Can a Better-Written CV Change Your Chances? Inside the 8% Finding

A small edit can change what a reader sees

Consider two descriptions of the same fictional task. One says, 'Responsible for various operational activities.' The other says, 'Checked supplier invoices against purchase orders and sent discrepancies to the finance team.' The second gives a reader something to assess. It is only better if it is true.

That distinction matters when using writing tools. A clearer account of existing experience and a newly invented achievement may both sound polished, but they are very different changes.

The experiment behind the headline

Emma van Inwegen, Zanele Munyikwa and John Horton studied a randomized writing-assistance intervention on an online labour platform. The experiment enrolled 480,948 new jobseekers in June–July 2021. Its main analysis of approved profiles with non-empty resumes included 194,700 people. The working paper was first posted in 2023.

The authors reported an approximately 8% relative increase in the probability of being hired. Table 9 puts the estimated change at 0.247 percentage points against a control-group hiring rate of 3.093%. In rounded terms, that is about 3.09% to 3.34%, not 3% to 11%. The reported 95% confidence interval for the relative increase was 3% to 13%.

These are group-level findings in a particular setting. They are not a calculator for an individual candidate's chances.

Read the primary paper, especially Section 3.6 and Table 9 →

What kind of assistance was tested?

MIT Sloan's account describes suggestions for spelling, grammar, punctuation, word usage, overused phrases, tone and style during profile creation. This was not an experiment in asking a chatbot to invent a candidate's career or generate hundreds of applications.

The platform served contract work and operated mainly in English, with many participants based in countries where English is not the native language. The researchers found no evidence that employers were less satisfied with assisted workers. That result supports taking writing help seriously, while leaving open how different tools and hiring markets would perform.

The university article and the paper describe the same experiment. They are not two independent replications. Neither establishes that a particular commercial CV product will improve your results today.

MIT Sloan's explanation of the intervention, April 2023 →

Use a tool as an editor with a limited brief

Our suggested workflow starts with a factual draft written from your own records. List the task, your contribution, the tools you really used and any result you can substantiate. You do not need a percentage improvement in every bullet. An accurate description of responsibility can be more useful than an impressive but unsupported number.

Then give the tool a narrow instruction: 'Improve grammar and clarity. Keep all dates, numbers, job titles and skill levels unchanged. Do not add achievements. If a detail is unclear, ask a question instead of filling it in.'

Compare the output with the original line by line. Watch for changes from assisted to led, used to mastered, or participated in to managed. Those are changes in meaning, even when the sentence becomes smoother.

Make the improvement visible without exaggeration

Here is an original illustrative example. Draft: 'Help with reporting and files.' Edited: 'Updated the weekly sales report and checked the spreadsheet for missing entries.' That edit is appropriate only if both tasks were actually performed.

A misleading version would add 'automated the reporting system and increased revenue by 30%' without evidence. Better grammar cannot make that claim legitimate. If a tool proposes a metric you have never measured, delete it.

Keep the language close enough to your own understanding that you can explain every sentence in an interview. For a practice project, say it was a practice project. For a team result, state your part in it.

A review you can finish before the next application

Read the finished CV beside the role description. Check whether the important requirement is answered with evidence, whether dates and titles agree across documents, and whether specialised abbreviations are understandable to the intended reader.

If possible, ask a trusted person to tell you what they think you actually did. When their interpretation differs from your experience, revise the wording. This is an editorial recommendation, not a method whose hiring impact was measured in the experiment.

Research reviewed on 20 October 2026. The study is a reason to take clear writing seriously, not a reason to outsource the truth of your application. Which CV sentence is hardest for you to explain in plain English?