The AI script clichés audiences recognize in seconds
Research points to the typical marks of AI generated writing. See how they show up in film scripts and how you can avoid them.
In 30 seconds
- Language models produce statistically likely text, that is, common language.
- Experienced readers recognize the so called AI vocabulary.
- Lists of three and vague conclusions also give away the origin of the text.
- Clichés are fought with specificity: real people, places and details.
Every scriptwriter knows the cliché: the ready made line, the scene the audience sees coming, the tagline that would work for any brand. Generative artificial intelligence brought a new family of clichés, hard to notice because the text comes out correct and fluent.
In this fourth article of the series, we bring together what research points out about the typical marks of AI generated writing and show how they appear in commercial scripts. We want you to arrive at approval time with a trained eye.
Why AI writes what everyone has already read
Language models learn from huge volumes of text and generate, at each step, the words most likely to come next. Mor Naaman, a professor at Cornell Tech, summed up this logic by saying that these systems produce statistically likely text, that is, common language.
In a 30 second film, common language works against your brand, because the audience recognizes the formula and tunes out before the tagline.
There is this risk of falling into the cliché of AI scripts and becoming a laughingstock instead of truly connecting with the audience.
The vocabulary that gives away the origin
The study by Jenna Russell, Marzena Karpinska and Mohit Iyyer, published in 2025, showed that readers who use AI often identify generated texts with very high accuracy. One of the clues they used most was the so called "AI vocabulary", made up of words that appear unusually often in these texts.
The analysis by Dmitry Kobak and colleagues of more than 15 million scientific abstracts arrived at a similar list, with English terms such as "delves", "showcasing", "underscores" and "crucial". In a script in any language, look with the same care at grandiose adjectives, showy verbs and abstract nouns that take up space in the sentence without saying anything specific about your brand.
Structural clichés
The clues include the way sentences are built. The experienced readers in the 2025 study also assessed formality, originality and clarity. When you read an advertising script, some structural patterns deserve extra attention:
- lists in groups of three, used out of habit, where the third idea adds nothing;
- generic openings about today's world, which could open a film from any industry;
- catchphrases built on opposition, which sound deep and repeat from campaign to campaign;
- vague conclusions, which talk about the future, connection and purpose without showing anything concrete;
- characters with no name, no place and no detail, who stand for an average of people.
These devices weigh down the script when they pile up and take the place of real observation of the audience and the brand.
Image clichés in generated films
In films produced with video tools, the cliché also shows up in the image. The cases in the third article of the series show how the audience reacted to faces, movements and settings that looked artificial. TechRadar's coverage of Coca-Cola's 2025 Christmas commercial pointed out visual inconsistencies between scenes and defined the piece as an imitation of nostalgia.
When script and image are born from the same statistical average, the film runs the risk of looking like a collage of familiar references. That is where the ironic comment finds room on social media.
How to swap the cliché for the specific
Clichés are fought with specificity. A strong script is born from concrete questions: who is the person who will watch, what they are living through in the week the film premieres, what detail only your brand has, what image or line will make someone recognize their own life on screen.
Here at Mentoniano, we hold that this work depends on research, listening and repertoire, and on someone willing to cut the pretty sentence that says nothing. In this polishing, the film gets the chance to move people, and the emotion shows up in the body of the person watching.
In practice
- Underline the grandiose words in the script and ask what each one says concretely about the brand.
- Count the lists in groups of three and keep only the ones that add information.
- Replace generic characters with people who have a name, a place and one true detail.
- Swap the vague tagline for a line only your brand could say.
References
- Cornell Bowers CIS. Human heuristics for AI-generated language are flawed (study published in PNAS), Mar. 2023. Access
- Russell, J.; Karpinska, M.; Iyyer, M. People who frequently use ChatGPT for writing tasks are accurate and robust detectors of AI-generated text. arXiv, 2025. Access
- Kobak, D. et al. Delving into LLM-assisted writing in biomedical publications through excess vocabulary. arXiv. Access
- TechRadar. Coca-Cola faces backlash again for missing the spirit of the holidays with another AI-slop ad, Nov. 4, 2025. Access
let's talk
If your brand's script sounds just like all the others, talk to Léo.
Talk to Léo on WhatsApp



