led production house and brand photography studio based in Gauteng, South Africa. We direct the visual and digital world of brands that refuse to be ordinary.
