# One Resume Per Job Is Now the Standard, and Not for the Reason You Think

LinkedIn now takes roughly 11,000 applications a minute. Workday told a federal court its software rejected 1.1 billion of them. Here is what actually screens you out, and a free prompt that builds a tailored resume without inventing anything.

Author: J.A. Watte
Published: August 9, 2026
Source: https://jwatte.com/blog/one-resume-per-job-ats-standard/

---

For about twenty years, the sensible advice was to keep one good resume, tweak the top line, and send it everywhere. That advice was correct. It stopped being correct somewhere in the last two years, and most of the writing about why is wrong.

The wrong version says a robot scores your resume and throws it out if you miss the keywords. That story has been repeated so many times it feels like knowledge. It is mostly folklore, and I will come back to where it came from, because clearing it out is the first step to seeing what is actually happening.

The real version is duller and more useful. Three things changed at once, and each one on its own would have been survivable.

## What actually changed

**The volume went vertical.** LinkedIn told CNBC in October 2025 that applications submitted on the platform had risen more than 45 percent in a year, running at roughly 11,000 a minute. A LinkedIn representative put the figure at nearly 9,500 a minute in a separate account, so treat the exact number as approximate and the direction as certain. Recruiters describe it as drinking through a fire hose. Popular roles collect hundreds of applications in the first seventy two hours.

**A machine layer got large enough to matter legally.** In *Mobley v. Workday*, a case in the Northern District of California, Workday represented in its filings that 1.1 billion applications were rejected using its software during the relevant period. In May 2025 Judge Rita Lin granted preliminary certification of a nationwide collective under the Age Discrimination in Employment Act, covering applicants aged 40 and over screened through Workday's tools since September 2020. Whatever you think of the merits, the number is the point. That is the scale of the layer sitting between a person and a hiring manager.

**Everyone got the same writing tool.** This is the shift almost nobody plans for. When every applicant can generate a competent, well organized, keyword aligned resume in ninety seconds, a competent, well organized, keyword aligned resume stops being a differentiator. It becomes the floor. And because most people prompt the same models with the same job description, the output converges. Recruiters now read the same document over and over with different names on it.

Those three together produce the situation we are in. The old strategy, one strong general resume, was optimized for a world where writing quality was scarce and volume was low. Both of those assumptions are gone.

## First, the folklore that has to go

You have read that 75 percent of resumes are rejected by applicant tracking software before a human sees them. It is on hundreds of career sites. It is in LinkedIn posts every week.

It traces back to a 2012 sales pitch from a company called Preptel, which sold resume optimization services. There was no published methodology, no disclosed sample, no study. Preptel shut down in 2013. The number survived because a 2014 article cited the pitch, later articles cited that article, and nobody walked back up the chain.

I want to be careful here, because debunking a number is not the same as saying the problem is imaginary. Parsers do mangle resumes. Recruiters do search inside these systems and do miss people who are in there. What is not supported is the specific mechanism everyone believes in: a score, a threshold, and an automatic bin.

A 2025 survey by Enhancv of recruiters across more than ten ATS platforms found the large majority did not configure their systems to auto reject based on content or formatting at all. Small sample, so hold it loosely. But it points the same direction as the vendor documentation, which describes ranking, filtering, and grading rather than deletion.

## So what actually screens you out

Four things, roughly in order of how often they bite.

**The screening questions.** Work authorization, years of experience, willingness to relocate, salary expectation, degree, clearance, license, shift availability. These are structured fields with structured answers, and they are the part of the application that genuinely disqualifies people automatically. Candidates agonize over resume verbs and then answer a dropdown carelessly. The dropdown is what rejects them.

**Not being findable.** Recruiters search inside the system. If your resume parsed badly, or your titles are internal jargon nobody searches for, you are in the database and invisible. This is where formatting actually costs you, and it costs you silently.

**Not being read.** With three hundred applications, the pile gets sorted somehow. Workday, for example, acquired HiredScore in 2024 and folded it in as a grading layer. HiredScore's own help documentation notes that a candidate can go ungraded if the CV arrives in a format it does not support. An ungraded candidate is not rejected. They just are not in the sorted list the recruiter is working down, which is worse, because nothing ever tells you.

**Reading as interchangeable.** If your document looks like the other two hundred generated documents, the six seconds a recruiter gives it are spent confirming that it looks like the other two hundred.

Notice that none of these is a keyword score. Three of the four are about being legible to the system and to the person. One is about the answers you type, not the file you upload.

## The real reason to tailor, which is not keywords

Here is the thing the keyword framing gets backwards.

Tailoring a resume is usually described as adding: put their words in, match their phrasing, raise your match score. That part is real but small. The expensive part of tailoring is **deciding what to leave out**, and that decision is different for every employer.

I rebuilt a resume last year for someone moving from defense work into corporate security engineering. The hardest, most technically impressive decade of his career was coalition intelligence networking. Genuinely difficult work, and to the company he was applying to, completely meaningless. A resume that led with it read as a mismatch, not as a strong candidate in the wrong clothes.

So nine bullets became two. The endpoint hardening and vulnerability work, which he considered routine, moved to the front and got the space. Nothing was invented. Nothing was even reworded much. The document simply argued for a different identity, built from the same facts.

You cannot do that once and reuse it. There is no universal strong resume, because the thing you sacrifice is different every time. That is the actual argument for one resume per job, and it survives even if every parser in the world got perfect tomorrow.

## The sameness problem, and how to beat it

This is the part I would most want a friend to understand, because it is where the advantage is now.

Generated resumes have a shape. Once you see it you cannot unsee it, and recruiters see it forty times a day:

- Bullets that are all about the same length, because the model balances them
- Achievement sentences built from three or four templates, most famously "Spearheaded X, resulting in a 30 percent improvement in Y"
- Round numbers everywhere. Fifty percent. Ten times. A hundred plus
- A competency block of abstract nouns with no systems named inside it
- Nothing ever going wrong. Every project on time, no constraints, no trade-offs
- A summary that would be equally true of any competent person in that field

Survey data on how employers respond is noisy and mostly comes from resume vendors, so I will not pretend the percentages are solid. The consistent finding across them is worth repeating anyway: the objection is not to AI assistance. It is to output that is not personalized. Specific, tailored detail reads as someone who actually wants this job.

The way to beat the shape is not better prose. Everyone has better prose now. It is access to detail that a model cannot produce because it does not have it.

Three tests. Run every bullet through them.

**The swap test.** Put another qualified person's name at the top of your resume. Does the bullet become false? If it stays true, it says nothing about you specifically. That is the definition of interchangeable, and it is the single most useful edit in this whole process.

**The "how do you know" test.** Imagine an interviewer stopping you and asking how you know that. If there is no answer, the claim is decoration and it will hurt you later.

**The specificity floor.** Does the bullet contain at least one thing only someone who was there would know? An exact figure. A version number. A constraint. A thing that failed. If not, it was written from the outside looking in.

Concretely, this beats:

> Improved deployment reliability across the enterprise.

with this:

> Found that the deployment tool silently accepted invalid hardening configuration and reported a successful install while enforcing nothing, then moved the settings to a policy hive that actually enforces and that scanners read.

The second one is longer, less grand, and impossible to generate, because it requires having been in the room. It also gives the interviewer something to ask about, which is the entire point of a resume.

One more thing about numbers, since this is where AI assistance does the most damage. Precise numbers are the most dangerous claims on a resume. Their specificity reads as evidence, which is exactly why they invite the follow up question that exposes them. A language model, asked to make a bullet stronger, will produce a plausible number rather than admit it does not have one. If you take nothing else from this post: **check every digit on your resume against something real, and delete the ones you cannot source.**

## Nobody is coming to fix this for you

It is tempting to assume regulation will sort out the machine layer. Over the past year it has moved the other way.

New York City's Local Law 144 requires annual independent bias audits of automated employment decision tools and candidate notice, and has been enforceable since July 2023. In December 2025 the New York State Comptroller published an audit of how the city enforces it. Reviewing the same 32 companies, the city's Department of Consumer and Worker Protection identified 1 violation. The state auditors identified at least 17 instances of potential non compliance. The report also found the complaint intake process was not reliably routing complaints to the department at all.

Colorado passed the most ambitious state AI law in the country, SB 24-205, then delayed it, then repealed and replaced it. Governor Polis signed SB 26-189 on 14 May 2026. The replacement drops the duty of care around algorithmic discrimination in favour of a disclosure framework for automated decision making technology, and it does not take effect until 1 January 2027.

In the EU, recruitment and candidate evaluation systems are classified as high risk under the AI Act. Those obligations were meant to apply from 2 August 2026. Under the Digital Omnibus agreed in May 2026, they are pushed to 2 December 2027, with AI embedded in regulated products moving to 2 August 2028.

None of this is an argument against the laws. It is an argument about timing. The screening layer is operating at a scale measured in billions of decisions right now, and the rules meant to govern it are eighteen months out at best and thinner than they were when drafted. Practically, you adapt to the system as it exists.

## The method, compressed

This is the process I have used on real rebuilds. Nine stages, and the order matters more than any individual step.

1. **Fact load before writing anything.** Old resumes, LinkedIn export, project docs, and a long interrogation of the person. Write badly and completely rather than well and thinly.
2. **Ask what they have not done.** Go through the posting line by line and write down every requirement they genuinely lack. This feels like arguing against yourself and it is the highest leverage question in the process, because a named gap becomes a writing problem instead of an interview ambush.
3. **Read the whole posting.** Every word, including the About the team paragraph, which is often where the real job hides. Extract it into a matrix: have it and can prove it, honest analog, or hard gap.
4. **Identify the system.** The apply link hostname tells you the ATS. It takes five seconds. More on this in a moment.
5. **Decide the positioning in writing, before drafting.** Two columns: what moves up, what gets compressed or cut. Show it to the person so they can object.
6. **Draft against the matrix,** not against a keyword list. What was broken, what you built, what happened after.
7. **Verify adversarially.** Turn on your own draft and try to break it. Every number needs a source. Every verb has to match what actually happened. On one build this killed an impressive, precise, entirely unverifiable figure that had come from a single approximate note.
8. **Verify the file by machine, not by eye.** Zero tables, zero text boxes, zero images, nothing in the header or footer, one font, and a round trip extraction that proves your name, phone, email, current title, and degree all survive.
9. **Write the gap memo.** Not a resume. An uncomfortable document listing what you cannot claim, what you will be asked, and what to say.

That last one is the step everyone skips. The resume gets the interview. The memo is what survives it.

## Finding out which system you are submitting into

Look at the URL behind the Apply button. It almost always names the vendor:

- `myworkdayjobs.com`: Workday, usually as `company.wd5.myworkdayjobs.com`
- `job-boards.greenhouse.io` or `boards.greenhouse.io`: Greenhouse. Anduril, for example, applies at `job-boards.greenhouse.io/andurilindustries`
- `jobs.lever.co`: Lever
- `jobs.ashbyhq.com`: Ashby, common at venture backed startups
- `taleo.net`: Oracle Taleo, still widely deployed at large employers
- `icims.com`: iCIMS
- `jobs.smartrecruiters.com`: SmartRecruiters
- `successfactors.com`: SAP SuccessFactors
- `apply.workable.com`: Workable
- `usajobs.gov`: US federal, and a completely different set of rules

That last one deserves its own warning, and it is where I have to correct something I believed until this week. Federal hiring was re-plumbed twice inside twelve months, and almost every federal resume guide online now describes a system that no longer exists.

The old advice was to write long. Five pages, every duty, every detail. Under OPM's Merit Hiring Plan of 29 May 2025, **the federal resume is capped at two pages**, and the USAJOBS resume builder enforces it. A stored five page resume from before the change will block you from applying. The same memo phased out self assessment questionnaires for rating and ranking, which used to be the thing everyone gamed, and replaced them with at least one technical or alternative assessment. Then the "Rule of Many" final rule, effective 7 November 2025, brought numerical scoring back as an option alongside category rating.

What did not change is the part people skip: every job block needs employer, title, **month and year** start and end, **hours per week**, and for federal jobs the pay plan, series and grade. Specialized experience is counted in months at a defined level, so a year only date range is one of the most common reasons a qualified person is rated ineligible.

Read the "How you will be evaluated" section of the announcement before you write anything. Whether you are being sorted into two quality categories or ranked into a top ten percent is a different writing problem, and the announcement has to say which.

Knowing the platform changes practical things. Workday's own documentation, for instance, states plainly that resume parsing does not fill the Skills and Languages fields. Those are manual, always, no matter how good your resume is. An empty Skills section makes you unfilterable in the exact tool the recruiter uses to narrow the pool. That is the vendor talking, not a blog.

It is less overwhelming than the list makes it look, too, because the systems share plumbing. One parsing engine, Textkernel, sits under Oracle Recruiting Cloud, SmartRecruiters, SAP SuccessFactors and Bullhorn. Another, DaXtra, sits under iCIMS, Jobvite and Zoho Recruit. Workable runs a language model. So you are not optimizing for forty systems. You are staying legible to about three generations of parser, and the rules for that are boring and stable.

A few more that are worth the ten seconds each:

- **Indeed's profile default is "Employers can't find you,"** and uploading a resume for an application does not change it. One switch puts you back into the largest resume database in the US.
- **LinkedIn Recruiter searches your profile, not your resume.** A clever headline removes you from the literal title filter recruiters actually run. And recruiters cannot sort search results at all, only narrow them.
- **Greenhouse will tell you the questions before you start.** Its board API returns every field on a posting and whether it is required, without any login. That is how you find out a cover letter is mandatory before you are twenty minutes in.
- **On Handshake, clicking through to an employer's own system counts you as an applicant to Handshake** whether or not you finished over there. Handshake documents this itself.

## The kit

I built this out as a reusable prompt for a friend, then generalized it. Both files are free, no signup, and they are plain markdown:

- **[The prompt](/downloads/resume-enrichment-prompt.md)**: paste it as the first message in a fresh AI session. It runs the nine stages, refuses to invent facts, produces a master resume plus a tailored version per posting, and finishes with the gap memo.
- **[The intake questionnaire](/downloads/resume-intake-questionnaire.md)**: fill this in first, save it as `my-answers.md`. This is the part that does the work.
- **[The ATS platform playbook](/downloads/ats-platform-playbook.md)**: what each system does to your resume, the apply-link lookup table, the file format decision rule, layout killers ranked, per platform instructions for about twenty systems, an expanded USAJOBS section, four free ways to test your parse, and an honest list of what could not be verified.

How to use them, in order:

1. Download both. Fill in the questionnaire. Budget an hour and do not skip the last section, which asks what you have not done.
2. Paste the entire job posting into it, plus the exact apply URL.
3. Open a fresh session in Claude, ChatGPT, or a coding agent that can read files. Paste the prompt first, then your answers.
4. It will come back with questions before it writes. Answer them. **If it hands you a finished resume with no questions, it made things up.** Start over.
5. It proposes a positioning before drafting. Push back if you disagree. Changing it then is free.
6. Read the gap analysis before you submit, not after you get the interview.

The prompt is deliberately hostile to the thing most people want, which is a finished resume in one shot with no effort. It will not do that, because the one shot version is exactly the interchangeable document this whole post is about.

If the wider problem behind your job search is that a salary has quietly become the only asset you own, [The W-2 Trap](https://thew2trap.com) is the book I wrote about that. Different problem from this post, same root cause.

## Fact-check notes and sources

- **LinkedIn application volume, 45 percent increase and roughly 11,000 applications per minute**: reported by CNBC, [Recruiters are 'drinking through a fire hose' of job applications](https://www.cnbc.com/2025/10/29/recruiters-are-drinking-through-a-fire-hose-of-job-applications-experts-say.html), 29 October 2025. A LinkedIn representative is quoted elsewhere giving nearly 9,500 per minute, so the per minute figure varies by account. The 45 percent year over year increase is consistent across reports.
- **Workday's 1.1 billion rejected applications, and the ADEA collective certification**: from filings in *Mobley v. Workday, Inc.* (N.D. Cal.). Preliminary certification granted 16 May 2025 by Judge Rita Lin. See [Holland & Knight](https://www.hklaw.com/en/insights/publications/2025/05/federal-court-allows-collective-action-lawsuit-over-alleged) and Proskauer's [Law and the Workplace](https://www.lawandtheworkplace.com/2025/06/ai-bias-lawsuit-against-workday-reaches-next-stage-as-court-grants-conditional-certification-of-adea-claim/). The case is ongoing; nothing here is a finding of liability.
- **Workday acquired HiredScore in 2024**, and HiredScore documents that a CV in an unsupported format may go ungraded. Format and parsing behaviour, including that parsing does not fill Skills or Languages and that image based resumes are discouraged, comes from Workday's own product documentation.
- **The 75 percent ATS rejection statistic** traces to a 2012 Preptel sales pitch with no published methodology; the company closed in 2013. Documented by [The Interview Guys](https://blog.theinterviewguys.com/ats-resume-rejection-myth/) and others who walked the citation chain back.
- **Enhancv 2025 recruiter survey**, roughly 25 recruiters across more than ten ATS platforms, found the large majority do not configure automatic content based rejection. Small sample; treat as directional.
- **NYC Local Law 144** requires annual independent bias audits and candidate notice for automated employment decision tools; enforceable since 5 July 2023. The enforcement audit is [Enforcement of Local Law 144, Audit 2024-N-6](https://www.osc.ny.gov/state-agencies/audits/2025/12/02/enforcement-local-law-144-automated-employment-decision-tools), Office of the New York State Comptroller, published 2 December 2025, covering July 2023 through June 2025. DCWP reviewed 32 companies and identified 1 violation; auditors identified at least 17 potential violations in the same set.
- **Colorado**: SB 24-205 was postponed by SB 25B-004 (signed 28 August 2025) and then repealed and replaced by SB 26-189, signed 14 May 2026, effective 1 January 2027. See [Norton Rose Fulbright](https://www.nortonrosefulbright.com/en-us/knowledge/publications/18733d31/colorado-enacts-revised-ai-law) and [Davis Wright Tremaine](https://www.dwt.com/blogs/privacy--security-law-blog/2026/05/colorado-ai-act-repeal-new-transparency-law).
- **EU AI Act**: Regulation (EU) 2024/1689 classifies recruitment and candidate evaluation systems as high risk under Annex III. Under the Digital Omnibus agreed 7 May 2026, Annex III high risk obligations move from 2 August 2026 to 2 December 2027, and Annex I to 2 August 2028. See [Gibson Dunn](https://www.gibsondunn.com/eu-ai-act-omnibus-agreement-postponed-high-risk-deadlines-and-other-key-changes/) and [Pinsent Masons](https://www.pinsentmasons.com/out-law/news/rules-high-risk-ai-delayed-under-eu-omnibus-deal).
- **Anduril's job board host** was verified directly at `job-boards.greenhouse.io/andurilindustries`.
- **Federal hiring rules.** The two-page federal resume cap and the phase-out of self assessment questionnaires for rating and ranking come from OPM's Merit Hiring Plan memorandum of 29 May 2025 and USAJOBS' own help pages. The restoration of numerical ranking is the "Rule of Many" final rule, published in the Federal Register 8 September 2025, effective 7 November 2025, revising 5 CFR 332.402 and 302.401. **A caveat I owe you:** `usajobs.gov`, `help.usajobs.gov` and `opm.gov` all refused connections from my machine when I went back to re-verify these quotes at source, so I am relying on the citation chain rather than a fresh fetch. These rules moved twice in twelve months. Confirm the current requirement on USAJOBS before you submit.
- **Parser engines.** Textkernel (owned by Bullhorn since June 2024) is confirmed for Oracle Recruiting Cloud, SmartRecruiters, SAP SuccessFactors and Bullhorn. DaXtra is confirmed for iCIMS, Jobvite and Zoho Recruit via subprocessor lists. Workable's is Google Gemini. Workday, Greenhouse, Lever, Ashby, Taleo, Dayforce, UKG, ADP, BambooHR and JazzHR do **not** disclose theirs, so any article naming one is inventing it. Note also that iCIMS runs DaXtra and not Textkernel, which contradicts a claim repeated on a lot of career sites.
- **Vendor-documented behaviour** cited above comes from each vendor's own product or help documentation: Workday on Skills and Languages, Greenhouse on auto-reject notification and its boards API, Indeed on delivering every application regardless of qualification criteria, Handshake on its applicant list, LinkedIn on not offering result sorting in Recruiter.
- **What I deliberately left out.** Every "hiring managers reject AI resumes at X%" figure, "250 applications per opening", the six-second resume scan, all ATS market share percentages (the same product has been reported at 40%, 15% and 0.47% for the same year), and every DOCX versus PDF fidelity comparison. None has traceable methodology.
- **Survey figures on hiring managers detecting or rejecting AI written resumes** are widely quoted and almost all originate with resume products surveying small panels. I have deliberately not cited specific percentages. The qualitative finding, that the objection is to lack of personalization rather than to AI use, is consistent across them.
- **Applications per opening figures**, commonly given as 250 or 300 plus, circulate without traceable methodology. I have avoided them.

## Related reading

- [What $200k AI Jobs Actually Ask For, and How to Practice Every Skill Free](/blog/blog-ai-ml-github-projects-200k-jobs/)
- [Mercor: The $10 Billion AI Hiring Platform, Its Data, and the Cracks Showing](/blog/blog-mercor-ai-hiring-data-connections/)
- [How to Build the Skills Behind a Lead AI/ML Platform Engineer](/blog/becoming-ai-ml-platform-engineer/)
- [The AI Posture Audit: From Per-Bot Matrix to One Master Prompt](/blog/blog-ai-posture-audit-master-prompt/)
- [Top AI CLIs and How To Feed Them The Prompts Our Generators Build](/blog/blog-ai-clis-with-our-prompts/)

*This post is informational, not legal or career advice. Mentions of Workday, LinkedIn, Greenhouse, Anduril, and other named companies and products are nominative fair use. No affiliation is implied. Litigation described here is ongoing and nothing above should be read as a finding of fact or liability against any party.*


---

Canonical HTML: https://jwatte.com/blog/one-resume-per-job-ats-standard/
RSS: https://jwatte.com/feed.xml
JSON Feed: https://jwatte.com/feed.json
Hero image: https://jwatte.com/images/blog-one-resume-per-job-ats-standard.webp