"Told Rather Than Seen": Comparatively Analyzing Human-Written and AI-Generated Screenplaysá

Ruth Li 
Alfaisal University

This assignment engages students in comparatively analyzing human-written and AI-generated texts including screenplays. I incorporated this strategy while teaching an inaugural upper-level English course on communication in the age of artificial intelligence and digital media (in spring 2024 and spring 2025). By attending closely to human-written and AI-generated texts, students investigate the extent to which AI-generated texts can capture the substance and style of the original texts. More broadly, the exercise encourages students to critically interrogate the extent to which AI outputs could be construed as creative. 


Learning Goals

  • Comparatively analyze human-written and AI-generated texts including screenplays 
  • Critically investigate the extent to which AI-generated texts can capture the substance and style of human expression 
  • Reflect on the possibilities and limits of AI as a tool for creative expression 

Original Assignment Context: Upper-level course on communication in the age of artificial intelligence and digital media 

Materials Needed: Selected readings, an AI tool such as ChatGPT, PC 

Time Frame: ~2-3 week-long unit on AI and creativity

Overview: In the following, I present an approach to supporting students in comparatively analyzing human-written and AI-generated texts. I incorporated this strategy while teaching an inaugural upper-level English course on communication in the age of artificial intelligence and digital media. I have taught this course and activity twice, once in spring 2024 and once in spring 2025. The course syllabus appears below: 

New technologies are constantly making our communication tasks both easier and more complicated. Students will learn about a variety of AI-driven communication tools, including ChatGPT, social media algorithms, speech analysis, chatbots, and many other tools. Interactive class discussions as well as firsthand exploration of these tools will help students understand the incredible power as well as the challenges and limitations of using artificial intelligence and digital media for communication. Students will have opportunities to analyze AI-generated texts, to design an AI innovation, and to compose using multiple media forms, including visual, audio, and linguistic components. This course counts toward the minor in strategic communication.

As part of a unit on AI and creativity, I immerse students in interpreting the substance and style of human written and AI-produced texts including film and television screenplays. I begin by asking students to write in response to a quotation by Justin Zackham: “For me, screenwriting is all about setting characters in motion and as a writer just chasing them. The characters should tell you what they’ll do in any scene you put them in.” I ask students to consider their thoughts on this quotation, including what makes a character come alive in a movie or show. One student reflected on the role of authenticity in characterization: “What makes a character come alive in a movie or show is authenticity and development. Having characters be stagnant in their growth is uninteresting and can create distance in the text and reader, so, seeing authentic and reasonable change in their character over the course of the text is what makes them feel real.” 

To model an example analysis, I engage the class in comparatively analyzing a clip from the show “Friends” alongside a ChatGPT-generated “Friends” script. Students then work in small groups, prompting an AI tool to generate a passage written in the style of their favorite book, movie, or show. Students take notes on their observations, interrogating the extent to which the AI-generated texts capture the substance and style of the original. Together, we disentangle the intricacies of language, considering the affordances and limitations of AI as a tool for creative expression. As a follow-up assignment, students write an essay that comparatively interprets the substance and style of a human-created and AI-generated text of their choice. 

In their reflections, students note the ways the AI texts elide the delicate distinctions of meaning and expression in the original texts. For instance, one student chose to analyze an AI-generated script of the show WandaVision. The student wrote, “In the AI text, Wanda’s development was told rather than seen, unlike the actual script.” The student’s reflection constructs an interesting juxtaposition between “seeing” and “telling” wherein “telling” could be construed as statements of declarative meaning — the words as stated in the script — whereas “seeing” conveys a sense of meaning that transcends description. This distinction between “seeing” and “telling” raises a question over whether AI-generated writing could capture an “ineffable” (Schnitzler & Vee, 2025) quality of meaning beyond the words themselves.  

Such philosophical questions invite deeper explorations over whether AI-generated screenplays could capture the nuances of human expression. As one student wrote in a post-unit reflection on the extent to which AI-generated texts could be construed as creative: “AI-generated art almost feels distant from human emotions; it does not accurately represent human art, neither does it portray any form of connection and relatability. The nuance in human art stems from life itself and situational experiences; these elements are absent in art generated by AI. Drawing a conclusion to my point, AI-generated art is only creative to some extent because human art requires far more deep-seated emotions, talent, and skill than AI-generated art ever will. Despite my opinion, creativity is subjective and is not a measurable property; therefore, it will always be in the eyes of the beholder.” 

The student’s reflection stimulates meditations on the intertwined nature of creativity and subjectivity — qualities that reaffirm the value of human experience in artistic expression. In future iterations of this activity, I might contextualize the discussions within theories of experimental poetics and computational creativity such as Oulipo, following Luman’s (2024) assignment in this collection. I might also encourage students to experiment with the iterative, recursive process of composing with AI and to explore the ways prompt engineering might shape or unshape a text, stimulating a perpetual unraveling of language and form. Ultimately, I inspire students to experiment with the possibilities and limits of creative expression at the interface of the human and the artificial. 


Assignment

Activity: Comparing human-written and AI-generated texts 

In small groups (2-3 members each): 

  1. Choose a book, movie, or TV show that you are all familiar with. Select a short passage from the text (1-2 pages in length). (10 minutes) 
  2. Prompt ChatGPT or another AI tool to compose a short passage in the style of your chosen text (1-2 pages in length). (10 minutes) 

Prompting strategies:

  • Specify the text, author, and genre you are prompting the AI tool to emulate (e.g., “Write song lyrics in the style of ‘happiness’ by Taylor Swift”).
  • Consider prompting the AI tool to revise the output multiple times based on your feedback, considering aspects such as theme, style, tone, or word choice (e.g., “Convey the complexity of emotions that accompany the end of a relationship,” “Incorporate imagery relating to light and darkness,” “Imbue the lines with a more melancholy tone”).
  1. Take notes on your observations: (20 minutes)
    • Compare the human-written and AI-generated texts side by side. Take notes on what you notice about the texts, including the substance, structure, and style. 
    • Compare the AI texts across revisions. Analyze the extent to which the revised version(s) is more specific, complex, or nuanced than the original.
  2. Discuss as a class: (20 minutes) 
    • How would you compare the themes and styles of the human-written and AI-generated texts? 
    • To what extent does the AI-generated text capture the nuances and subtleties of the original? 

AI Text Analysis Essay 

In this assignment, you will comparatively analyze the content and style of a human-created text and an AI-generated text written in the style or genre of the human-created text. 

Possible genres to analyze: a poem, song lyrics, an essay, a book, a movie, a TV show, a report, a case study, a painting, a photograph, etc. 

  1. First, choose a text that you are familiar with. Select a short passage from the text (1-2 pages in length). 
  2. Next, prompt ChatGPT or another AI tool to write a short passage in the style or genre of the text (1-2 pages in length). 

Prompting strategies:

  • Specify the text, author, and genre you are asking the AI tool to emulate (e.g., “Write song lyrics in the style of ‘happiness’ by Taylor Swift”).
  • Consider prompting the AI tool to revise the output based on your feedback on the themes and style (e.g., “Convey the complexity of emotions that accompany the end of a relationship,” “Incorporate imagery relating to light and darkness”).

Compare the human-written and AI-generated texts side by side. You might consider issues such as: 

  • Substance: What ideas does each text express? How well-developed are the themes? How complex, detailed, or specific is each text? If applicable, how are the characters, scenes, or plot represented in the human-written story and in the AI-generated version?  
  • Structure: How is each text organized? How does each text progress? 
  • Style: What word choices does each text employ? How effective are the language choices in each text? 
  • Overall effectiveness: Evaluate the overall effectiveness of the AI-written text. How well does the AI-generated text convey the substance, structure, and style of the original work? To what extent does the AI text convey the nuances (the subtle shades or distinctions in meaning or expression) of the original? Are there any aspects in the original text that are not captured by the AI version, or vice versa? To what extent does the AI text offer original or inventive ideas? What suggestions, if any, would you make to improve the AI-generated text? 
  1. If you would like, you may prompt the AI tool to revise the generated text based on your comments or critiques. Analyze the extent to which the revised version(s) is more specific, complex, or nuanced than the original. 

Acknowledgements

I am indebted to my colleagues and students for their generosity and insights, which inspired this piece. 

References

Luman, D. (2024). cmpttnl cnstrnt: An Exercise in Constraint and Prompt Engineering. In C. Schnitzler, A. Vee, & T. Laquintano (Eds.), TextGenEd: Continuing Experiments. The WAC Clearinghouse. https://doi.org/10.37514/TWR-J.2024.2.1.01 

Schnitzler, C., & Vee, A. (2025). Ghostwriting: Your Voice in the Machine. In C. Schnitzler, A. Vee, & T.  Laquintano (Eds.), TextGenEd: Continuing Experiments. The WAC Clearinghouse. https://doi.org/10.37514/TWR-J.2024.2.1.01