The AI Voice: False Enthusiasm and List Addiction
Session 1.3 · ~5 min read
AI writing has two modes of emphasis: exclamation marks and bullet points. When it wants to convey importance, it reaches for superlatives. When it wants to organize information, it reaches for lists. Both are formatting shortcuts that substitute for actual thought.
False Enthusiasm
Real enthusiasm is specific. "I stayed up until 3 AM rewriting this function because the benchmarks finally made sense" is enthusiasm. "This is a truly powerful and exciting approach!" is a formatting pattern. The difference is that specific enthusiasm reveals experience. Generic enthusiasm reveals nothing except a tendency toward exclamation marks.
AI defaults to enthusiasm because RLHF training rewards positive, encouraging responses. A rater comparing "This method can be useful in some situations" against "This is a powerful method that can transform your workflow!" will often rate the second as more helpful, because it sounds more confident. The model learns that superlatives correlate with higher ratings.
| AI Enthusiasm Pattern | Example | What It Actually Means |
|---|---|---|
| Superlative stacking | "A truly remarkable and groundbreaking approach" | "An approach" (the adjectives add no information) |
| Exclamation inflation | "The results were amazing!" | "There were results" (no specifics given) |
| Transformation claims | "This will revolutionize your workflow" | "This might help" (no evidence of revolution) |
| Empowerment language | "Unlock your full potential" | Filler. Means nothing measurable. |
| Certainty without evidence | "You'll be amazed at the difference" | Promise without proof |
Real enthusiasm is specific. AI enthusiasm is generic. "This changed how I think about database indexing" is enthusiasm. "This is a game-changer!" is punctuation.
List Addiction
Lists are not inherently bad. A list of ingredients, a list of steps in a procedure, a list of error codes: these are appropriate uses of the format. The problem is when lists become the default structure for everything, regardless of whether the content benefits from list formatting.
AI defaults to lists for the same reason it defaults to hedging: lists are safe. A list of five points does not require the writer to build an argument, show cause and effect, or connect ideas. Each point stands alone. There is no logical throughline to maintain, no transitions to write, no cumulative reasoning to develop.
'Explain project management'"] --> B{"Default formatting
decision"} B -->|"Lists: easy"| C["Bulleted list
5-10 items"] B -->|"Prose: harder"| D["Connected paragraphs
building an argument"] C --> E["Each point is
independent, generic"] D --> F["Ideas build on
each other"] E --> G["Reader gets fragments"] F --> H["Reader gets understanding"]
How Lists Break Arguments
Consider a question like "Why do software projects fail?" An AI response will almost always produce a list: poor requirements, scope creep, inadequate testing, communication failures, unrealistic timelines. Five items. Each one is true. None of them connect to each other.
A practitioner writing about project failure would tell a story. Poor requirements led to scope creep because the team kept discovering things the client forgot to mention. Scope creep compressed the timeline, which forced the team to skip testing. Skipped testing meant bugs in production, which caused communication breakdowns between the dev team and the client. The failure was a chain, not a list.
Lists present information as if each item is equally important and independent. In most real-world situations, that is not true. Causes are connected. Effects cascade. Priority matters. A list of five equal bullet points obscures the structure that would make the information useful.
The Combined Effect
False enthusiasm and list addiction often appear together, producing a distinctive AI output pattern. The structure typically follows this formula:
Opening paragraph with superlative claim about the topic's importance. Followed by a numbered list of 5-10 points, each beginning with a bold keyword and a colon. Each point contains 2-3 sentences of generic explanation. Closing paragraph restating how exciting and powerful the topic is.
This structure is the equivalent of a template. It works for any topic because it commits to no specific argument about any topic. You can swap the subject and the output looks identical. An article about project management, database optimization, and cake decorating can all follow the same structure because the structure does not emerge from the content. It is imposed on the content regardless of what the content is.
| Aspect | List Format (AI Default) | Prose Format (Human Writing) |
|---|---|---|
| Structure | Parallel, independent items | Sequential, ideas build on each other |
| Causality | Implied at best | Explicit, with transitions showing relationships |
| Priority | All items appear equally important | Emphasis reflects actual importance |
| Memorability | Low (lists blur together) | Higher (narrative sticks) |
| Effort to produce | Low (no argument structure needed) | Higher (requires logical threading) |
When Lists Are Correct
Lists serve specific purposes well. Reference data (a list of API endpoints), procedural steps (a deployment checklist), and truly parallel items (features of competing products) belong in list format. The rule is simple: if the items are genuinely independent and parallel, use a list. If the items have relationships, causality, or hierarchy, use prose.
The problem with AI is not that it uses lists. The problem is that it uses lists when prose would communicate the idea better, because lists are easier to generate and harder to get wrong.
Further Reading
- A Survey of AI-generated Text Forensic Systems (arXiv)
- Stylometry: How AI Detectors Identify Writing Style (NetusAI)
- Presenting Bulleted Lists (Nielsen Norman Group)
- Creating Helpful, Reliable, People-First Content (Google Search Central)
Assignment
- Find an AI-generated article that uses excessive lists (3 or more bulleted/numbered lists in a single article).
- Rewrite the main section as flowing prose: connected paragraphs with transitions, arguments that build on each other, explicit causal relationships between ideas.
- Compare the two versions. Which communicates a complete idea? Which would you trust more as a reader?
- Write a 1-paragraph reflection on when lists are appropriate versus when they are a crutch.