CREATIVE
Why AI campaign copy all sounds the same
Not because the model is not good enough. Because nobody decided what kind of idea to have, so it reaches for the same device every time.
Ask any AI to build a campaign and you get the same shape: state the problem, state the benefit, add a testimonial, close with a call to action. The shape is reasonable. The shape is also identical for a coffee shop, a law firm, and a SaaS app.
The natural response is to blame the model and write a longer prompt. That rarely works, and the reason has nothing to do with prompting.
The decision nobody made
Writing a campaign involves two decisions, not one. First: what kind of idea are we having? Second: how do we write it well.
Almost everyone, human or machine, jumps to the second. The first decision still happens; nobody just makes it. Left unmade, it is filled by the default, and a language model's default is the most common shape in its training data. The most common shape is, by definition, the one most similar to everyone else's.
A longer prompt gives the model more context to fill the same shape with. What changes is the filling, not the shape.
FREE DOWNLOAD
Seven Angles · 10 pages · Seven creative devices, one from each family, each with the conditions it fails under. Get it →
Devices, not moods
The fix is to choose the device first. Not a mood. Not bolder or more premium, because those describe a feeling and a model cannot obey a feeling. A device is an instruction that can be followed.
Compare the two. Make it more engaging changes nothing structurally. Whereas: write the problem as a running cost they are already paying, in their own units of time or money, and do not mention the product until the cost has been stated, produces a structurally different campaign rather than a differently worded one.
The first is a preference. The second is a constraint. Only a constraint can be obeyed.
The half everyone skips
Every creative device has conditions under which it fails, and that half is almost never written down. Dramatising the problem works when the audience does not think they have one, and collapses into nagging when the problem is small. Admitting the hard part builds trust, and collapses into false modesty when the admission is secretly a boast.
A deck of creative devices that only tells you when something works will get played on the wrong hand. That is why when-it-fails deserves the same weight as when-it-works, rather than a footnote.
READ NEXT
CAMPAIGN
Nine questions to answer before any budget moves
Most campaigns do not fail on execution. They fail because a stage was never answered, and nobody noticed until the money was spent.
TOOLS
A general AI agent or a Campaign HQ: when to use which
This week we spent a night inside a general AI agent that had just raised a large round. It made a good client report fast. What it did not do is the thing that keeps an agency trusted.