We timed our own CV formatting pipeline, and it turned a raw CV into a branded, client-ready PDF in a median of about 3 seconds, between roughly 2 and 4 seconds across five different layouts. The branded PDF build itself took about 0.6 seconds every time. Most of the rest is the AI reading the CV.
"How long does it take to format a CV?" is a fair question to ask before you trust a tool to do it at volume, and like parsing accuracy, almost nobody answers it with a real number. This post does. We ran five real CV files through the exact pipeline RefineCV uses on the homepage demo, parse the CV then render it into a branded template, and timed each stage.
Key takeaways
- In our own test, RefineCV turned a raw CV into a branded, client-ready PDF in a median of about 3 seconds, and between roughly 2 and 4 seconds across five different CV layouts.
- The branded PDF build itself was the fast, steady part: about 0.6 seconds every time, whatever the layout. The variable part is the AI extraction, which ran from about 1.5 to 3.4 seconds depending on the CV and how busy the AI provider was.
- Speed did not depend much on how complex the CV was. The five-employer CV formatted in about 2.8 seconds, no slower than a simple one. The time is driven by the AI model, not by the length of the document.
- These are seconds, measured end to end. We are not going to put a minutes-per-CV figure on doing it by hand, because no independent study measures that and every number in circulation comes from a company that sells formatting software.
- The measured time does not include the human review step, which we recommend on every CV, or upload and network time. It is the raw pipeline: file in, branded PDF out.
The result
Across five CVs, RefineCV turned a raw file into a branded, client-ready PDF in a median of about 3 seconds, ranging from about 2.1 seconds for the sidebar CV to about 4 seconds for the dense one. The branded PDF build was a steady 0.6 seconds on every CV. Everything above that is the AI reading the document, and it is the only part that moves.
How we timed it
We timed the real pipeline, not a stopwatch on a screen recording. For each CV we ran the two stages the product actually runs: AI extraction (reading the CV into structured fields) and rendering (building the branded PDF from our template). We measured each stage with a monotonic timer, ran one untimed warm-up per file to load fonts and code, then took the median of three timed runs.
The five CVs are the same golden set we use for our parsing accuracy test: a clean single-column CV, a dense CV, a two-column CV, a CV with a sidebar, and a complex CV with five employers. They cover the layouts recruiters actually receive.
Two honest caveats. First, we ran this on a developer machine talking to the live AI providers, so the extraction time moves with provider load. On one run a provider was briefly rate-limited and a parse took noticeably longer, which is exactly why the AI stage is the variable part. Second, this is the raw pipeline time. It does not include uploading the file, the human review step we recommend before export, or any network round trip in a browser.
Time by CV layout
Here is every CV, split into the two stages and the total. Sorted fastest to slowest.
| CV layout | AI extraction | PDF build | Total |
|---|---|---|---|
| Sidebar (PDF) | 1.5 s | 0.6 s | 2.1 s |
| Complex (five employers, TXT) | 2.2 s | 0.6 s | 2.8 s |
| Simple (single-column PDF) | 2.6 s | 0.6 s | 3.2 s |
| Two-column (DOCX) | 2.8 s | 0.6 s | 3.3 s |
| Dense (one-column DOCX) | 3.4 s | 0.6 s | 4.0 s |
We publish our own speed and accuracy numbers because almost no formatting vendor does, and a checkable, first-party figure is worth more than a round marketing claim. This is five CVs measured on our own pipeline, not an independent benchmark, so read it as a transparent sample. For the companion accuracy test, see how accurate CV parsing is in practice.
What the numbers show
The number that surprised us was how steady the PDF build is. Rendering the branded document, cover page, snapshot table, every section, page frame and footer, took about 0.6 seconds on every CV, from the simple one to the five-employer one. That is the part we fully control, and it does not get slower as the CV gets longer.
The variation is all in the AI extraction, and it tracks the model and the provider, not the document. The five-employer complex CV extracted in about 2.2 seconds, faster than the dense single-column CV at 3.4 seconds. So "a longer CV takes longer to format" is not really true here. What moves the clock is how quickly the AI model responds, which is why we report a range rather than a single hard number.
Either way, the whole thing is seconds. The honest comparison is not to a precise by-hand figure, because no independent study measures how long a recruiter takes to reformat a CV by hand, and every number you see quoted comes from a vendor that sells the alternative. The point stands without one: a pipeline that runs in seconds changes what you can do at volume, and it leaves the recruiter's time for the part that needs judgement, reviewing and correcting the output before it goes to a client.
Frequently asked questions
How long does it take to format a CV?
With automated CV formatting software, seconds. In our own test, RefineCV turned a raw CV into a branded, client-ready PDF in a median of about 3 seconds, and between roughly 2 and 4 seconds across five layouts. The branded PDF build itself was about 0.6 seconds; the rest is the AI reading the CV. By hand, the honest answer is that no independent study measures it, and the figures vendors quote range from about 15 to 45 minutes.
Why does automated formatting time vary between CVs?
Almost all the variation is in the AI extraction step, which depends on the AI model and how busy its provider is at that moment, not on how long or complex the CV is. In our test the PDF build itself was a steady 0.6 seconds on every CV, while extraction ranged from about 1.5 to 3.4 seconds. A five-employer CV was not slower than a simple one.
Does the measured time include reviewing the CV?
No. The time we report is the raw pipeline: file in, branded PDF out. It does not include uploading the file, the human review step we recommend on every CV before export, or the network round trip in a browser. Formatting is fast; the recruiter should still check the parsed fields before the pack goes to a client.
Is automated CV formatting faster than doing it by hand?
Yes, by a wide margin, though we will not attach a precise manual figure to the claim. No independent study measures how long a recruiter takes to rebuild a CV in Word, and every number in circulation (commonly 15 to 45 minutes) comes from a company that sells formatting software. What we can measure is our own pipeline, which runs in seconds. The time you save is best measured on your own desk against your own template.
Does formatting speed matter more than accuracy?
No, accuracy matters more, and speed is only useful if the output is right. We publish both: this speed test and a separate field-by-field accuracy test where our parser averaged about 99 out of 100. Read them together. A fast formatter that garbles the data just lets you send a bad CV faster.
The bottom line
Automated CV formatting is a seconds-long step, not a minutes-long one. In our test the branded document was built in about 0.6 seconds and the whole pipeline ran in a median of about 3 seconds, with the AI extraction being the only part that varied. That speed is what makes formatting at volume practical, and it leaves the recruiter's time for the judgement step: reviewing and correcting the output before it reaches a client.
Related reading: how accurate CV parsing is in practice, the ROI of faster CV formatting, and our guide to the best CV formatting software for recruitment agencies.
Time it yourself on your worst CV
Upload a candidate CV and watch RefineCV build a branded, client-ready pack in seconds. Start free with 10 CVs, no credit card. Then $0.40 per CV, or $50 per month for 200.
Sources
- RefineCV formatting-time measurement (2026): First-party test described in this post: five golden CVs run through the live demo pipeline (AI extraction plus branded PDF render), median of three timed runs per CV.