AI won't follow our document template
You have a standard structure for your documents, but the AI reorders sections, drops the parts it thinks are unimportant, and invents its own headings, so every draft needs restructuring by hand.
Open the fixLast verified: Claude Fable 5 / GPT-5 era models · July 2026
The table looks perfect in the chat. Then you paste it into Google Sheets and the whole thing lands in one cell, or into Google Docs and the pipes show up as literal text. So someone rebuilds it column by column, which is exactly the work the AI was meant to remove. This page gives you a prompt that makes the AI output the right format for wherever the table is going, plus the two paste paths that actually work.
Paste this into a fresh Claude or ChatGPT chat. It interviews you about where your tables go and how your numbers should look, then hands back a table block you paste at the top of any chat so the AI produces the right format instead of a table that shatters on paste.
You're going to build my company's TABLE OUTPUT BLOCK: a reusable instruction block I'll paste into any AI chat that produces tables, so they land cleanly wherever I paste them. Interview me first, one question at a time. Where I don't know, propose a sensible default and move on: 1. Where do our AI-produced tables usually go: Google Sheets or Excel (a spreadsheet), or Google Docs or Word (a document)? Sometimes both? 2. What kinds of table do we make most (pricing, comparisons, schedules, data summaries) and roughly how many columns? 3. Number conventions: currency format, decimal places, thousands separators, how we show percentages and dates. 4. Any structure rules: header row style, a totals row, columns that must always appear, maximum width before it should become a list instead. 5. What goes wrong today: everything in one cell, wrong split points, literal pipes in Docs, numbers turning into text or dates? Then output my TABLE OUTPUT BLOCK as one copy-paste block containing: (a) one line of context, (b) the destination rule: for a spreadsheet, output the table as CSV or TSV in a code block, never as a markdown table; for a document, output a standard markdown table; ask which destination if I haven't said, (c) our number, currency, date and percentage conventions, (d) structure rules: header row, totals, mandatory columns, and when to use a list instead of a table, (e) a self-audit that checks the format matches the destination and the number conventions before output, then appends "Table check: pass" or lists what is off. Keep it under 400 words so it fits any project-instructions field. After you output it, tell me the exact steps to paste each format cleanly (CSV or TSV into a spreadsheet, markdown into Docs) and the two mistakes most likely to still break a table.
The block makes the AI output the right format. These are the paste steps that keep it intact.
Here is the block the generator hands back, for our running example company, the fictional consultancy Fenwick & Cole.
CONTEXT: Fenwick & Cole tables are usually pricing and option comparisons, going into either client documents (Google Docs) or working spreadsheets (Google Sheets). DESTINATION RULE: if the table is for a spreadsheet, output it as CSV or TSV inside a code block, never as a markdown table. If it is for a document, output a standard markdown table. If I haven't said which, ask before producing the table. NUMBERS: currency as £4,000 (no decimals unless I ask), thousands separated with commas, percentages as 12% (no decimal unless needed), dates as 4 July 2026. STRUCTURE: header row always present, and bold in document tables. Pricing tables end with a Total row. Maximum four columns; if the content needs more, offer a list instead. Never put long prose in a cell. SELF-AUDIT: before output, confirm the format matches the destination (CSV/TSV for a spreadsheet, markdown for a document) and the numbers follow the conventions above. Append "Table check: pass", or list what is off.
The AI gives you a markdown table: text with pipe characters and dashes. Spreadsheets don't read markdown, so the whole thing pastes into a single cell. Spreadsheets do read CSV and TSV, where columns are separated by commas or tabs. Ask the AI for CSV or TSV instead of a markdown table and the columns land in the right cells.
Tell it to output the table as CSV or TSV in a code block rather than a markdown table. Copy that, paste into the sheet, and if it arrives in one cell use Data then Split text to columns. TSV usually splits on paste automatically. The generator prompt on this page bakes that destination rule into a reusable block so you don't ask each time.
Ask the AI for a standard markdown table, then in Google Docs use Edit then Paste from Markdown instead of Ctrl+V. Paste from Markdown turns the table into a real Docs table; Ctrl+V leaves the pipes and dashes as literal characters. This is the same paste path that fixes AI Markdown generally in Google Docs.
You have a standard structure for your documents, but the AI reorders sections, drops the parts it thinks are unimportant, and invents its own headings, so every draft needs restructuring by hand.
Open the fixThe AI praises weak ideas, caves when you push back and tells you what you want to hear. For decisions, plans and drafts, that makes its feedback close to useless.
Open the fixAI leans on the em dash, 'not X but Y' sentences, lists of three and stock openers. Readers spot them instantly, so someone edits every draft to take the tells out.
Open the fixThe table block you just generated is exactly what an AI Brain stores as always-on context. Paste it and this table lands cleanly; installed, the AI already knows your destinations and number conventions and formats every table for where it's going, so nobody rebuilds columns by hand. This page proved it on your own tables. The install is how the right format becomes automatic.
What else does AI get wrong that you fix by hand? Browse the other fixes
Free founder tools
Each tool gives you a first-pass answer in your browser. No email gate and nothing stored.