The AI Agency Myth – The Copy of A Copy Of A Copy (Idea: Johannes Faupel, Execution: Google Nano Banana)

The AI Agency Myth

Part of the Beyond SEO series — on what search optimization was never designed to solve.

From Reproductive Hallucination to True Human Creativity

David Ogilvy Big Idea or the campaign is like a ship in the night.

David Ogilvy: “It takes a big idea to attract the attention of consumers and get them to buy your product. Unless your advertising contains a big idea, it will pass like a ship in the night.”

David Ogilvy: “It takes a big idea to attract the attention of consumers and get them to buy your product. Unless your advertising contains a big idea, it will pass like a ship in the night.”

That sentence was written before the internet existed.

It has not aged.

It has not aged because it is not about technology. It is not about channels, formats, reach, or frequency. It is about the one thing that has never changed in the entire history of human commerce: a person needs to feel, in the first three seconds of encountering your message, that you understand something true about them that they had not yet found words for.

That is what a big idea does.

That is the only thing that has ever worked.

What the Market Believes Right Now

The proposition being sold to marketing departments in 2026 is something like this:

[The myth: AI can generate your content, your copy, your campaigns, your social posts, your email sequences, your product descriptions, and your SEO articles — faster, cheaper, and at unlimited scale.]

This is true as a technical matter of fact.

Every word of it is true, and every word of it is beside the point.

Because the question was never: Can this be produced quickly?

The question to be asked is:

Does this contain something true enough about the reader that they stop, recognise themselves, and stay?

Speed and volume are answers to a different question. They are excellent answers to that different question. But when they are applied to the question of persuasion — of genuine contact between a message and a mind — they produce something that looks exactly like communication and functions as noise.

The online market is currently drowning in it.

3 (amongst others) Men Who Understood What AI Cannot Source

David Ogilvy wrote the Rolls-Royce headline — “At 60 miles an hour the loudest noise in this new Rolls-Royce comes from the electric clock” — after reading 26 Rolls-Royce engineering documents and understanding that one technical detail meant something to the reader that the engineers themselves had not noticed. The detail said: this object is so refined that the world’s ordinary irritations have been engineered out of it. The reader who wanted that did not want a car. They wanted a proof of arrival. The headline was not written to that desire. It was the desire, made visible.

No training corpus produces that perception. The corpus contains the headline — after the fact.

Ted Nicholas spent decades studying what he called the most important words in any piece of copy: the headline and the first sentence. Not because they were the beginning, but because they were the contract. The reader decides in the first sentence whether the writer has earned the right to continue. Nicholas understood that this contract is made not with technique but with truth — the writer must have been inside the reader’s specific fear or specific want, not a general approximation of it. His direct mail pieces achieved response rates that are still cited as benchmarks because the opening line demonstrated, without stating it, that the writer had lived adjacent to the reader’s problem.

AI has read every Nicholas headline. It can generate headlines in his register. What it cannot do is have spent the years earning the understanding that made each headline inevitable rather than constructed.

Eugene Schwartz gave us the concept of market sophistication — the understanding that a market’s awareness of its own desire determines everything about how you can speak to it. A naive market can be reached with a direct promise. A sophisticated market has heard every direct promise and stopped believing them. A saturated market requires you to speak to the desire underneath the desire — the want behind the stated want, the fear that the stated want is trying to resolve.

Schwartz’s Breakthrough Advertising remains the most structurally precise analysis of how persuasion actually works, because it was built from observation of real market behaviour over decades. It is not a style guide. It is an epistemology of desire.

AI trained on Schwartz can reproduce his vocabulary. It cannot accumulate the observational database that made his framework necessary to build in the first place.

What is The Difference Between Execution and Perception?

Perception is the mental work, the foundation. Like Andy Warhol has had the idea to print Marilyn Monroe in different colors. Execution is the technical work. Like Andy Warhol’s workers did the printings after the big idea was there.

Here is the distinction that the AI Agency debate almost never makes clearly:

Execution is the translation of a known idea into competent form. Given a clear brief, a defined audience, an established voice, and a true insight — execution is the work of rendering those inputs into output. This is work that AI does extraordinarily well. It is tireless, consistent, fast, and increasingly indistinguishable in quality from average human execution.

Perception is the prior act. It is the moment when a human being — through experience, attention, and the kind of uncomfortable proximity to another person’s reality that cannot be simulated — understands something true that was not previously understood. It is the act that produces the brief, defines the audience, establishes the voice, and identifies the insight.

Perception precedes execution. Always.

When an organisation hands its communication to an AI agency — meaning: removes the human act of perception from the process and replaces it with generated execution — it does not save money on creativity. It eliminates creativity and pays production costs for its absence.

The output looks like advertising. It moves like advertising. It occupies the same channels as advertising. And it passes, as Ogilvy said, like a ship in the night.

What Reproductive Hallucination Actually Means

The term hallucination in AI refers to the generation of plausible-sounding content that is not grounded in fact. But there is a subtler form of hallucination that receives less attention:

The generation of plausible-sounding relevance.

A language model trained on existing content learns the patterns of what has already resonated. It learns the vocabulary of connection, the structure of empathy, the rhythm of a well-placed insight. It can produce text that has the shape of genuine understanding.

What it cannot produce is genuine understanding — because genuine understanding requires a perceiving subject with something at stake, encountering a reality that resists simple pattern-matching, and staying with that resistance long enough to extract something true.

This is reproductive hallucination: content that reproduces the surface appearance of original thought, generated without the act of original thought that made the surface worth imitating in the first place.

The market is developing, slowly and collectively, an immune response to this. Not consciously. Not articulately. But the same instinct that makes a person stop on a single sentence in a book of 80,000 words — the recognition instinct, the this writer has been here response — is also the instinct that registers its absence.

Readers cannot always say why certain content feels hollow. They simply do not stay.

What Compounds and What Depreciates

Ogilvy’s Rolls-Royce headline was written in 1958.

It is still being cited, analysed, and taught in 2026. Not because it has been re-optimised. Because the perception it contains was true, and true things do not depreciate.

Nicholas’s opening lines still generate response when studied — not imitated, studied — because they encode an understanding of human motivation that has not changed.

Schwartz’s market sophistication framework is applied today to AI-era audiences with the same accuracy it had for direct mail audiences in 1966, because human desire has the same structure it always had.

The work of genuine perception compounds. It accumulates authority not through citation count but through continued accuracy. The person who encounters it years after it was created finds it as useful as the person who encountered it on the day of publication.

The benchmark article optimised for this quarter’s keyword cluster is invisible within the year. Not because it was poorly made. Because it contained nothing that was not already known.

The Human Work That AI Makes More Visible

There is an inversion available to those who understand the distinction.

When AI handles execution — competently, quickly, at scale — the human who is freed from execution is not freed from work. They are freed for the harder work that execution was obscuring: the act of perception itself.

This means:

Sitting with a customer’s actual operational reality long enough to find the true friction, not the stated friction
Building the frameworks that name what a field has not yet named — Information Embedding as the discipline of occupying semantic vacua before consensus forms
Asking the question underneath the question, the way the PAA data reveals the retirement anxiety underneath the ROI query
Doing what Ogilvy did with those 26 engineering documents: reading past the surface until one true thing becomes unmistakable

This is not faster than generating content at scale. It is categorically different work — and in a market saturated with generated execution, it is the only work that cannot be replicated by the next model update.

The organisations that understand this will not use less AI. They will use it for exactly what it does well, while protecting, with deliberate discipline, the human act of original perception that gives any AI output something true to execute from.

If someone tells you to replace your advertising agency by AI – run. Fast.

Replacing Advertising Agencies with Artificial Intelligence Agents

If you here “We replace your Advertising Agency by AI-Agents” – run. Fast.

A Note on Ogilvy, Nicholas, Schwartz — and Why They Are Not Nostalgic References

These three names are not invoked here as a lament for a golden age of advertising.

They are invoked because they were the people who most rigorously documented what actually works — not what looks like it works, not what generates activity, but what produces the specific cognitive and emotional event in a reader that results in a decision.

Their findings have not been superseded. They have been buried under the vocabulary of digital marketing, which developed its own metrics for activity and largely stopped asking whether the activity was producing the event that matters.

The event that matters is recognition. The moment a person reads something and thinks: this is for me. This person has been where I am. I am in the right place.

That event is not a product of optimisation. It is a product of understanding.

Understanding is still, in 2026, a human act. The tools available to support and distribute that understanding have never been more powerful. The understanding itself still has to come from somewhere.

It comes from the same place it always has.