Copying from PDFs sucks. AI finally fixes it.

Welcome to this edition of Prompt, Tinker, Innovate—my AI playground. In each edition, we dive into a practical AI use case, explore its impact, and give you a hands-on experiment to try.

Your mission? Test the prompt, tweak it, and see how AI can enhance your workflow. Before you know it, you’ll be innovating with AI in your daily life—one experiment at a time.

This week’s playground: Clean PDF extraction with AI

Why this matters

PDFs were built for reading—not for editing or reuse. The result? Copying from them often turns into a formatting nightmare. Sentences break awkwardly, tables flatten into gibberish, and headers vanish. What looks crisp on screen turns into chaos in your editor.

AI can fix that.

It doesn’t just copy—it interprets. It understands layout, reconstructs structure, and delivers usable, clean content. Whether you're pulling policy text, editing old decks, or reusing report data, this saves hours of manual cleanup.

Your AI experiment: Try this prompt

👉 Time to tinker: Copy and paste the prompt below into ChatGPT, Claude, Gemini, or your favorite AI tool. While all models are improving, I’ve found the most accurate results come from the more advanced “thinking” models.

📝 Prompt: “You are a document analyst trained to extract verbatim, clean, readable text from PDF files. Your task is to accurately convert the contents of this PDF into structured text while removing unnecessary line breaks and preserving headings and bullet points. Return the word-for-word output as plain text that’s ready for editing.”

(Upload your own PDF when using this prompt.)

💡 Pro tip: Get more from your extracted text

• “Format this as a Google Doc outline with H1s, bullet points, and page breaks preserved.”

• “Extract and organize content into labeled blocks—headlines, CTAs, data points, and examples—for reuse in future assets.”

• “Reformat this into a set of tweet threads, social captions, or talking points based on the extracted content.”

What did you discover?

Did AI clean up your PDF better than expected? Or did it fumble something critical—like table layouts or footnotes?

📩 See you next week for another AI experiment.