Accessible PDF Converter

Integrations

ChatGPT Custom GPT

Use the Accessible PDF Converter directly inside ChatGPT with our Custom GPT action

The Accessible PDF Converter is available as a Custom GPT action in OpenAI's ChatGPT. This lets you convert PDFs to WCAG 2.1 AA compliant accessible HTML without leaving the ChatGPT interface.

What you can do

With the Custom GPT installed, you can ask ChatGPT to:

  • Convert a PDF from a URL to accessible HTML
  • Check the progress of a conversion
  • Retrieve and display the converted HTML result
  • Ask follow-up questions about the converted content

ChatGPT handles the entire workflow — uploading, polling for progress, and retrieving results — through natural conversation.

Prerequisites

Before using the Custom GPT, you need an API key from the Accessible PDF Converter. Contact your team administrator or reach out to support to obtain one.

Installing the Custom GPT

  1. Open ChatGPT in your browser
  2. Navigate to the GPT Store or use the direct link provided by your administrator
  3. Click Start Chat to begin using the Accessible PDF Converter GPT
  4. When prompted, enter your API key to authenticate

Converting a PDF

Once the GPT is installed, simply paste a link to a PDF and ask ChatGPT to convert it:

"Convert this PDF to accessible HTML: https://example.com/report.pdf"

What happens behind the scenes

  1. Upload — The GPT sends the PDF URL to the converter API
  2. Processing — The converter downloads the PDF, analyzes it with AI vision, and generates WCAG-compliant HTML
  3. Polling — ChatGPT automatically checks progress every 10-15 seconds
  4. Result — Once complete, the accessible HTML is returned directly in the chat

Conversion time

Processing time depends on document length:

Document sizeApproximate time
1-5 pages30-60 seconds
5-20 pages1-3 minutes
20-50 pages3-8 minutes
50+ pages8-12 minutes

ChatGPT will keep you updated on progress as the conversion runs.

Viewing results

Small documents (under 50KB HTML)

For most documents, the full accessible HTML is returned directly in the chat. You can:

  • Read the converted content in the conversation
  • Ask ChatGPT questions about the content
  • Request specific sections or summaries

Large documents (over 50KB HTML)

For larger conversions, ChatGPT receives a preview of the first portion of the HTML along with a link to download the full file from your dashboard at pdf.theaccessible.org.

ChatGPT will provide the download link and let you know the full document is available on the web dashboard.

Example conversation

You: Convert this syllabus to accessible HTML: https://university.edu/courses/cs101-syllabus.pdf

GPT: I've started the conversion for "cs101-syllabus.pdf" (estimated 8 pages). Let me check the progress...

The conversion is 45% complete — currently processing pages.

...

The conversion is complete! Here's the accessible HTML version of your syllabus:

(accessible HTML content displayed)

You: Summarize the key dates from this syllabus.

GPT: Based on the converted content, here are the key dates...

Troubleshooting

"Invalid or missing API key"

Your API key may be expired or incorrectly entered. Re-enter your API key when ChatGPT prompts for authentication, or contact your administrator for a new key.

"The file is not a valid PDF"

The URL must point directly to a PDF file. Some links may point to a webpage that contains a PDF viewer rather than the PDF itself. Try right-clicking the download link and copying the direct URL.

"The remote server took too long to respond"

The server hosting the PDF did not respond within 30 seconds. This usually means the file server is slow or the PDF is behind authentication. Try downloading the PDF first and hosting it on an accessible URL.

"Conversion timed out"

Very large or complex documents may exceed the 12-minute processing limit. Try splitting the PDF into smaller sections and converting each one separately.

Conversion failed

If a conversion fails, ChatGPT will display the error message. Common causes include:

  • Scanned images without OCR text layers
  • Password-protected PDFs
  • Corrupted PDF files

You can retry by sending the same URL again in the conversation.

Limitations

  • URL-only uploads — The GPT action accepts PDF URLs, not direct file uploads from your computer. Upload the PDF to a file hosting service first if needed.
  • File size limit — PDFs must be under 50MB.
  • Large HTML results — Documents that produce more than 50KB of HTML will be truncated in the chat, with a link to the full file on your dashboard.
  • Rate limits — API key usage is subject to your plan's rate limits.

Security and privacy

  • All PDF transfers use HTTPS encryption
  • Files are processed on secure servers and stored in your account
  • The GPT action authenticates every request with your API key
  • No PDF content is shared with OpenAI beyond what ChatGPT displays in the conversation