A CV That Gets Read
What screening software actually does, why generated CVs read as forgettable, and how to use AI on the parts where it genuinely helps.
Two things are true at once, and job-search advice usually only tells you one of them. AI is very useful for writing a CV. And a CV written entirely by AI is worse than the one you would have written yourself.
The reason is specific: a model produces the most typical version of a CV for your role. Typical is exactly what does not get shortlisted when a recruiter is on their fortieth of the morning.
What screening software really does
Applicant tracking systems are mostly search and filing, not judgement. The mythology around them causes more bad decisions than the systems themselves.
- They parse your document into fields and index the text so a recruiter can search it
- They rank or filter on keywords and requirements a human configured — years of experience, a qualification, a location
- They do not have a mysterious score that rejects you for formatting flourishes, though genuinely broken layouts do parse badly
- White text stuffed with keywords is detected trivially and is treated as dishonesty when found
Where AI genuinely helps
- Turning duties into results. 'Managed social media' becomes 'Grew the company's Instagram from 2,000 to 11,000 followers in eight months' — you supply the numbers, it fixes the sentence
- Extracting the real requirements from a long posting, so you can see what to lead with
- Cutting length. Ask it to reduce a bullet to twelve words without losing the number
- Translating vocabulary across industries when you are changing field, so a hiring manager recognises the skill
- Catching the passive, hedging phrasing you stop seeing in your own writing
Where it makes things worse
- Writing your bullets from scratch. It will invent plausible achievements, and you will have to defend them in an interview
- The summary paragraph at the top. Generated summaries are uniformly bland — 'results-driven professional with a passion for excellence' is the most skippable sentence in recruitment
- Inflating scope. 'Led a team of engineers' when you coordinated two interns is a five-minute conversation away from collapsing
- Adding skills the posting mentions that you do not have. It gets you into the interview where it falls apart
A prompt that improves rather than fabricates
Here is a bullet from my CV. Rewrite it three ways.
Rules:
- Use only facts in what I give you. Invent nothing
- Lead with the outcome, not the duty
- Keep the number in
- Under 20 words
- Plain language, no "leveraged", "spearheaded" or "passionate"
If a number is missing that would make it stronger, ask me
for it instead of guessing.
Bullet: "Responsible for handling customer support tickets
and improving response times."
Facts: reduced average first reply from 14 hours to under 3;
handled about 60 tickets a week; alone in the role.The tailoring that pays off
You do not need a different CV per application. You need the top third to match the job. Paste the posting and your CV and ask which of your existing bullets map to the stated requirements — then reorder so those are first, and cut two that map to nothing.
Ten minutes per application, using only material that is already true.
What to take from this chapter
- Screening systems index and filter — layout sanity matters, formatting mythology does not
- Use AI to sharpen bullets you supply, not to write achievements you did not have
- Numbers you provide are what make a bullet strong; the model cannot supply them
- Skip the generated summary paragraph — it reads as filler to every recruiter
- Tailor by reordering real bullets to match the posting, not by rewriting everything
Try it
Take the three weakest bullets on your CV and run the prompt above on each, supplying real numbers. If you cannot supply a number for a bullet, that is a sign the bullet describes a duty rather than a result.