# Column settings (https://www.clavy.ai/help/column-settings)

The prompt, web search, reasoning, citations, and model for each column.

Each auto-fill column has its own settings. Select the column's header to change them.

## Prompt [#prompt]

The prompt is the question the column answers for every row, and it shapes the results more than any other setting. See [Write column prompts](/help/write-column-prompts).

## Web search [#web-search]

Turn on **Web search** when the answer isn't in your table or documents, like a company's latest funding or a person's current role. The agent searches the web for each row and lists the pages it used.

Leave it off when the answer should come only from the row's own data or documents. That keeps answers grounded in your sources.

## Reasoning [#reasoning]

Turn on **Reasoning** to have the agent work through each answer before writing it, and keep its explanation. It helps with judgment calls, like scoring or classifying, and lets you see why the agent answered the way it did. Reasoning takes longer and can use more credits.

## Citations [#citations]

Turn on **Citations** to link each answer to the exact passage in the document it came from. The option appears when the prompt mentions a file column. See [Review answers and sources](/help/review-answers-and-sources).

## Model [#model]

Each column can use a different AI model, from OpenAI, Anthropic, or Google.

- **Larger models** are more thorough. Use them for nuanced reading, like reviewing contracts.
- **Smaller models** are faster and use fewer credits. Use them for simple lookups and extraction across many rows.

Credits follow what each model costs to run, so the model choice is the biggest lever on how many credits a column uses. See [Credits](/help/credits).

## Changing settings [#changing-settings]

Changing a setting marks the column's filled cells as outdated. Nothing reruns until you choose to. See [Run columns](/help/run-columns).