AI Content: If "Nobody" Creates It, Does It "Mean" Anything?

Here's the thought experiment that keeps surfacing in content marketing circles, usually late in a conversation after someone has had enough coffee to say it out loud: if an AI generates content that no human truly authored, and another AI summarizes it for a reader who never finishes it, did anything get communicated? Was anything actually said?

It sounds philosophical. It is philosophical. And marketing leaders who dismiss it as such are missing the most practical question in their industry right now.

Key Points

• Researchers have a name for what happens when AI trains on AI-generated data: model collapse. The distinctive, the specific, the unexpected — all of it smooths out into a statistical average. The same thing is happening to brand voice.
• A 2025 Journal of Business Research study, based on seven experiments, found that AI-generated emotional content elicits measurable moral disgust in consumers — not annoyance, not skepticism. Disgust. The kind that reduces loyalty and kills word of mouth.
• 52% of new English-language articles on the web are now AI-generated, up from 5% before ChatGPT launched. When everyone is producing the same kind of nothing, the brands that are saying something real become disproportionately visible.
• Consumers aren't taking AI content less seriously because it's low quality. They're taking it less seriously because they sense no one was accountable for it.
• The question isn't whether to use AI. It's whether there's a human signal for AI to amplify — and what happens to your brand when there isn't.

We Built a Very Expensive Game of Telephone

The mechanics are worth understanding clearly. Large language models are trained on human-generated text. They produce output. That output gets published, indexed, and scraped. It feeds the next round of training data. At a sufficient scale, the internet becomes a hall of mirrors — models increasingly reflecting models, the original human insight that sparked any of it diluted to the point of homeopathy.

This isn't speculation anymore. Research published in Nature in 2024 found that indiscriminate use of model-generated content in AI training causes irreversible defects, specifically that the tails of the original content distribution disappear. The tails are the interesting part. The rare observations. The unexpected angles. The friction and specificity that make human writing useful. When AI models are repeatedly fed their own output, rare patterns disappear, and the system drifts toward bland averages. 

Researchers call this model collapse. It's a technical problem with a deeply human corollary: the same thing happens to brand voice when AI operates without genuine human direction. The distinction gets smoothed out. The specific becomes generic. What remains is content-shaped — structurally correct, topically relevant, and hollow at the center.

As of November 2024, 50.3% of new web articles were generated primarily by AI, up from just 5% before ChatGPT launched. We have, in roughly two years, flooded the information environment with the statistical average of what content in every category tends to sound like. The floor is everywhere. The ceiling is increasingly hard to find. e

What Consumers Are Actually Feeling

Here's the part that should concern marketing leaders more than any SEO metric: consumers aren't just noticing AI content. They're having an emotional response to it that the research describes in unexpectedly strong terms.

A 2025 study published in the Journal of Business Research ran seven preregistered experiments on how consumers respond to AI-authored versus human-authored marketing communications. The findings showed that AI-generated emotional marketing communications are perceived as less authentic, leading to higher moral disgust and decreased positive word of mouth and loyalty among consumers.

Moral disgust. Not mild skepticism. Not reduced engagement. The same psychological category researchers use for reactions to ethical violations.

When participants believed that most emotional marketing communications were written by AI, they expressed disgust. The reverse was true when they believed most communications were written by a human. The researchers named this the "AI-authorship effect" — and it showed up consistently across different emotions, different business types, different communication formats. 

Why disgust specifically? Because of what emotional content implies. When a brand sends you something meant to inspire, congratulate, empathize, or connect — there's an implicit contract. Someone felt something and wanted to communicate it. Research demonstrates that when consumers believe emotional content has been generated by AI, they experience disgust that comes from a fundamental expectation: emotional content should originate from emotional beings. Discovering that no one was actually behind the feeling reads, neurologically, as a kind of deception. 

This intersection — of consumer perception, AI-generated content, and authenticity — was significant enough that Merriam-Webster selected "authenticity" as their 2023 word of the year, citing consumers' concerns about generative AI as a driving factor. That's not a marketing trend. That's a cultural signal. 

The Meaning Question

Roland Barthes declared "the death of the author" in 1967 — arguing that meaning lives in the reader, not the writer. It's a useful frame, but one Barthes didn't anticipate would need to be extended this far. He assumed there was a writer to theorize about. He wasn't imagining a statistical process that generates plausible text with no intent, no stake, no lived moment informing the choice of those words in that order.

Meaning in content — the kind that moves people, earns trust, gets shared, accumulates into brand equity — tends to require things AI cannot originate:

Someone had to decide this argument was worth making. Someone had to care whether you believed it. Someone had to exist in a specific cultural moment and bring that context into the text. These aren't decorative qualities. They're structural. They're what make a piece of writing feel like it was addressed to you rather than being directed at you.

When those elements are absent, what's left are content-shaped objects. They have paragraphs, transitions, and calls to action. They pass a surface-level read. But readers — the ones worth reaching — feel the absence even when they can't name it. The flatness. The sense that no one was home when this was written.

As one content strategist put it: "AI content is good for generating traffic but bad at building trust — it's like reading a Wikipedia page. Even if you solve the reader's problem, they won't remember you." Traffic without memory isn't brand building. It's noise in a room that's already too loud.

The Brand Equity Slow Bleed

The damage playing out right now won't show up in quarterly content metrics. Brand voice is accumulated trust — readers learn a brand's perspective, its sense of humor, its intellectual habits over time. That recognition is worth something real. It's the difference between content that gets shared and content that gets scrolled past.

Pure AI content regresses toward the mean by design. Structurally, it is the average of what has come before in its category. Average brand voice is an indistinguishable brand voice. And indistinguishable brand voice compounds, over time, into brand irrelevance — not dramatically, not all at once, but steadily enough that by the time your metrics catch it, a lot of equity has already drained.

86% of top-ranking Google pages are still human-authored. Only 14% of top results are AI-generated. That gap matters less as a tactical SEO fact and more as a signal about what search systems — which are essentially mirrors of human attention patterns — still reward. 

The brands winning this period are not the ones producing the most content. They're the ones whose content has a point of view that couldn't have been generated by a prompt.

What This Requires

The resolution isn't to abandon AI tools. It's to be precise about where AI creates value and where it destroys it.

AI as infrastructure: outlines, research summaries, distribution logistics, volume at the edges of a human-directed program. AI as identity: having a language model determine your angles, your voice, your arguments without meaningful human editorial direction. The second is where the meaning drains out.

The research is instructive here, too — messages that were edited but not written by AI were significantly less penalized on authenticity than those fully authored by AI. The human signal doesn't have to be the only signal. It has to be present and in charge. 

The most honest test for any content program: if you removed the human editorial direction, would the output be different? If the answer is no — if the AI is operating without a genuine human signal to amplify — that's not a content strategy. It's a content simulation.

And the data suggests readers know the difference. They may not be able to explain it. They may not even consciously register it. But it shows up in their behavior — in reduced loyalty, in lower word of mouth, in the particular emotional response that researchers, in a finding that deserves to sit with marketing leaders for a while, decided to call disgust.

Frequently Asked Questions About AI Content and Meaning

Here's some more info.

Do consumers really respond differently to AI-generated content?

Yes, and the research is more striking than most marketing leaders expect. A 2025 Journal of Business Research study found that AI-generated emotional marketing communications produce measurable moral disgust — not just reduced engagement — as well as decreased word-of-mouth and brand loyalty. The effect is strongest for emotional content; factual content is less penalized.

How much of the web is now AI-generated?

As of late 2024, approximately 50-52% of new English-language articles on the web are primarily AI-generated, according to an analysis of 65,000 URLs by SEO firm Graphite. Before ChatGPT's public launch in November 2022, that figure was roughly 5%.

What is model collapse and why does it matter for marketers?

Model collapse is a documented phenomenon in which AI models trained recursively on AI-generated data experience degradation—rare, distinctive, and specific content patterns are smoothed out into statistical averages. The brand voice parallel is direct: AI content produced without genuine human direction regresses toward the mean, sounding like everyone and differentiating from no one.

What's the practical difference between AI as infrastructure and AI as identity?

AI as infrastructure means using AI to accelerate, scale, and support content that a human has already shaped editorially — the direction, voice, and argument are human-originated, AI-assisted. AI as identity means letting AI determine those things without meaningful human input. The first amplifies a human signal. The second replaces it, and that's where the trust cost accumulates.

At Winsome Marketing, we build content programs where the human signal leads and AI accelerates it — not the other way around. If that distinction matters to your brand, let's talk.