CSV data for LLMs: compact tables that fit your prompt

Workflow

1. 1Upload the CSV. Large files stream during processing, so the browser stays responsive.
2. 2Open Customize output and find the CSV options. Check "Compact table output for AI" to reduce token use, or pick the Tables/data output goal to apply table-friendly settings in one step.
3. 3Confirm "First row contains headers" matches your file and set the delimiter if the export uses semicolons or tabs.
4. 4Set "Maximum rows" for the first pass so the prompt stays focused, then process and copy the output into your AI tool.

Extraction tips

• Start with a few hundred rows. A focused sample usually answers the question better than a full dump that crowds the context window.
• Keep "Skip empty lines" on so blank export rows do not waste tokens.
• Ask the model to restate the column headers before analyzing so you can confirm it read the table correctly.

Frequently asked questions

How large a CSV can I paste into an LLM?

It depends on the model’s context window. Compact table output and a row cap shrink the data substantially. If a file is still too large, filter to relevant columns or split it by topic before converting.

What does "Compact table output for AI" change?

It compresses the table representation to reduce repeated formatting and lower token use while keeping headers and values readable for the model.

Does my data stay private?

Yes. The CSV is parsed in your browser and is not uploaded during the standard conversion flow. You can also turn on Privacy cleanup to label values such as emails and phone numbers before export.

My file uses semicolons or tabs. Will it parse correctly?

Yes. Set the delimiter in the CSV options to match the export. The header checkbox and row limit work the same regardless of delimiter.

Related guides

• Excel to Claude
• Prepare documents for AI