Recruiters convert a scanned CV to Word in one of two ways: manually — running the scan through OCR, then re-typing and restructuring the text into a template (typically 25-45 minutes per document) — or with AI CV formatting software, which reads the scan, extracts the content, and outputs an editable branded Word document in about 60 seconds. FormaCV does the latter for $0.99 per CV.
This guide covers both routes: why scanned CVs are harder than they look, where OCR goes wrong, and the workflow that gets you from a photographed or scanned document to a client-ready DOCX without an afternoon of re-typing.
Why are scanned CVs such a problem for recruiters?
A scanned CV is a picture of text, not text. Your ATS can't search it, your client's ATS can't parse it, and you can't edit it. Yet scans keep arriving: older candidates with a CV last updated on paper, international pipelines where documents are photographed on a phone, public-sector and healthcare candidates whose certificates and CVs live as PDF scans, and referrals forwarded as photos in WhatsApp or email. For an agency, an unreadable CV is a candidate who silently disappears — they don't surface in keyword searches and they can't be submitted to a client system. Before anything else can happen, the scan has to become real, structured, editable text. That conversion step — image to editable Word document — is what this post is about, and it's where most manual time gets burned.
How does OCR work — and where does it go wrong?
OCR (optical character recognition) software looks at the image and guesses which shapes are which characters. On a clean, flat, 300-dpi scan of a single-column document, modern OCR is genuinely good. Recruitment reality is rarely that kind. The common failure modes:
- Skewed or photographed pages — phone photos taken at an angle warp lines, and accuracy drops sharply
- Two-column layouts — OCR reads left to right across both columns, mashing the sidebar into the work history
- Tables and text boxes — content comes out in the wrong order, or not at all
- Low contrast and small fonts — 9pt grey text on a faxed page produces character soup
- Names and proper nouns — OCR fixes dictionary words but mangles surnames, employers, and qualifications, which are exactly the fields recruiters care about
The deeper problem: even perfect OCR gives you a wall of raw text. The document structure — sections, roles, dates — is gone, and rebuilding it is most of the work.
The manual workflow: scanned CV to Word by hand
If you only see a scanned CV occasionally, the manual route works:
- Re-scan if you can. Ask the candidate for the original file first — a surprising number of "scans" have a DOCX ancestor somewhere.
- Run OCR. Word itself opens PDFs and attempts conversion; Adobe Acrobat, Google Drive (open image as Google Doc), and free tools like OCR.space all extract raw text.
- Proofread every proper noun. Names, employers, job titles, qualifications, dates — check each against the original image.
- Rebuild the structure. Reverse-chronological roles, standard section headers, clean dates — the rules in our ATS formatting guide apply doubly here, because the output usually goes into an ATS next.
- Reformat into your agency template. Branding, fonts, layout.
Budget 25-45 minutes per document, more for photographed or multi-column originals. At one or two CVs a week, that's tolerable. At agency volume, it isn't.
What does the AI workflow look like?
With AI-based conversion the steps collapse into one. Upload the scanned PDF or photo to FormaCV, and the AI performs OCR, identifies the candidate's name, roles, employers, dates, skills, and education semantically, and restructures everything into your branded, ATS-safe template — output as an editable DOCX or PDF in about 60 seconds. Because the AI understands what a CV is, it recovers the structure OCR loses: it knows "2019 – 2023" next to "Operations Manager" is an employment period, even when the scan scrambled the column order. You review the result against the original image (still do this — especially names and dates), make any edits, and submit. The same pipeline handles a scanned resume from a US candidate, a LinkedIn export, or a plain DOCX — scans aren't a special case, just another input. At $0.99 per CV with no subscription, the economics work whether scans are 2% or 40% of your intake. This is AI CV formatting in practice: input-agnostic conversion into one consistent branded output.
Manual vs AI: when is each the right call?
Manual conversion makes sense when scans are rare, the document is short and clean, or you're dealing with a one-off format so unusual that you'd review every line anyway. AI conversion wins on everything else: volume (ten scanned CVs is an afternoon manually, ten minutes with software), consistency (every output lands in the same branded template), and ATS safety (the output is structured text that parses, not a patched-up OCR dump). A sensible hybrid rule for agencies: any scanned or photographed CV that's heading to a client goes through the AI pipeline; documents that are purely for internal record can wait. Whichever route you choose, never submit the raw scan to a client ATS — an image-based PDF either fails to parse or parses as an empty candidate, and it reflects on your agency, not the candidate. If you're comparing tools for this, our 12-tool comparison flags which formatters handle scanned input.
Quality checklist before you submit
Whatever produced the Word file, check five things against the original image before it leaves the building. Names and contact details — one transposed digit in a phone number costs you the candidate. Dates — OCR confuses 1 and 7, 3 and 8; a wrong start year misstates experience. Employer and qualification names — proper nouns are OCR's weakest area and the first thing a client googles. Completeness — multi-page scans sometimes drop a page silently; count the roles. Formatting hygiene — single column, standard headers, real text everywhere (select-all in Word; anything you can't highlight is still an image). Two minutes of checking protects both the candidate's chances and your agency's brand — and it's the part of the workflow that should never be automated away.
Frequently asked questions
How do I convert a scanned CV to Word for free?
Open the scanned PDF directly in Microsoft Word (it attempts OCR automatically), or upload the file to Google Drive and open it as a Google Doc. Both extract raw text from clean scans. You'll still need to proofread proper nouns and rebuild the structure and formatting by hand — that re-typing step is where most of the time goes.
Can FormaCV handle scanned and photographed CVs?
Yes. Scanned PDFs and photos are standard inputs alongside DOCX, TXT, LinkedIn exports, and call transcripts. The AI performs the text extraction and restructures the content into your branded, ATS-safe template in about 60 seconds, output as editable DOCX or PDF, at $0.99 per CV.
Why does my scanned CV come out as gibberish after OCR?
Usually one of four causes: the page was photographed at an angle, the original uses a two-column layout (OCR reads across both columns), the scan resolution is too low, or the text is low-contrast. Re-scanning flat at 300 dpi fixes most of it; for multi-column originals you need software that reconstructs structure, not just characters.
Will an ATS read a scanned CV?
Mostly no. An image-based PDF either fails to parse or imports as a near-empty record, so the candidate never appears in keyword searches. Some ATS attempt built-in OCR, but accuracy is poor and structure is lost. Convert the scan to a structured Word document before loading it into any ATS.
How long does it take to convert a scanned CV manually?
Typically 25-45 minutes: running OCR, proofreading every name and date against the image, rebuilding sections in reverse-chronological order, and reformatting into your agency template. Photographed or multi-column originals take longer. AI formatting software compresses the same job to around 60 seconds plus a short review.
Is OCR accuracy good enough to trust without checking?
No. Modern OCR is strong on clean scans but weakest exactly where recruiters need precision: surnames, employer names, qualifications, and dates. Always check proper nouns and numbers against the original image before submitting — whether the conversion was manual or AI-assisted, that two-minute review is non-negotiable.
Try it on your worst scan
Take the most battered scanned CV in your inbox and run it through a 30-day FormaCV free trial — no credit card. If the AI can turn it into a clean, branded, editable Word document in 60 seconds, the rest of your pipeline is easy. See pricing for the full $0.99-per-CV details.