We tested CV formatting tools on an awkward synthetic CV. Here's what we saw

On 10 October 2026 we tested recruiter-focused CV formatting tools through free accounts and public demos, using one synthetic CV built to be awkward. No tool was flawless, and the differences that mattered weren't about looks. They were about how each tool handled small qualifiers like "degree not awarded", whether contact removal reached the exported files, and how much you can try before you pay. This isn't a ranking of the best CV formatting tools tested on one CV, because one CV can't support a ranking. It sets out what we saw, how we tested, why we don't attach brand names to individual errors, and how to run the same test yourself.

Best CV formatting tools tested: the short version

• Qualifiers were the weak spot. Of the three tools that processed our CV, two showed a withdrawn MSc without the withdrawal on the first pass, so it read as completed. One kept it. Both tools that dropped it restored it after a correction.
• A vendor's own demo showed an education entry that ended after two days as "1995 – Present".
• Contact removal varied by route. One public anonymiser removed email and phone but left the name and address. The same vendor's main product passed our limited identifier checks in both PDF and Word.
• Mapping created new facts. "Basic conversational Spanish" became a formal language level, and a location from one part of the CV was attached to a job that had none.
• Keyword scores were fragile. "No Python experience" counted as a Python match, and deleting one full stop moved a score from 67% to 100%.
• Recovery features were good where we tried them, including Undo and Redo and regenerating after a correction.
• Trial access ranged widely, from ten free exports with no card to sales-led routes we couldn't enter.

How we tested

Who and when. The CVPitch team, on 10 October 2026. We make a competing product, which is a good reason to read this with care and a good reason for us to be strict about method.

The test CV. A fictional candidate with two jobs with exact dates, a six-person reporting line, a documented pilot that cut processing time from 12 days to 9, an MSc marked "Withdrew in May 2018; degree not awarded", a completed BSc, three skills, "basic conversational Spanish", and fictitious contact and referee details. No real candidate data was used anywhere.

Access. We screened eleven recruiter-focused tools and went deeper on six. Four of those allowed a free account; others had public demos or free tools. We used no payment details, submitted no sales or demo forms, and didn't create extra accounts to get around limits.

What we checked. The on-screen result, and wherever possible the exported files. We rendered and viewed PDF pages, extracted the text, and searched Word files' internal XML for our test contact strings.

How we labelled results. Verified (we did it and saw the result), Issue (the result lost or changed information), Inspected (we saw the controls but didn't complete the action) and Gated (it needed a paid plan, a sales contact or access we didn't have).

Why we don't name tools beside individual errors

Every observation below comes from one synthetic CV, on one day, at whatever depth each tool's free access allowed. These products change weekly. Attaching a brand name to a single-sample error would read like a verdict, and the evidence doesn't support a verdict. So we refer to the tools we tested as Tool A to Tool F. What matters for your evaluation is the type of problem, because you can test for it on any tool.

What we found

Facts: qualifiers are where meaning changes

• Tool A (a public editor) extracted both jobs, dates, the reporting line and the 12-to-9-day claim correctly. Its first pass showed "MSc Operations Research, 2017–2018" without the withdrawal. When we asked it to add the wording, it did, and Undo and Redo removed and restored the fix cleanly.
• Tool B (a free account) kept "Withdrew in May 2018; degree not awarded" through parsing and into the PDF export.
• Tool C (a free account) omitted the withdrawal from its first PDF and Word files. After we corrected the field and regenerated, both formats kept the correction along with the employment facts we checked.
• Tool D (a public demo using the vendor's own sample CV, not ours) showed a course the source said ended after two days as "1995 – Present", without the withdrawal.

Mapping, derivation and association

Tool B converted "basic conversational Spanish" into "Elementary proficiency (A2)", a level the source never stated. It showed a total of "6 years" where the source had only date ranges, and it attached a city to a job whose entry had no location (the city appeared elsewhere in the CV). In the vendor's own demo profile, a "10+ years" skill duration appeared as "10 yrs" in the exported PDF. None of these is outrageous on its own. Each is a statement the candidate didn't make.

Contact removal

• Tool A: a chat request cleared email and phone from the visible fields, and Undo restored them. We couldn't export a file to check further.
• Tool B: on the vendor's demo profile, with phone and email switched off, neither value appeared in the PDF's extracted text or the Word file's XML in our checks. The candidate's name and location stayed. That's visibility control, not anonymisation.
• Tool C: the public, account-free anonymiser replaced emails and phone numbers but left the candidate's name and street address. It also turned the "References" heading into "[REDACTED NAME]" while the referee's actual name remained. The main product, in a free account, behaved differently: nine test identifiers were absent from the PDF text and from all the Word file's XML parts we checked.

Our contact checks were string searches, not a full forensic review of images, annotations and every hidden surface.

Layout and recovery

Switching Tool A to a different layout preset showed "Unnamed candidate" and placeholder "Start date / End date" fields in the preview, and Undo brought the data back. Tool D's demo output had an empty "Nationality" row and a sparse final page. Tool B's demo PDF split a job's last bullet onto the next page, away from its heading. Tool C saved and reloaded a brand kit, applied renamed headings and a new section order, and regenerated correctly.

Keyword matching

Tool C's free keyword checker counted "No Python experience" as a Python match. With the same text, deleting only the full stop after "Spanish" moved the score from 67% to 100%. Keyword presence isn't evidence of skill, and a confident percentage can hide that.

Exports and deliverables

Tool C produced PDF and Word files from our CV. Tool B produced a PDF from our CV, and both PDF and Word from its demo profile. Tool D's public demo produced a PDF. Tool A's export sat behind a "contact us" step. Tool E offered a broad template gallery and a custom builder, but saving a custom template needed a paid plan. Tool F offered a free PowerPoint shortlist-deck template to download, a useful reminder that some agencies want more than a CV. We didn't test an AI-generated deck.

Trial access: what you can do before paying

Tools we didn't test

The market is bigger than our sample. It includes vendors we didn't test, for example CVFormatter, RemakeCV and iReformat. We didn't get hands-on access to their products in this round, so we make no statement about how they perform. It also includes formatting features built into some ATS platforms, which we didn't cover. For a wider overview, see our CV formatting software comparison.

What about CVPitch?

We didn't run CVPitch through this session, so we make no claim that it would have done better on this CV. What we can tell you is how it's designed to deal with these failure types. Every extracted fact sits beside its source passage, uncertain items are flagged, recruiter additions are labelled, and approval is tied to the exact revision. Contact removal covers the generated DOCX, PDF and hidden metadata. Optional AI writing tools return proposals you accept change by change, the editor has undo and redo, and version history can restore an earlier revision. Job match marks each requirement as evidenced, partly evidenced or not shown in the CV, without a fit score. You can check all of it yourself: the free trial gives you 10 exports with no card, and the guide shows the workflow.

Run the same test yourself

Make a synthetic CV containing these lines, run it through each tool you're considering, and check the exported files rather than the preview:

• An MSc with "withdrew, degree not awarded".
• A role with only a start year and no end date.
• "10+ years" of a skill and "basic Spanish".
• "No Python experience".
• A location given for one job only.
• A numeric claim, such as "cut processing time from 12 to 9 days in a pilot".
• Contact details in the header, the footer and a hyperlink, plus a referee's name and email.

Record what went missing, what changed and what appeared from nowhere, separately from layout issues. The buyer's guide has the full trial method, and AI CV formatting accuracy explains why these lines catch problems. To check contact removal by hand, use our contact removal guide.

Limits of this test

• One synthetic CV, tested on one day. That's not an accuracy rate, and it's not a ranking.
• Access depth varied. Some tools we exercised fully in a free account, others only through public demos.
• Products change. Something we saw on 10 October may already be different.
• Contact checks were string searches in extracted text and XML, not a forensic review.
• We didn't time workflows or measure speed.
• We make a competing product. We've tried to report neutrally, and we've included what tools did well.

Frequently asked questions

What is the best CV formatting tool?

It depends on your clients' formats, your volume and which errors you can't afford. On our single test CV, no tool was flawless. Test shortlisted tools on your own awkward CVs and judge the exported files.

Did you test CVPitch the same way?

No. CVPitch wasn't part of this session, so we make no comparative claim. You can run the same synthetic test on CVPitch in the free trial.

Why don't you name the tools in the findings?

Each finding comes from one synthetic CV on one day, and products change quickly. Naming tools beside single-sample errors would suggest a verdict the evidence doesn't support. The types of problem are what you should test for.

How can I test CV formatting tools myself?

Build a synthetic CV with a withdrawn degree, a missing end date, a negative skill statement, a language level, contact details in headers, footers and links, and a numeric claim. Run it through each tool and check the exported Word and PDF files.

Bottom line

Good-looking output tells you very little. Test qualifiers, mapped values and contact removal on the exported files, read the trial terms, and choose the tool whose mistakes you can see and fix before the client does.