AI for the job hunt: 10 steps through the ATS filter

Most applications are never read by a human. Before your documents reach a recruiter, they run through an ATS, an applicant tracking system, that filters by keywords. If the CV doesn't fit the ad, it lands in the no pile without anyone ever having looked at it.

This is exactly where AI helps, but differently than most think. A chatbot writes you a nice-sounding CV. What it doesn't do: check whether that CV gets through the filter. A 2026 CareerBuilder analysis shows the difference brutally. CVs that come raw out of a chatbot pass the ATS in about 29 percent of cases. CVs that are specifically tailored to the ad reach around 71 percent. So the chatbot is your writing tool, not your checking tool. The checking you do yourself, with the process in this playbook.

The whole run-through costs you about 20 minutes per application. Sounds like a lot when you've got twenty positions in view. But it's the reason why nothing comes back for others and something does for you. Ten quick applications that nobody reads are worth less than three that get through.

Step 1: Read the ad as raw material, not as a wish list

Most people read a job ad to decide whether to apply. From now on you read it differently, namely as a list of the words that have to appear in your CV. Every skill, every tool, every qualification, every software name in the ad is a keyword that the ATS will later search for.

Copy the complete ad and throw it into your chatbot with this prompt:

Here is a job ad. Extract for me as a list all the hard
keywords an applicant tracking system would filter for:
skills, tools, software, certificates, qualifications, job titles.
Sort by importance. No explanations, just the list.

[paste the job ad here]

What comes out is your target list. You work it into the CV piece by piece in the next steps.

Step 2: Mirror the exact job title

That's the cheapest and most effective trick in the whole playbook. Jobscan analyzed 2.5 million CVs. If the exact job title from the ad appears somewhere in the CV, the probability of an interview is 10.6 times higher. Not 10 percent higher. Tenfold.

If the position is called "Junior Content Manager" and you were previously an "editor", then you put "Content Manager" in visibly, for example as a line under your name or in the short summary. Don't lie about experience, but name what you did with the word from the ad. The same thing often has three names. Take the one from the ad.

Step 3: Spread keywords across three zones

Writing a keyword in just once at the very bottom isn't enough. The ATS weights where a word appears. Spread your most important terms across three zones: the short summary at the very top, the skills block, and the individual bullet points at your roles.

An example. If "project management" is on your list, then it appears in the summary ("editor with experience in project management"), in the skills block as its own entry, and in a bullet at one of your roles ("project management for three parallel campaigns"). Three times, in three places, built in naturally. That way a human doesn't later read it as word spam, but the machine reliably finds it.

Step 4: Find the gap between ad and CV

Now you compare. Take your keyword list from Step 1 and your current CV and let the chatbot run them against each other:

Here is a list of required keywords and my current CV.
Tell me which keywords are MISSING from the CV. Don't invent
anything, only suggest where I could name existing real
experience with the right term.

Keywords: [list from step 1]
CV: [your CV]

The important part is the second sentence in the prompt. The chatbot shouldn't attribute skills to you, but show where you've only labeled real experience wrongly. Everything that's missing and that you really can do goes in now.

Step 5: Rewrite bullets into the employer's language

Your experience is already in there, but in your words. The ATS and the recruiter think in the words of the ad. The chatbot does this translation well. Take one bullet after another:

Rewrite this bullet point so that it picks up the terms from
the job ad. Keep the facts exactly the same, change only the
wording. No marketing speak, concrete and concise.

Bullet: [your original bullet]
Relevant terms from the ad: [2-3 keywords]

A sentence like "looked after the social media channels" becomes "content strategy and community management for four social media channels". Same truth, but now in the language that's being filtered for.

Step 6: Numbers in, but never invented ones

Bullets with a number beat bullets without a number, always. "Increased reach" is weak. "Increased reach by 40 percent in six months" is strong. Per role, three to six such result bullets belong in, each with a number if possible.

This is the point where you must not trust the chatbot. A chatbot likes to fill gaps with plausible-sounding numbers if you let it. It doesn't know that you did 40 percent, it guesses. Never let it invent numbers. The numbers come from your head or from old reports. The chatbot may build the wording around them, you supply the value.

Phrase this bullet as a result sentence with the number I give
you. Do NOT invent any additional numbers.

Activity: [what you did]
My real result: [your number]

Step 7: Switch to an ATS-friendly format

A pretty CV with two columns, graphics and a photo looks good to humans and like gibberish to the ATS. Many systems read multi-column layouts, text boxes, tables and content in headers or footers simply wrong or not at all.

Switch to a single-column layout. Standard headings like "Work experience", "Education", "Skills", no creative titles. No graphics, no tables, no icons, nothing in the footer. As a file format, a .docx file is the safest choice, because older ATS also read it cleanly. Modern systems handle PDF fine, but if you don't know which system the company uses, go with .docx.

Step 8: Write abbreviations both ways

The ATS is often dumber than you think. If a recruiter searches for "Search Engine Optimization" and your CV only says "SEO", the machine may find nothing. The other way round too.

Write out every important abbreviation once in both forms, ideally right after each other: "Search Engine Optimization (SEO)". That way you hit both search variants. This applies to all abbreviations in your field, from "CRM" to "KPI". Once spelled out, once as an abbreviation, then the machine is happy.

Step 9: Check the match rate, target 75 to 80 percent

Before you send, you check how well your CV now fits the ad. As a rule of thumb you want to reach 75 to 80 percent agreement on the important terms. Not 100 percent, that looks like blunt copying. But well above half.

Compare my CV with the job ad and estimate what share of the
important keywords from the ad appear in the CV. Name me the
keywords that are still missing and that I could honestly add.

Ad: [job ad]
CV: [your current version]

A quick note, because many fall for this. The 2026 generation of ATS can also match semantically, so it understands that "customer care" and "Customer Success" are related. Still, the rule holds: mirroring the term from the ad literally still performs better than rewriting. If the ad says "Customer Success", write "Customer Success", not "customer care".

Step 10: The cover letter last, and individual

Only once the CV is set does the cover letter come. It draws its keywords from the same list. Let the chatbot build a rough draft, but give it enough to work with so it doesn't sound like a template:

Write a short cover letter, maximum four paragraphs. Pick up
these three points from the ad: [3 most important requirements].
Connect them with this real experience of mine: [2-3 roles].
Tone: direct, no marketing speak, no clichés like "I hereby
apply for".

You then polish the rough draft yourself. The first paragraph must name the company and the specific position, otherwise it smells like a form letter. Two minutes of handwork here make the difference between "just another one" and "she made an effort".

What's next

This process works once in full and then per application as a quick repeat. You build your base CV cleanly once, after that you only adjust Steps 1, 2 and 5 for each new position, those are the 20 minutes per application.

If you want to go deeper into prompt work so the rewrite results from Steps 5 and 6 get better, look into Level 2 of the Academy. It's about system prompts and about how to give a model enough context that it stops guessing. And if you apply regularly, the playbook "Using AI productively in everyday life" is worth it, because you then set up your application process once as a repeatable routine instead of starting from zero every time.

Source

The facts and figures in this playbook were checked against public expert sources on the topic of ATS and AI-assisted job applications (as of July 2026). Numbers change, the method behind them, mirroring keywords from the ad, has been stable for years.

• ATS optimization and keyword-match process: matcha.ai/blog/ats-resume-tips-2026
• Jobscan analysis of 2.5 million CVs (exact job title, 10.6x interview chance): jobscan.co/blog/ats-resume
• ChatGPT in the application process, strengths and limits: aiapply.co/blog/how-to-use-chatgpt-for-a-resume

An honest note to close: no AI tool and no playbook guarantees you the job. What this process does is get your application past the automatic filter so that a human sees it at all. Everything after that, the interview, the impression, the fit, is up to you.