> AI agents: this is one page from Mammoth Analytics documentation. The index of all pages as Markdown is https://mammoth.io/docs/llms.txt. Append `.md` to any docs URL, or send `Accept: text/markdown`, to get Markdown.

# Add PDFs to a dataset from a watched folder

> How a Mammoth watched folder reads new PDFs and messy spreadsheets with the destination's recipe, and what to do when a file is kept separate or not read.
>
> Applies to: Where: Add data › read a PDF or spreadsheet, then the watched folder's Data Consolidation

If the destination dataset was made by reading a PDF or a messy spreadsheet in **Add data**, it has a recipe. New files in the folder are read with that recipe, then added. The type picker points to this path: "Reading a spreadsheet or PDF? Start from **Add data**, and the automation comes with it." See [Extract tables from PDFs](https://mammoth.io/docs/learn/get-data-in/extract-from-pdfs/).

## When a PDF or spreadsheet is not added

When the recipe can't be used, the file is not added:

| Reason shown on the file | What to do |
| --- | --- |
| "Sales's recipe couldn't read it, so it was kept as its own dataset" | It stays as its own dataset, marked **Kept separate** and **Meant for Sales**. Click **Review** |
| "No table in it matches Sales's recipe, so it was kept as its own dataset" | Check the PDF is the same kind of document as the earlier ones |
| "Sales has no recipe yet, so it was kept as its own dataset" | Read a file once in the interpreter to create the recipe |
| "Read it once in the interpreter, so Sales can take files like it" | Click **Review** and read the file in the interpreter |
| "Not added to Sales yet: the first one waits for you to check it was read right" | The first file read with a new recipe waits for your check. Click **Review** |
| "Not added to Sales: its figures don’t add up" | The read figures don't add up to the total the recipe expects. Check the file against the document |
| "Not read: this month's AI pages for PDFs are used up" | Wait for next month's AI pages, then add the file again |
| "Not read: PDFs need an AI key, and none is set up" | Ask your admin to set up the AI key |
| "Its rule names several datasets, and a PDF can go to one" | Give the PDF its own rule with a single destination |

When a PDF or image batch couldn't be matched to the recipe, the automation's status shows **Needs your input** until someone deals with it.

## Related

- [Fix files a watched folder didn't add](https://mammoth.io/docs/learn/automate/watched-folder-files/): every chip and action on a waiting file.
- [Extract tables from PDFs](https://mammoth.io/docs/learn/get-data-in/extract-from-pdfs/): create the recipe that reads new PDFs.
- [Combine files dropped in a folder into one dataset](https://mammoth.io/docs/learn/automate/consolidate/): set up the watched folder.

---
Source: https://mammoth.io/docs/learn/automate/pdf-recipe-files.md · Updated: 2026-11-08