Getting More Interviews: Real Results From AI-Customised CVs

Does a CV edited with AI actually get you further than the one you wrote on a Sunday night three years ago? It is a fair question, and most of the answers online are marketing dressed up as research.

There is real evidence, though. One large randomised trial, published through the US National Bureau of Economic Research, and one very large dataset from a CV tool vendor. Neither is perfect, and both are worth reading carefully rather than quoting as a headline.

Here is what they found, what they did not find, and what that means for a UK job search in autumn 2026.

The Strongest Evidence: A Trial With Nearly Half a Million Jobseekers

In 2021, economists Emma van Inwegen, Zanele Munyikwa and John Horton ran a field experiment on a large online labour market. As 480,948 new jobseekers created their profiles, which act as the CV on that platform, half were randomly given algorithmic writing assistance and half were not.

Because the groups were randomly assigned, this is about as close to cause and effect as hiring research gets. The treated group also formed 7.8% more contracts overall, and their hourly wages in the jobs they won were 8.4% higher.

Two findings matter as much as the headline. First, the treated jobseekers did not send more applications, so the gain came from the document rather than extra effort. Second, employers were not later disappointed with the people they hired. The researchers found no evidence that better writing tricked anyone; it seems to have helped employers see ability that was already there.

What the study did not test

Read the method before you repeat the number. The assistance was a writing checker that flagged spelling, grammar and phrasing errors. It was not a generative AI rewriting CVs, and it did not tailor anything to a specific job. The trial also ran on a freelance platform, not in UK corporate recruitment.

The gains were concentrated among the weakest writers. People whose profiles were already clean had little to gain. And the paper measured hires, not interviews; on that platform the number of interviews given did not measurably change, which the authors link to how new entrants appear in employer searches.

So the honest summary is narrower than most blogs make it: removing errors and making a CV easier to read causes more hires, especially for people whose writing was holding them back.

What Jobscan’s Data Adds

The second source is Jobscan, a US CV-checking tool. Its State of the Job Search 2025 report analysed more than 2.5 million job applications made through its tools.

That is a striking figure, and it needs a health warning. This is observational data from a vendor’s own users, not an experiment. People whose previous title already matches the vacancy are often simply closer fits for the role, so the title alone is not doing all of that work.

The same report surveyed 384 recruiters. Of those, 55.3% said they use job titles to filter applicants in their applicant tracking system, and 55.32% said a CV tailored to the job description makes a candidate stand out. That is a recruiter preference, not a measured outcome, but it points the same way as the trial.

What “Tailored” Means in Practice

Put the two sources together and a practical picture emerges. The trial says clarity and correctness pay. The Jobscan data says alignment with the specific vacancy matters. Neither says that stuffing a CV with keywords, or letting AI invent achievements, will help.

For a UK application, tailoring usually means four things:

• A title the sifter recognises. If your employer called you “Customer Success Lead” and the advert says “Account Manager”, say both, honestly, in your headline or role line.
• Relevant evidence first. Reorder bullet points so the two or three that answer this advert sit at the top of each role.
• The advert’s vocabulary, where it is true. Our guide to mirroring job description language covers how to do this without copying the advert back at the reader.
• Clean, error-free writing. The one change with randomised evidence behind it, and the one AI does most reliably.

The line to hold is between wording and facts. AI is good at rephrasing what you did so a stranger can follow it. It should never add experience you do not have, and recruiters are alert to keyword matching that looks forced.

Realistic Expectations for a UK Job Search Now

Context matters. The ONS estimated 702,000 UK vacancies in June to August 2026. Outside the pandemic period, the last time vacancies were that low was August to October 2014. More candidates are competing for fewer openings, which makes a CV that is easy to read and obviously relevant more valuable, not less.

But an 8% lift in hiring probability is a nudge, not a guarantee. If your chance of landing a particular role was modest, a better CV makes it a little less modest. The compounding effect comes from applying that improvement consistently across every application you send.

Tailoring also takes time, and that is the honest cost. A tailored CV for every application is more work than one generic version. The research suggests it is worth it; the practical question is how to make it sustainable across twenty or thirty applications. Keeping the result in your own voice matters too, because the interview will be with you, not the CV.

A Faster Way

Tailoring by hand works. Reading the advert, adjusting your title line, reordering bullets and checking every sentence for errors is exactly what the evidence supports, and plenty of people do it well with a notepad and patience.

Zappli does the repetitive part for you: it reads the job description against your CV, shows where the two line up and where they do not, and drafts a tailored version from your real experience. The free tier needs no card and no account. You get an ATS score, plus three CVs and three job matches, a lifetime allowance rather than a trial.

If you are applying in volume, there are passes: 7 days for £9.99 with up to 25 applications, 30 days for £24.99 with up to 100, or 90 days for £49.99 with up to 250. One payment only, never auto-charged, and nothing to cancel. Zappli is paid by candidates, never by employers.

If a free diagnosis sounds more useful than tailoring every version by hand, pick whichever of these is closest to where you are.