Tables and structured documents are easier to read and reuse — when the model produces them consistently.
Specify Columns Exactly
"Create a table with columns: Option, Cost per month, Setup time, Main risk, Recommendation." Named columns in a set order produce predictable tables.
Define Cell Contents
State units, formats and length: "Costs in AUD without decimals; risks in under ten words."
Handle Missing Information
"Write 'unknown' where the information isn't in the source." This prevents invented values.
Choose the Format
Markdown tables for chat and documents, CSV for spreadsheets, JSON for programs. Ask explicitly.
Comparison Matrices
For comparisons, provide the criteria and ask the model to fill in each option, then summarise which option wins on which criteria.
Structured Reports
Provide a template with headings and a description of each section's content; the model fills it in.
Validate
For tables destined for spreadsheets or programs, parse the output and check column counts and types.
Large Tables
For many rows, generate in batches, or have the model write code that produces the table from data, which is more reliable than generating many rows of text.