Creating Searchable PDFs

← Back to Blog

Image-Only PDFs vs Text PDFs

Not all PDF files contain selectable, searchable text. When a physical document is scanned on a photocopier or flatbed scanner and saved as a PDF without further processing, the result is an image-only PDF. Each page is simply a raster image wrapped in a PDF container. There is no text layer: no word on the page can be selected, copied, or found with Ctrl+F. The document looks like a normal PDF on screen, but from a machine-readability perspective it behaves like a photograph.

A text PDF, by contrast, encodes the actual character data alongside glyph positions. This text can be selected, copied to the clipboard, indexed by search engines, and processed programmatically. Any PDF produced by exporting from a word processor, desktop publishing application, or modern office suite will normally be a text PDF. The distinction matters wherever content needs to be found, reused, or made accessible.

Why Searchability Matters

The practical benefits of searchable PDFs are significant across several domains. Users can locate specific terms within a long document in seconds rather than reading page by page — critical for legal contracts, technical manuals, and research reports. Text can be copied and pasted into other applications without re-keying. Desktop indexers such as Windows Search and enterprise document management systems can index content and surface documents in relevant searches. Web crawlers can parse PDF text for SEO purposes, meaning searchable PDFs rank in search results while image-only PDFs are effectively invisible to search engines. Screen readers and other assistive technologies depend entirely on a text layer to read content aloud — accessibility frameworks including PDF/UA (ISO 14289) and WCAG explicitly require that PDF text be machine-readable rather than image-based.

How OCR Creates a Hidden Text Layer

Optical Character Recognition (OCR) is the process of analysing a raster image and identifying the characters it depicts. When OCR is applied to a scanned page, the engine outputs a stream of recognised characters with bounding-box coordinates corresponding to their positions on the page image. In a searchable PDF, this output is stored as an invisible text layer positioned directly beneath the visible page image. The image remains the visual representation; the text layer provides the machine-readable substrate. This architecture — sometimes called a sandwich PDF or image-over-text PDF — preserves the exact appearance of the original scan while enabling all the benefits of embedded text.

Using Acrobat’s Recognize Text (OCR) Feature

Adobe Acrobat Pro includes a built-in OCR engine accessible via Tools > Scan & OCR > Recognize Text. You can process the currently open document or a batch of files. Acrobat analyses each page, detects image-only pages automatically, and runs recognition on those pages. Pages that already contain a text layer are skipped unless re-recognition is explicitly requested.

The primary settings to configure before running OCR are:

• Document language: The OCR engine uses language models to improve accuracy. Selecting the correct primary language significantly reduces recognition errors, particularly for common words and proper names.
• Output type: Searchable Image preserves the original scan as the visual layer with text placed invisibly underneath — the safest choice for archival and distribution. Editable Text and Images attempts to replace the scan with live text, which can alter appearance unpredictably for complex layouts.
• Downsample images: Acrobat can reduce image resolution after OCR to lower file size. This is best deferred until after recognition, since higher resolution improves accuracy.

Resolution and OCR Accuracy

Scan resolution has a direct and well-documented impact on OCR accuracy. The industry-standard recommendation is 300 dpi for body text, with 400–600 dpi preferred for small type, degraded originals, or documents with fine detail. At 200 dpi or below, character separation becomes ambiguous and error rates rise sharply. Colour mode also matters: black-and-white or greyscale scanning provides maximum contrast and keeps file sizes manageable; full-colour scanning adds file size with no benefit to recognition quality for most documents.

Verifying Searchability

After running OCR, always verify the result. Open the document in Acrobat or a PDF reader and press Ctrl+F (Cmd+F on macOS). Search for a word visible on the page — if it is highlighted when found, the text layer is present and functional. You can also try selecting text with the cursor: if individual characters highlight under the text selection tool, OCR has succeeded. For a more thorough check, Acrobat’s Accessibility Checker (Tools > Accessibility > Full Check) will flag image-only pages as accessibility failures when no text layer is detected.

Making PDFs Searchable Programmatically

For high-volume workflows — scanning bureaux, document management systems, or automated ingestion pipelines — interactive OCR in Acrobat is impractical. Acrobat Pro’s Action Wizard supports unattended batch processing, applying OCR with consistent settings across folders of files. For fully programmatic control, the Acrobat SDK exposes the OCR engine through its API, allowing developers to trigger recognition, configure language and output settings, and handle results entirely from code — suitable for server-side document processing at scale.

For a deep dive into automating OCR with the Acrobat SDK and API, see our dedicated article on OCR in Adobe Acrobat.