You scroll past a post and think: any company could have published this. The copy is fine, the image is neat, and nothing sticks. That feeling is not a matter of taste. There are recognisable patterns behind it, and those patterns appear in predictable ways.
How to spot it in text
Generic AI copy has a set of tics you cannot unsee once you know them. The opening that makes a grand statement about the era we live in. The rule of three, where everything arrives in tidy triplets because that sounds good. The fake contrast, of the type: it is not about X, it is about Y. Sentences of suspiciously even length. And adjectives that carry no weight, like powerful, effortless or next generation.
There is also a substantive tell that matters more than style: not one number, not one name, not one situation you could not have invented on the spot. Good brand copy contains things only you could know. How long your delivery actually takes. What customers ask on the phone. Why a particular product choice was made.
Watch for structure as well. Generic copy tends to answer a question nobody asked, then close with an encouraging line that commits to nothing. Real copy usually starts in the middle of something and ends with a thing the reader can do.
How to spot it in image, audio and video
In images it shows up as gloss. Too symmetrical, too clean, light arriving from every direction at once, people looking slightly too pleased. Colours that sit close to your palette without landing on it, which is often more jarring than a colour that is obviously wrong. Typefaces that resemble your brand font without being it. And compositions where the logo looks pasted on instead of the image being built around it.
In audio and video the tell is rhythm. A voice at an even volume throughout, with no breath, no hesitation, no stress that means anything. Video where every shot runs exactly the same length. Music that never responds to what is happening. It sounds like someone reading a script rather than someone telling you something.
Where it comes from
Generic output is almost never a model problem. It is an input problem. A language or image model gives you the average of everything it knows unless you steer it clearly somewhere else. If your request is general, the answer is general. The most common causes:
- The system has never seen real brand material, only a short description of the desired tone.
- Nothing good was supplied. Without examples of what you approve, nobody knows what good means.
- The request contains no facts, only a topic and a channel.
- Everyone on the team uses the same tool the same way, so everything converges.
- What you approve never travels back into the system, so nothing is learned.
How to prevent it
Feed the system your actual material, not your description of it. Your brand book, logo files, colour values, typefaces, and above all twenty to thirty pieces of content you are proud of. That last part is the strongest steer available, stronger than any tone of voice document. A system that is retrained on your brand at every generation and carries forward what you approved before does not drift back to the average.
Then put a hard check on top. Colour values, typefaces, logo usage and banned words are all objectively testable. Anything you can fix in advance does not belong in a review round. That leaves your attention free for the question that actually matters: is this a good idea.
Two more habits help. Put something specific in every request: a number, a customer question, an internal detail, a place. And vary who asks. If one person writes every request, the output inherits that person's defaults, and after a few months everything sounds like one voice on a slow day.
Run a short test this week. Put five recent pieces side by side and mentally remove the logo. Can you still tell they are yours? If not, you know exactly where to start.
Further reading
Want to see what Content-Agent makes for your brand?
We load your brand rules and styling once. After that you make images, video, audio and text in minutes, always on brand.