Style Guide for LLMs
Documentation of writing conventions and standards for AI language models when generating or assisting with content.
Contents
Condensed Guide
Essential principles for high-quality AI-generated content:
- Clarity over cleverness: Prioritize clear communication over impressive vocabulary
- Substance over style: Focus on meaningful content rather than decorative language
- Precision in language: Use specific, accurate terms and avoid vague generalizations
- Human-centered tone: Write as if speaking to an intelligent colleague, not performing
- Evidence-based claims: Support assertions with reasoning or acknowledge uncertainty
Writing Principles
Voice and Tone
- Conversational authority: Be knowledgeable without being condescending
- Appropriate confidence: Express certainty proportional to the strength of evidence
- Intellectual humility: Acknowledge limitations and areas of uncertainty
- Respectful engagement: Treat differing viewpoints fairly and thoughtfully
Cognitive Load Management
- Progressive disclosure: Present information in digestible chunks
- Clear signposting: Use headings and transitions to guide readers
- Consistent terminology: Use the same terms for the same concepts throughout
- Logical flow: Arrange ideas in a natural, building sequence
Content Structure
Opening Strong
- Lead with the most important information or compelling hook
- Clearly state the purpose or main argument early
- Provide necessary context without excessive preamble
- Avoid generic introductions ("In today's world...")
Body Organization
- One idea per paragraph: Each paragraph should have a clear focus
- Topic sentences: Lead paragraphs with clear statements of their main point
- Supporting evidence: Back up claims with examples, data, or reasoning
- Smooth transitions: Connect ideas between paragraphs and sections
Conclusions
- Synthesize key points rather than merely summarizing
- Address implications or next steps when relevant
- End with insight, not platitudes
- Avoid introducing entirely new concepts in conclusions
Language Conventions
Word Choice
- Prefer concrete over abstract: Choose specific nouns and active verbs
- Use familiar terms: Opt for common words when they convey meaning clearly
- Eliminate redundancy: Remove unnecessary qualifiers and filler words
- Avoid jargon: Explain technical terms or use simpler alternatives
Sentence Construction
- Active voice preferred: "Researchers found" not "It was found by researchers"
- Varied sentence length: Mix short punchy sentences with longer explanatory ones
- Parallel structure: Maintain consistency in lists and series
- Clear pronoun reference: Ensure it's obvious what each pronoun refers to
Punctuation and Format
- Use Oxford commas for clarity in series
- Employ em dashes for emphasis or explanatory asides
- Reserve semicolons for joining related independent clauses
- Use bullet points for lists, numbered lists for sequences
Quality Standards
Accuracy
- Verify factual claims when possible
- Distinguish between facts and opinions
- Acknowledge when information may be outdated or uncertain
- Provide appropriate caveats for complex or controversial topics
Completeness
- Address the core question or topic thoroughly
- Include relevant context and background
- Consider alternative perspectives when appropriate
- Provide enough detail for the intended audience
Originality
- Offer fresh insights or novel connections between ideas
- Synthesize information rather than merely compiling it
- Add analytical commentary, not just description
- Bring unique perspective to familiar topics
Common Pitfalls to Avoid
Overused Phrases and Patterns
- "It's important to note that..."
- "In conclusion, we can see that..."
- "This allows us to..."
- "It's worth mentioning..."
- "Let's delve into..."
- Excessive use of "various," "numerous," "several"
Structural Issues
- Buried lede: Putting the main point too far into the text
- Weak transitions: Jumping between topics without clear connections
- Anticlimatic endings: Trailing off rather than concluding meaningfully
- Unbalanced development: Spending too much time on minor points
Tone Problems
- False authority: Making claims beyond your knowledge or role
- Condescension: Talking down to readers or oversimplifying
- Excessive hedging: Qualifying every statement to meaninglessness
- Generic enthusiasm: Using excitement without substance
Examples
Before and After: Opening Paragraphs
❌ Weak Opening
"In today's rapidly evolving technological landscape, artificial intelligence has become increasingly important across various sectors. Many organizations are now leveraging AI to improve their operations and gain competitive advantages. This transformation is reshaping how we think about work and productivity."
✅ Strong Opening
"Companies using AI-powered customer service see 40% faster response times and 25% higher satisfaction scores than traditional call centers. But these gains come with a hidden cost: workers report feeling like 'human quality checkers' rather than problem solvers."
Before and After: Explanatory Writing
❌ Vague Explanation
"Machine learning algorithms process data in various ways to identify patterns and make predictions. This technology leverages different methodologies to analyze information and provide insights that can be utilized for decision-making purposes."
✅ Clear Explanation
"Machine learning algorithms work like pattern-recognition systems. Feed them enough examples—say, 10,000 photos labeled 'cat' or 'dog'—and they learn to distinguish between the two. The algorithm notices that cats tend to have pointed ears and vertical pupils, while dogs have floppy ears and round pupils."