Your resume passes two readers in sequence, and they are profoundly different. The first is software, which looks for text matches and understands no nuance. The second is a person with another twenty applications to get through, deciding in seconds whether to read past the first screen. Writing for only one of them means losing to the other.
This guide tries to be precise about how both filters work, separating what’s documented from what’s my own judgement. It also points out where the topic is overstated, because there are plenty of myths around it.
What an ATS actually does (and doesn’t)
ATS stands for Applicant Tracking System: the software companies use to collect, catalogue and search applications. The most common platforms in larger organisations are products like Workday, Greenhouse and Lever.
Let’s kill the biggest myth straight away: in most configurations the ATS does not automatically reject your resume. It is first and foremost a database. It does two things that genuinely affect you:
- It extracts the text (parsing) and maps it into structured fields: name, roles, dates, skills.
- It lets the recruiter filter and search hundreds of applications by keyword.
The difference matters. If parsing fails, your resume enters the system mangled or empty, and never shows up in the recruiter’s searches. You weren’t rejected: you became invisible. That’s a worse outcome, because it leaves no trace.
Some setups do include knockout questions on objective requirements: work authorisation, a required licence, willingness to travel. Those genuinely do exclude automatically, but they have nothing to do with the words in your resume.
Operating consequence: problem number one isn’t keywords, it’s whether the file can be read. Keywords come immediately after.
The formatting rules that actually matter
Ordered by how much risk they remove.
| Rule | Why |
|---|---|
| Single column | Multiple columns often get read in the wrong order, interleaving lines |
| No text inside images, text boxes or page headers | Many parsers don’t read them at all |
| Tables only if simple, better avoided | Merged cells are the most frequent cause of corrupted parsing |
| Standard headings: Experience, Education, Skills | The parser looks for known labels, not “My Journey” |
| Consistent date format (mm/yyyy) | Dates rebuild the chronology; mixed formats create gaps |
| PDF saved from a text editor, not a scan or image | An image-PDF is, to the software, a blank page |
| Standard fonts, and no icons replacing words | A phone icon is not a phone number |
One last thing that sounds trivial: the filename. Jane_Smith_CV.pdf can
be found again; cv_final_v3_last.pdf tells the reader something you didn’t
mean to tell them.
Keywords, done properly
The rule is simple: the job ad is the glossary. If the ad asks for “client portfolio management”, that exact string should appear in your resume, not “nurturing customer relationships”. It isn’t a matter of style: the recruiter’s search will use the ad’s own terms.
Two habits that raise your odds considerably:
- Acronym and full form together, at least once: “SEO (Search Engine Optimization)”, “GDPR (Regulation EU 2016/679)”. You don’t know which one they’ll search for.
- Skills inside the experience entries, not only in a closing list. A keyword that appears in context (“I ran the sales team’s CRM”) is credible; the same word in an isolated list isn’t.
And a hard limit: no hidden keywords. White text on white, invisible lists at the foot of the page, keyword stuffing. They barely worked before and almost never work now, and when they’re caught the application is over. It does happen, because the recruiter sees the text the system extracted, not your PDF. Not worth the risk.
The second reader: how much time you really have
There’s a much-quoted number here. An eye-tracking study commissioned by Ladders found recruiters spend an average of just a few seconds on the first pass over a resume, on the order of seven to eight seconds. It’s research commissioned by a company in the sector, so I take it for what it is: an order of magnitude, not a precise measurement.
But the order of magnitude is enough for the practical conclusion, which holds even if the figure were double: the first pass is won or lost in the top third of page one. Current role, company, one recognisable result. Everything below serves to confirm a decision that’s already leaning, not to make it.
The structure, line by line
Header. Name, target role (not “seeking opportunities”: the role you want), city, email, phone, link to your LinkedIn profile. A full home address serves nobody and hands over one more personal data point.
On photos: they’re still common in parts of Europe, but they aren’t required and in several countries they’re actively discouraged to reduce discrimination risk. The Europass format makes it optional. My advice: drop it, unless the role has a public-image component. It takes up prime space in the top third and adds no useful information.
Summary, two or three lines. Not a generic profile, but the answer to “why you for this role”. A formula that works: [role] with [X] years in [field], specialised in [the thing the ad asks for], with results in [measurable area].
Experience, reverse chronological. For each: role, company, dates, one line of context (sector and size, because “marketing manager” at an eight-person company and at an 800-person company are different jobs) and then the results.
Skills, selected against the ad. A list of twenty says nothing; six relevant ones say you read the posting.
Education and certifications, short. If you’ve been working more than five years, your degree is a line, not a section.
Results instead of duties: how to actually do it
It’s the most repeated advice in the field and it’s correct, but it’s rarely explained how to apply it when your job doesn’t produce obvious numbers.
The structure that works is: action + effect + measure.
❌ “Handled customer requests” ✅ “Reorganised the first-line support process, cutting average response time from 48 to 32 hours”
If you don’t have percentages, you still have something measurable:
- volume: how many cases, clients, people coordinated
- time: it took X before, Y after
- scope: budget managed, markets covered, team size
- verifiable qualitative outcome: “process later adopted by the other two branches”, “document became the internal standard”
An honesty rule that is also self-defence: every number you write must be something you can talk through in the interview. If you can’t explain where it came from, leave it out. That’s exactly what gets probed in interview questions.
Tailoring: how much it’s actually worth
Here I am flagging that this is my opinion, because I don’t know of solid public data quantifying the return on tailoring. The logic, though, points clearly one way: a few careful applications beat many identical ones, because the reader immediately recognises a text written for that posting.
There’s a rational explanation too: fifty generic applications each compete with hundreds of other generic applications, while one application aligned to the ad competes with a handful of aligned ones. The denominator changes.
The efficient way to do it is to keep a master resume with everything you’ve done, and produce a cut-down version for each ad: you change the summary, the order of skills, and three or four phrasings. Ten minutes, not two hours.
Using AI without sounding like everyone else
A generative tool is useful for three concrete things: comparing resume and job ad to list missing terms, rephrasing a weak sentence, and spotting where you wrote duties instead of results.
For generating the resume from scratch, though, it isn’t useful, it’s harmful. The output has a recognisable register, repeats the same formulas across every resume the recruiter gets that day, and above all it invents: if it produces a number you didn’t supply, that number is false and you’re signing it. The method for using it well, with verification at the centre, is in using ChatGPT at work and writing effective prompts.
The costliest mistakes
- Over two pages without an executive or academic profile: it signals no hierarchy in the information.
- Claims without evidence: “excellent interpersonal skills” isn’t information. Soft skills are demonstrated with an episode, not an adjective. On which ones actually count, see the most in-demand soft skills.
- Unexplained gaps: one honest line is worth more than a silence the recruiter will fill however they like.
- Typos: in a one-page document you should have reread ten times, they communicate something specific.
- One resume for everyone: it cancels out every other thing you did well.
Frequently asked questions
PDF or Word? PDF, unless the ad says otherwise. It preserves the layout and modern parsers handle it correctly. The only exception is portals that explicitly ask for .doc/.docx: follow the instruction, not the habit.
Is the Europass format fine? It’s fine in public administration and in many European contexts, where it’s sometimes required. In the private sector it’s often seen as rigid and padded. If it isn’t required, a single-column resume you wrote yourself communicates better.
Should I include my date of birth? It isn’t required and generally doesn’t help. Anti-discrimination practice points the other way: the fewer irrelevant data points, the better. Same for marital status and nationality, except where work authorisation is a genuine requirement.
How many pages for a senior profile? Two, and the second has to earn its space. Beyond two pages only makes sense for academic, research or executive roles with a list of publications or projects.
Do creative resumes work? In design and communication roles they can, if paired with a portfolio, but then it’s the portfolio doing the work. In every other case an unusual layout raises parsing risk and slows reading, which means it damages both filters at once.
Sources
- Europass. The European Commission’s CV tool, guidance on required and optional fields.
- Workday, Greenhouse and Lever. Public documentation from major applicant tracking systems on how parsing and search work.
- Ladders. Eye-tracking study on recruiter reading time (commissioned research; read as an order of magnitude).
- Regulation EU 2016/679 (GDPR). Data minimisation principle, applicable to the personal data included in an application.
- Italian Data Protection Authority. Guidance on processing candidates’ data during recruitment.



