Convert PDF to Text File in Power Automate: Turn Scanned Contracts into a Searchable SharePoint Corpus with PDF4me

PDF4me Extract Text and Images is a Power Automate action that returns the text layer of a PDF as an array. Paired with OCR upstream and a Compose join downstream, this flow turns every PDF dropped in a SharePoint library into a matching .txt file, ready for AI search.

Search for how to do this and the first result is Microsoft's own PDF actions reference, which describes Power Automate Desktop actions. They do not exist in a cloud flow. That single detail is why the next two results are a Reddit thread and a Power Platform community thread of people asking the same question and not getting a straight answer. This flow is the straight answer, and it runs entirely in the cloud on standard-tier connectors.

Why-Based Q&A​

What You'll Get​

Input: any PDF dropped into a SharePoint document library, scanned or born-digital, no naming convention required. Output: one plain text file per PDF in a second library, containing the full readable contents of the document.

RagInput. Drop a PDF here and the flow starts.

RagText. One plain text file per PDF, ready to index.

Here is what comes out the other side for the sample contract, exactly as written to disk:

MASTER SERVICES AGREEMENT
 Between Initech LLC and Umbrella Inc.
Effective  date:  1 January 2026.  Auto-renews annually.
 Liability cap:  USD  100000. Confidentiality: 5 years.
 CLEAN TEST MARKER: PDF4ME-CLEAN-TEST-2026.

Note the leading spaces. OCR preserves layout whitespace, and that is fine for search and for feeding a language model. Trim it later if a downstream parser is fussy.

What You Need​

The Flow at a Glance​

1. When a file is created (properties only) (SharePoint trigger) scoped to /Shared Documents/RagInput.
2. Get file content using the trigger's Identifier.
3. PDF - Convert PDF to editable PDF using OCR with quality Draft and OCR Only When Needed true.
4. PDF - Extract Text and Images with Extract Text Yes and Extract Images No.
5. Compose running a join() over the returned texts array.
6. Create file (SharePoint) writing .txt into /Shared Documents/RagText.

Complete flow overview​

OCR is the only slow step at 11 seconds. The other five finish in under half a second each.

Step 1: How do you watch a SharePoint library for new PDFs?​

Flow so far: nothing yet, this is the trigger.

Use When a file is created (properties only). It fires on new items and returns their metadata. Scope it to one folder, because a library-wide watch fires on every upload anyone makes anywhere in that library.

1. Create an Automated cloud flow and pick When a file is created (properties only).
2. Configure:
   • Site Address: https://ynoox1.sharepoint.com/sites/PDF4meSharepoints
   • Library Name: pick Documents from the dropdown
3. Open Advanced parameters, add Folder, and set it to /Shared Documents/RagInput.

SharePoint trigger configuration​

Library Name is the library itself, Documents. The Folder parameter underneath is what narrows the trigger to one folder inside it.

Step 2: Why do you need Get file content after the trigger?​

Flow so far: SharePoint trigger.

The properties-only trigger hands over metadata, not the document. Get file content exchanges the trigger's Identifier for the actual bytes that PDF4me needs.

1. Add SharePoint > Get file content.
2. Configure:
   • Site Address: the same site as the trigger
   • File Identifier: Identifier from the trigger
3. Open Advanced parameters and set Infer content type to Yes.

Get file content configuration​

Infer content type Yes is what hands the next action real PDF bytes instead of a generic blob.

Step 3: Why run OCR before extracting text?​

Flow so far: SharePoint trigger plus Get file content.

A scanned page is an image. There is no text to extract until OCR puts a text layer on it. This action does that, and it is smart enough to leave already-readable PDFs alone.

1. Add PDF - Convert PDF to editable PDF using OCR.
2. Configure:
   • File Content: File Content from Get file content
   • File Name: the trigger's file name token
   • QualityType: Draft
   • OCR Only When Needed: true
   • Language: leave blank unless your documents are not in English
   • Is Async: No

OCR parameters​

QualityType is set to Draft here. On a clean scan it reads just as well as High and costs less time.

Step 4: How does Extract Text and Images return the text?​

Flow so far: SharePoint trigger plus Get file content plus OCR.

This action reads the text layer and hands it back as an array called texts. Point it at the OCR output, not back at Get file content, or scanned pages come back empty.

1. Add PDF - Extract Text and Images.
2. Configure:
   • File Content: File Content from the OCR action
   • File Name: the trigger's file name token
   • Extract Text: Yes
   • Extract Images: No

Extract Text and Images parameters​

Look at the icons. File Content carries the PDF4me mark because it comes from OCR. File Name carries the SharePoint mark because it comes from the trigger.

Step 5: How do you turn the texts array into one string?​

Flow so far: SharePoint trigger plus Get file content plus OCR plus Extract Text and Images.

texts is an array. A file needs a string. One join expression bridges the two.

1. Add a Compose action.
2. In Inputs, switch to the expression editor and paste:

join(body('PDF_-_Extract_Text_and_Images')?['texts'], decodeUriComponent('%0A'))

3. Click Update.

The decodeUriComponent('%0A') is how you write a literal newline in a Power Automate expression. There is no escape sequence for it, so this is the idiom.

Compose configuration​

The action name inside body() uses underscores for spaces and hyphens. Rename the action and this expression breaks.

Step 6: How do you save the text as a .txt file in SharePoint?​

Flow so far: SharePoint trigger plus Get file content plus OCR plus Extract Text and Images plus Compose.

The last step writes the joined string into a second library.

1. Add SharePoint > Create file.
2. Configure:
   • Site Address: same site as the trigger
   • Folder Path: /Shared Documents/RagText
   • File Name: the trigger's file name token followed by the literal text .txt
   • File Content: the Compose output token

SharePoint Create file configuration​

File Content is the Compose output, carrying the purple data-operation icon, not anything from PDF4me. The text has already been assembled by this point.

Run the Flow and Verify​

1. Save the flow at the top right.
2. Upload contract-sample.pdf into the input folder. The trigger polls, so give it a minute.
3. Open the flow's run history and check all six actions carry a green tick. OCR will be the slow one.
4. Open /Shared Documents/RagText. A .txt file should be sitting there. Open it and search for PDF4ME-CLEAN-TEST-2026. If that marker is present, OCR and extraction both worked end to end.

Common Variations You Can Add Without Rebuilding​

Common questions​

Can Power Automate extract text from PDF?​

Not on its own in a cloud flow. The PDF actions in Microsoft's documentation belong to Power Automate Desktop and are unavailable in cloud flows, which is why so many community threads on this question go unresolved. Adding the PDF4me connector gives a cloud flow a real Extract Text and Images action that returns the document's text as an array.

The whole extraction is one action. The surrounding five steps are just moving the file in and the text out.

How to extract text from PDF using Power Automate?​

Trigger on the new file, fetch its bytes with Get file content, run Convert PDF to editable PDF using OCR so scanned pages get a text layer, then run Extract Text and Images with Extract Text set to Yes. That returns a texts array. Join it with a Compose expression if you want prose rather than an array.

Point the extraction step at the OCR action's output rather than the file you downloaded. Reusing the downloaded copy is the reason scanned documents come back empty.

Can Power Automate do OCR?​

A cloud flow has no native OCR action. AI Builder offers one on a premium, credit-metered licence, and Power Automate Desktop has its own OCR engine tied to a machine. The PDF4me Convert PDF to editable PDF using OCR action runs in a standard-tier cloud flow with no machine involved.

Its OCR Only When Needed setting is worth knowing about. Set to true, it inspects each file and skips OCR when a text layer already exists, so a mixed library does not pay the cost on every document.

Why is Power Automate extract text from PDF not working?​

Three causes cover almost every case. The document is a scan and never went through OCR, so there is genuinely no text to find. The extraction step is reading the downloaded file rather than the OCR output, which produces the same empty result on scanned input. Or the text arrived fine but was written to the file as a raw array, so it looks like JSON rather than a document.

Open the run history and inspect the Extract Text and Images output directly. If texts has content there, the problem is downstream in Compose or Create file.

How do you convert PDF to a text file for free?​

For one document, any online converter will do it. The reason to build this flow is volume and repetition: it runs unattended on every file that lands in the folder, handles scans, and needs no one to be at a keyboard. PDF4me's free tier covers a meaningful number of documents per month, and every connector used here is standard tier, so there is no premium Power Automate licence involved.

Troubleshooting​

Next Steps​

The same six-step pattern (watch a folder, fetch, OCR, extract, flatten, save) turns any document library into a machine readable corpus.