AI CV builder that starts from your PDF

The short answer

Yes. An AI CV builder from a PDF reads the CV you already have, pulls the facts out and writes a new CV from them. The risk is the middle step: a model left to improvise on top of your file adds numbers and titles you never had. CitedCV lists every fact it finds and writes only from the ones you confirm.

In CitedCV the import takes a PDF, DOCX or TXT file up to 5 MB, costs no credits and needs an account. It does not edit your PDF in place; it builds a new CV in one of eight single-column templates, and each later line edit costs 1 credit.

In the UK, Europe and Turkey the document is a CV (özgeçmiş in Turkish); in the US it is a résumé, and Harvard's career office describes a CV as the longer, mostly academic record. This page says CV for both.

The first minute: what to do and what it costs

Create an account, open Facts and choose your file. Reading it costs no credits, creating an account asks for no card, and a new account starts with 6 credits.

Run the audit first if you want a verdict

Audit scores the same file out of 100, flags lines a recruiter would skim past and shows the CV as a parser stores it. It calls no model, stores nothing, costs nothing and needs an account.

What costs nothing and what does not

How a PDF becomes a CV, in five steps

1. Your browser extracts the text: pdf.js for a PDF, mammoth for a DOCX, a TXT as it is.
2. The original and its text are stored, with control and zero-width characters stripped and the text cut at 30,000 characters.
3. A language model drafts the facts, copying each one word for word from your text; a fact not found in the file is dropped. Code, not the model, reads the numbers, dates and names.
4. You review and confirm. Every fact arrives as Unconfirmed and a build uses only confirmed ones; the builder says how many it skips.
5. You build. Each bullet carries the ids of its facts, a deterministic check compares every number, date and name with them, and a failing line is rewritten once, then dropped.

What the review step looks like

The Facts page lists each fact with its kind, its source and an Unconfirmed or Confirmed badge. Click a sentence to edit it, confirm one fact, delete it (the page asks twice) or confirm all; saving an edit confirms the line. Importing the same CV again adds only what is new. Careers services ask for the same discipline: MIT wants factual accomplishments, and Prospects says AI should not invent or embellish details.

What the import keeps and where your file goes

A fact is one statement from your CV: a role with employer and dates, a bullet, a skill, a degree. Facts keep the file's language; the new CV can be in another.

Where your file goes

The original and its text sit in private storage on Cloudflare until a nightly sweep deletes them 30 days after upload. The text also goes to a large language model provider, currently MiniMax, to draft the facts; the privacy policy says it may not use the data for its own purposes and that identifiable CV content does not train third-party models. Facts stay until you delete them or your account, plus up to 30 days for backups. GDPR Article 17 gives a right to erasure in the cases it lists; the Privacy page exports your data or deletes your account with your facts, CVs and uploads.

From LinkedIn: the same importer

CitedCV does not sign in to LinkedIn or fetch a profile from its address. Save your profile with LinkedIn's Save to PDF and import it from the From LinkedIn panel on the Facts page. LinkedIn's help page says the option is not on its mobile app and needs English as the profile language. CitedCV removes the page footers and bracketed durations LinkedIn adds, so "5 years 7 months" is not read as a number.

The numbers behind this page

• 5 MB: the largest file CitedCV accepts (CitedCV upload check).
• 2.5 MB: the size above which Greenhouse says it cannot parse a resume (Greenhouse Support, Unsuccessful resume parse, read 2026-10-03). Keep the file you send employers under it.
• 30,000 characters: the most text one import reads (CitedCV import code).
• 0, 6, 10 and 1 credits: an import, a quick build, a full build and a line edit (CitedCV pricing page).

No measured accuracy rate for the import exists in CitedCV's documentation, so this page quotes none.

Who this is not for

• You want to keep your own design and change a few words. CitedCV rebuilds in its templates; edit the DOCX yourself, or paste the text into ChatGPT, Claude or Gemini and check every line.
• You send one trusted CV to many similar postings. RecastCV keeps a master CV and rewrites it per posting.
• You need a two-column or design-led layout. Kickresume and Enhancv offer them; CitedCV has eight single-column templates.

Other ways to start from a PDF

As each tool's own pages described it on 2026-09-08 (our comparison pages); check before relying on them.

What the import cannot read or do

Scanned PDFs and CVs saved as pictures

A PDF with no selectable text, such as a scan or a CV exported as an image, is refused; CitedCV runs no OCR. Lever's help page gives the test: if you cannot highlight text with your cursor, the file is likely not parseable. Save it again as a PDF from Word, or upload the DOCX. A password-protected PDF is refused too.

Two columns, text boxes, tables and headers

Greenhouse names columns, graphics, tables, and contact details in a header, footer or text box as causes of a failed parse. In a DOCX, CitedCV counts text boxes, drawings, tables, columns and header text from the XML. In a PDF it detects columns, picture-only pages and pages full of filled shapes such as skill bars, and says so on the Facts page and in the audit. It finds columns from line positions; a layout it cannot tell from one column is read in the order the file stores it.

• Only PDF, DOCX and TXT, recognised by content and not name. An old .doc, a Pages file or an image must be saved as PDF or DOCX first; Lever lists seven file types, CitedCV imports three.
• A photo in your file is not carried over; add one on the Facts page, used on Turkish and German CVs on request and never on US or UK ones.
• No CV tool guarantees an interview, and Workday, Greenhouse, Lever, Taleo, SuccessFactors and iCIMS do not detect AI text.

Questions people ask

Sources

1. Greenhouse Support, Unsuccessful resume parsesupport.greenhouse.io/hc/en-us/articles/200989175-Unsuccessful-resume-parse
2. Lever Help Center, Understanding resume parsinghelp.lever.co/s/article/Understanding-Resume-Parsing
3. LinkedIn Help, Save a profile as a PDFwww.linkedin.com/help/linkedin/answer/a541960
4. MIT Career Advising and Professional Development, Resumescapd.mit.edu/resources/resumes/
5. Prospects, How to write a CVwww.prospects.ac.uk/careers-advice/cvs-and-cover-letters/how-to-write-a-cv
6. Harvard Mignone Center for Career Success, Create a resume, CV or cover lettercareerservices.fas.harvard.edu/channels/create-a-resume-cv-or-cover-letter/
7. Regulation (EU) 2016/679 (GDPR) on EUR-Lexeur-lex.europa.eu/eli/reg/2016/679/oj