# Jumbled, multi-column and unusual layouts > RemakeCV detects when a PDF's text comes out in the wrong reading order and automatically re-extracts it via OCR to recover the intended order. Source: https://www.remakecv.com/help/troubleshooting/jumbled-and-multi-column-cvs Last updated: 2026-08-21 --- Multi-column and heavily designed PDFs often yield text in the wrong reading order — columns interleaved, sections torn apart. RemakeCV detects this automatically and re-extracts the document through OCR, which recovers the intended order from the page layout. Where the source is genuinely ambiguous, correct the affected fields in the editor. ## Why does this happen? A PDF does not store a document. It stores instructions for painting glyphs at coordinates. There is no "this is column one, read it first" — that structure exists only in the visual arrangement. So a naive text extraction of a two-column CV reads across the page, producing lines that alternate between the sidebar and the main body: ```text Alex Morgan Operations Manager SKILLS Northwind Logistics Lean Six Sigma 2019 – Present SAP Led a team of 12 across three Stakeholder mgmt distribution centres, reducing ``` Parsed as written, that is unusable — skills interleaved with employment history. ## What does RemakeCV do about it? It checks. After extracting the text layer from a PDF, RemakeCV assesses whether the result is in a coherent reading order. If it comes back jumbled, the document is automatically re-extracted through OCR, which reads the page as a human sees it and recovers the intended order. Three things are worth knowing about that check: - **PDFs only.** A DOCX is never assessed and never falls back to OCR. In practice that matters less, because a Word file carries real structure — but it does mean a genuinely broken DOCX has no safety net. - **It runs only when enough text was extracted.** If a PDF yields fewer than 50 words, the low-text fallback fires first and the jumble check is skipped entirely. The two are alternatives, not both. - **It adds time.** The check runs on every qualifying PDF and typically adds a few seconds. That is often the answer to "why did this ordinary PDF take so long?" > **Note:** There is no "this CV is two-column" setting. Detection and fallback are part of processing. You can, however, force OCR yourself with the **Image Process** option — see [scanned PDFs and OCR](https://www.remakecv.com/help/troubleshooting/scanned-pdfs-and-ocr.md). ## The check is not perfect, in both directions **It sometimes flags a clean CV.** Roughly one in six flags is a false positive, sending a perfectly good document through OCR unnecessarily. If a clean PDF comes back looking oddly OCR-ish, that is the likely explanation — reprocessing usually resolves it. **It sometimes misses, silently.** If the check cannot reach a verdict — a timeout, or a service problem — processing continues with the original text rather than stopping. There is no warning. This is the main reason reprocessing the same CV a minute later can give a noticeably different result. > **Tip:** That silent-miss behaviour is worth remembering when a badly ordered CV appears out of nowhere on a document that worked last week. Reprocess before concluding anything about the source file. ## What if the output is still wrong? Some layouts are genuinely ambiguous, and no extraction strategy resolves them reliably: | Layout | Problem | |---|---| | Sidebar with no clear boundary | Which section does the sidebar content belong to? | | Timeline graphics | Dates positioned visually rather than adjacent to their role | | Tables used for layout | Cell order may not match reading order | | Text inside images | Not extractable text at all | | Overlapping decorative elements | Confuses both text extraction and OCR | For these, the practical path is: ### Reprocess once Free, and sometimes resolves differently. ### Correct the affected sections Usually only one or two sections are wrong. Fix those rather than starting over. ### Ask for the Word version A DOCX carries structure, so the intended order is far more likely to survive extraction. > **Tip:** Design-heavy CVs are common in marketing, design and creative recruitment, and they are the hardest category to parse. If you work those desks, making "please send the Word version" part of your standard candidate request removes the problem entirely rather than managing it. If content is missing rather than merely reordered, see [the formatted CV looks wrong](https://www.remakecv.com/help/troubleshooting/formatting-looks-wrong.md) and [checking what was extracted](https://www.remakecv.com/help/formatting-cvs/checking-the-extraction.md). ## Does the client see the original layout? No, and that is the point. The output uses your agency template regardless of how the source looked. A candidate's two-column creative CV becomes a standard branded document — the layout problem is an extraction concern, not something your client ever sees. ## Frequently asked questions ### Do I need to flag a two-column CV? No. RemakeCV checks whether the extracted text is in a coherent reading order and re-extracts via OCR automatically when it is not. You can also force OCR with the Image Process option. ### Why did a clean PDF come back looking OCR'd? The reading-order check occasionally flags a clean document — roughly one flag in six is a false positive. Reprocessing usually resolves it, and costs no credit. ### Why does a two-column PDF cause problems at all? A PDF stores positioned text fragments, not reading order. Extracted naively, a two-column layout reads across both columns line by line, interleaving unrelated content. ### Does the check apply to Word documents? No — PDFs only. A DOCX carries real structure so the intended order is usually recoverable directly, but it also means a Word file never falls back to OCR.