OCR Not Recognizing Text in PDF: Causes and Fixes

OCR accuracy is almost entirely determined by the quality of the input image. The recognition engine, no matter how sophisticated, cannot recover information that was never captured in the scan. Low resolution, poor contrast, skewed pages, and compression artifacts all degrade OCR accuracy dramatically. Resolution is the most critical factor. OCR engines are calibrated to work best at 300 DPI. Scans below 200 DPI produce noticeably worse results, and scans below 150 DPI often fail entirely for small text. If your PDF was scanned at 72 or 96 DPI — common when scanning to email or using a phone camera — the resolution is simply too low for reliable character recognition. Contrast matters almost as much. Faded documents, pencil text, and colored backgrounds all reduce the contrast between characters and the background, making it harder for the engine to distinguish letter edges. Many OCR tools include built-in preprocessing that enhances contrast automatically, but severely faded documents may need manual enhancement before OCR.

1. 1Re-scan the document at 300 DPI minimum — most scanner apps and flatbed scanners offer a DPI setting in advanced options.
2. 2When scanning, use black-and-white or grayscale mode rather than color to improve contrast and reduce file size.
3. 3Ensure the document is flat on the scanner glass — even slight warping causes the text to curve, which confuses character segmentation.
4. 4For phone scans, use an app like Microsoft Lens or Adobe Scan that automatically corrects perspective and enhances contrast.

OCR engines use language models to interpret ambiguous characters and predict word boundaries. When the wrong language is selected, the engine applies incorrect probability weights — it might read an 'l' as a '1', a 'rn' as an 'm', or miss diacritical marks entirely. This is especially noticeable with non-Latin scripts. Running Arabic text through an English OCR profile, or running French text through a Spanish profile, produces dramatically worse results than using the correct language. Many tools default to English, so if your document is in another language, you must explicitly change the language setting before running OCR. LazyPDF's OCR tool uses Tesseract, which supports over 100 languages. Always select the correct primary language before processing. For multilingual documents, some Tesseract configurations allow combining multiple language packs (e.g., 'eng+fra') to improve accuracy across language boundaries within a single document.

OCR engines read text in straight horizontal lines. When a page is scanned at an angle — even a few degrees of skew — the character segmentation breaks down. Letters that overlap row boundaries get misclassified, and word spacing becomes inconsistent. A page tilted just 3–5 degrees can drop OCR accuracy from 95% to below 70%. Page orientation errors are even more damaging. A page scanned upside down or rotated 90 degrees will produce complete OCR failure — the engine cannot recognize characters that are sideways or inverted. Always check the orientation of every page before running OCR. Modern OCR tools include automatic deskew and orientation detection, but they do not always work correctly. If your output is garbled, manually rotate and straighten the pages first using LazyPDF's Rotate tool, then run OCR on the corrected document. This extra step reliably improves accuracy on scans with orientation issues.

1. 1Open the PDF and visually inspect every page for skew (diagonal tilt) or incorrect rotation.
2. 2Use LazyPDF's Rotate tool to correct any pages that are sideways or upside down.
3. 3For pages with diagonal skew, look for a 'deskew' option in your scanner app or image editor before creating the PDF.
4. 4After rotating, run OCR again on the corrected file to see if accuracy improves.

Standard OCR is trained on printed text and performs poorly on handwriting. Cursive handwriting in particular — where letters are connected and shapes vary by writer — is beyond the capability of most general-purpose OCR engines. Specialized handwriting recognition tools (ICR — Intelligent Character Recognition) exist but are not commonly available in free online tools. Decorative or unusual fonts also cause problems. Script fonts, condensed fonts, and fonts with heavy stylization share visual characteristics with multiple characters, causing frequent misclassification. Technical documents with mathematical symbols, chemical formulas, or musical notation are similarly problematic because the OCR engine has not been trained on those symbol sets. For these cases, the most practical solutions are: use a specialized tool designed for the content type (handwriting recognition apps, math OCR tools), manually correct the OCR output after the fact, or accept that automated OCR is not appropriate for this document type.

Even good OCR produces some errors. Common patterns include: '0' and 'O' confusion, '1', 'l', and 'I' confusion, split words where a space is incorrectly inserted mid-word, and merged words where word boundaries are not detected. These errors follow predictable patterns and can be corrected efficiently with find-and-replace operations. For important documents, always review OCR output manually before using it. Search for impossible character combinations (like '0pen' where it should be 'Open') and correct them. In Word or Google Docs, the spellcheck will flag many OCR errors automatically. For high-volume OCR work, scripting find-and-replace corrections or using a specialized post-correction tool is more practical than manual review. LazyPDF's OCR tool provides a clean text layer embedded in the PDF — you can also copy the recognized text directly from the output and paste it into a text editor for manual cleanup.