Can AI Write Long Copy? I Tested It.
The results were instructive, though not in the way the enthusiasts would hope.
I decided to conduct an experiment. The question was simple, and I wanted an honest answer, unburdened by the prejudices that an old copywriter inevitably carries. The question was this: can AI write long-form advertising copy — the kind of copy that builds an argument, sustains attention over many paragraphs, and persuades through accumulated evidence and rhetoric? I had my suspicions, but suspicions are not evidence, and I wanted evidence.
The methodology was straightforward. I selected three briefs from campaigns I had worked on during my career — campaigns for which I still had the original copy, the original briefs, and in two cases the original results data. I gave each brief to ChatGPT and to one of the dedicated AI copywriting tools. I asked each tool to produce long-form body copy of approximately eight hundred words, following the brief as closely as possible. I then compared the AI output against the human copy that had actually run, evaluating both on the criteria I have used throughout my career: clarity, persuasiveness, specificity, rhythm, and what I can only call conviction — the sense that the writer believes what they are saying.
I did not tell the tools that this was a test. I did not prime them with my own copy or give them examples to imitate. I gave them the brief, the product information, and the target audience, and I let them work. I wanted to see what they would produce when left to their own devices, because that is how they will be used in practice — not as aids to experienced copywriters, but as replacements for them.
The First Brief: A Financial Product
The first brief was for a retirement investment product. The target audience was men and women in their fifties, approaching retirement with insufficient savings. The proposition was that this product offered a straightforward, low-cost way to build a supplementary retirement income. The tone was to be authoritative, reassuring, and direct.
The AI produced copy that was competent. It opened with a general statement about the challenge of retirement planning. It listed the product's features. It addressed common objections. It closed with a call to action. The grammar was correct. The tone was appropriate. The copy could have been published without embarrassment.
But it could not have been published with pride. The copy was general where it needed to be specific. It spoke of "financial security" and "peace of mind" — phrases so worn from overuse that they no longer communicate anything. It did not include a single concrete number, a single specific example, a single detail that would make the reader think "this person understands my situation." It was, in the truest sense of the word, generic. It could have been written for any financial product by any writer in any decade.
The human copy that had actually run — my copy, I should acknowledge — opened with a specific calculation. It told the reader exactly how much money a fifty-five-year-old with a specific salary and a specific existing pension could accumulate over ten years using this product. It named the amount. It showed the maths. It was not exciting writing. It was not poetic writing. But it was specific writing, and the specificity is what made it persuasive. The reader could see themselves in the numbers. They could not see themselves in "financial security."
The Second Brief: A Consumer Product
The second brief was for a kitchen appliance — a high-end food processor that cost significantly more than its competitors. The challenge was to justify the premium. The target audience was serious home cooks who already owned a cheaper model and needed to be convinced that the upgrade was worth the investment.
Here the AI performed better, because the product had tangible features that could be listed and compared. The AI produced a competent feature comparison, described the build quality, and made reasonable claims about durability. It was informative and organised. It read like a decent product review.
But it did not read like advertising. It did not persuade. The difference between a product review and advertising copy is that a product review informs the reader and lets them decide. Advertising copy guides the reader toward a decision. It anticipates their objections, addresses their hesitations, and builds, paragraph by paragraph, toward the moment when the reader thinks: yes, I want this. The AI copy did not do this because it did not understand the reader's psychology. It did not know that a person who already owns a food processor and is considering spending three times as much on a new one is not looking for information. They are looking for permission. They are looking for someone to tell them that the indulgence is justified, that the money is well spent, that they deserve a better tool. The human copy understood this and spoke to it directly. The AI copy listed features.
The Third Brief: A Luxury Brand
The third brief was the most revealing. It was for a luxury watch — not one of the Swiss giants, but a smaller brand with a distinctive heritage and a particular story to tell. The brief asked for copy that communicated the brand's history, its craftsmanship, and the emotional significance of owning a watch that was made, by hand, in a specific workshop in a specific town by specific people.
The AI produced prose that was, on the surface, beautiful. It used words like "heritage" and "craftsmanship" and "timeless." It constructed flowing sentences about the art of watchmaking. It was polished, elegant, and utterly empty. It sounded like every piece of luxury copy ever written, because it was, in a statistical sense, the average of every piece of luxury copy ever written. It was the platonic ideal of generic luxury prose.
The human copy told a story. It described a specific watchmaker — his name, his bench, the tools he used, the light that came through the window of his workshop in the afternoon. It described the specific movement of the specific watch, the number of components, the hours required to assemble them. It was not more eloquent than the AI copy. It was more true. And truth, in advertising, is the ultimate luxury, because it cannot be faked and it cannot be mass-produced.
The Verdict
My experiment confirmed what I suspected but needed to verify: AI can write long copy, in the sense that it can produce many words that are arranged in paragraphs and that address the subject of the brief. But it cannot write long copy that works — copy that builds an argument, that sustains the reader's interest, that persuades through the accumulation of specific, truthful, psychologically astute detail.
The reason is structural, not accidental. Long copy works because the writer understands the reader. Not the reader in the abstract — not the demographic, not the psychographic profile, not the audience segment. The reader as a person, with specific fears and desires and objections and aspirations. This understanding comes from experience, from empathy, from the irreplaceable act of being a human who has spoken to other humans and who knows, intuitively, what will move them. A large language model does not have this understanding. It has a statistical approximation of this understanding, which is good enough for short copy — a headline, a tagline, a social media post — but which collapses under the sustained weight of long-form argument.
Long copy is the most demanding form of advertising writing, and it is the form that most clearly reveals the difference between human craft and machine fluency. The machine is fluent. The human is persuasive. In advertising, persuasion is the only metric that matters.
Can AI write long copy? It can produce it. That is not the same thing.