Troubleshooting AI by Zapier PDF processing from Google Drive to Google Sheets

Hi ​@pinneapple 

Help links for using AI by Zapier: https://zapier.com/apps/ai/integrations#help

Make sure you are using the current AI by Zapier Zap app version in the Zap step.

For the input field, make sure you are setting the type as FILE. (see screenshot)

For us to have more context, post screenshots showing how your full AI by Zapier step is configured, so we can see the field mappings and settings.

The symptom sounds like the AI step is receiving a file reference/URL but not the actual readable PDF content.

I would test it in this order:

1. In the Google Drive trigger, confirm the field passed to AI is the actual file object or downloadable file field, not just the Drive web URL.
2. Add a small Formatter/Text or Code step before AI only to log the file name, mime type, and whether the file field is populated. Do not log the PDF contents.
3. Try one simple text-based PDF first. If that works but scanned PDFs fail, you need OCR before the extraction step.
4. Ask the AI step for strict JSON with the 8 fields and default blanks, then map those JSON fields to Google Sheets. This makes it easier to see whether extraction failed or only the Sheets mapping failed.
5. If the PDF is stored in a restricted Drive folder, test with one temporary non-sensitive PDF in a folder owned by the connected Google account.

I would keep the Zap as: Drive file trigger -> confirm readable file -> extract structured JSON -> validate required fields -> add row to Sheets. That keeps PDF access problems separate from Sheets mapping problems.

Hi everyone,

I'm building a Zap that extracts data from PDF files stored in Google Drive and populates a Google Sheets spreadsheet. I've had this automation working a few months ago with the same setup, so I'm not sure what changed.

What I'm trying to do:

1. Trigger: New file in Google Drive folder (PDF reports)
2. AI step: Extract 8 specific fields from the PDF (date, project name, company, responsible person, number of workers, incidents, preventive measures, signature)
3. Action: Add extracted data to Google Sheets

The Problem: When I test the Zap, ALL fields in the spreadsheet are filled with "No especificado" (Not specified). This tells me the AI is not accessing the PDF content at all, it's receiving something but can't read it.

What Zapier Copilot told me: The copilot suggested that the issue is likely that I'm passing the PDF URL (application/pdf export link) to the AI, but AI by Zapier cannot directly download and process PDFs. The suggestion was to:

1. Change the document_content input field from {{=gives["367881179"]["exportLinks"]["application/pdf"]}} to {{=gives["367881179"]["exportLinks"]["text/markdown"]}}
2. Enable a checkbox called "Is 'document_content' a file URL?" to tell AI to download and process the file

What I found in the documentation: According to the AI by Zapier help docs, Input fields should have a "What type of content is this?" dropdown with three options:

• Text (default)
• Image (for image URLs)
• File (for file URLs like PDFs)

However, this dropdown is NOT appearing in my UI nor the checkbox “Is ‘document_content’ a file URL?”, even after changing the field value. I only see:

• Field name
• Field value
• No "What type of content is this?" option below

My question:

• Should the "What type of content is this?" dropdown always appear after adding an input field, or only under certain conditions?
• Is there another way to tell AI by Zapier to process a PDF from a URL?
• Could this be a UI bug or am I missing something?

I'll attach screenshots showing exactly what I see in my Zap configuration:
 

Any help would be greatly appreciated!

Hi, ​@pinneapple, Has this been resolved yet? If not, 

Since all fields are returning "No especificado," I'd focus on confirming whether the AI step is actually receiving the PDF content rather than the extraction prompt itself.

Given that this worked previously, I'm also wondering whether the AI by Zapier interface or file-handling behavior has changed since you first built the workflow.

If you'd like, happy to take a look at the screenshots and setup with you:

--link removed by moderator--

The missing dropdown is the clue everyone skipped. That selector only exists on the current version of the AI by Zapier app, and Zaps built months ago keep running their old step version forever, editing the field value never upgrades it. Delete the AI step and add a fresh AI by Zapier step in its place, and the "What type of content is this?" option should appear under the input field.

Two more things once it shows up. Map the trigger's File field (the hydrated file object), not the exportLinks URL, because export links require Google auth, so the downloader gets a login page instead of your PDF, which is exactly how every field comes back as "No especificado". And the fact that this worked a few months ago fits the same story, so rebuilding the step on the current version usually clears both problems at once.