Crafting the President's Words: How ChatGPT Measures Up to French Presidents' Speeches

Crafting the President's Words: How ChatGPT Measures Up to French Presidents' Speeches

In our tech-savvy world, AI's role in shaping communication is growing daily. One intriguing area where AI is making its mark is speechwriting, traditionally a human endeavor filled with nuance and emotional appeal. Can a machine like ChatGPT fill this role effectively? Researchers Dominique Labbé, Cyril Labbé, and Jacques Savoy have delved into this question by comparing the linguistic style of ChatGPT-generated speeches with those delivered by recent French presidents. Let's dive into their fascinating findings to see how artificial intelligence measures up to human eloquence.

How Does ChatGPT Compare?

The research attempts to decode the written style of ChatGPT by holding it up against the polished speeches of French presidents like Jacque Chirac, Nicolas Sarkozy, François Hollande, and Emmanuel Macron. The high-wire test? Generate end-of-the-year addresses, a staple event in any president's calendar, and see how they stack up against the real deal.

Cutting Through the Jargon

At its heart, ChatGPT, like a magic trick, pulls off written communication through a flipbook of possibilities, using massive amounts of data to predict what comes next in a sentence. This AI isn't just stringing words together; it's trying to mimic the flair and nuance of presidential speechcraft. Yet, ChatGPT shows some tell-tale signs in its writing that are distinct from human-created works.

Some Observations

Fine-Tuning the Speech: ChatGPT seems to have a penchant for nouns and numbers, but takes a backseat when using lively verbs, pronouns, and adverbs. This results in speeches that might sound more factual and structured—like a well-organized closet but lacking the spontaneous flair of human passion.

A Style with Limits: When given a sample text, ChatGPT closely mirrors the given style, but as numerous tests reveal, the magic dissipates when it's left to create independently. It churns out more universal and standardized sentences, differing in stylistic nuances from the presidents' speeches.

Parsing Language Like a Pro

Engaging with ChatGPT is like conversing with your most knowledgeable friend who, while not particularly creative, has a vast reservoir of knowledge. Yet, when tasked with speechwriting, this friend tends to play it safe, opting less for the flair in French (pun intended) and more for uniformity.

The Syntax Breakdown

  • Posessing Possessives: ChatGPT frequently uses tools like possessive determiners. Forsaking the juicy action-packed verbs, ChatGPT often describes what "we" must do, rather than plunging us into the action.
  • Numbers and Data: ChatGPT likes numbers, a reflection perhaps of its data-driven nature. Presidents focus on data to back their speeches, but ChatGPT seems to elevate numbers to center stage, reflecting a slight tilt toward technicality over emotion.

Sentence Construction: A Lengthy Discussion

One of the most distinctive features in the study is ChatGPT's tendency to stick with medium-length, average sentences. Presidential speeches, filled with naturally varying sentence lengths, lend a rhythm that holds the listener's attention. In contrast, the AI's middling sentence structure can at times feel like reading the nutritional information on a cereal box.

Real-World Implications

In the grander scheme of things, while ChatGPT's output might not yet oust the human touch in presidential speeches, it shows promise in areas where clarity, consistency, and speed are paramount. For environments that value clarity over creativity—think policy briefs or technical documentation—ChatGPT could play a beneficial role.

Key Takeaways

  1. Strength and Weakness in Style: While ChatGPT can mimic the structure of a speech when provided with examples, it struggles with carrying the same emotional and stylistic nuances.

  2. Sentence Uniformity: ChatGPT tends to favor medium-length sentences—structured but lacking the engaging variability found in natural texts.

  3. Fine Line Between Data and Dialect: ChatGPT doesn’t fully master the art of weaving descriptive language and data, highlighting its mechanical underpinning beneath human-like text.

  4. Potential in Specific Niches: AI like ChatGPT is ideal for standard messages needing clear structure and uniformity, but still lacks the emotive touch required for uplifting speeches.

  5. The Future of AI in Speechwriting: As AI models develop, the challenge will be to close the gap between structured predictability and animated storytelling.

As AI continues its stride into domains traditionally dominated by humans, the challenge for machines will always be: how close can they get to imitating the dance in human dialogues—the pauses, the crescendos, and the nuanced whispers of tone and intention? For now, technological magic, it seems, can only approximate the real wizardry of human speech. But who knows what tomorrow holds?


This study serves as a reminder of AI's capabilities and limitations, inviting us to reflect on what makes our human touch truly unique. Whether you're a speechwriter or just someone crafting your next perfect email, understanding these nuances in AI-driven texts can empower you to challenge the machine and rise above its charming, yet predictable, prowess.

Stephen, Founder of The Prompt Index

About the Author

Stephen is the founder of The Prompt Index, the #1 AI resource platform. With a background in sales, data analysis, and artificial intelligence, Stephen has successfully leveraged AI to build a free platform that helps others integrate artificial intelligence into their lives.