# The Sent-Resume Question Sheet, One Tailored Version Traced Through a Loop

*Pull the resume an employer holds, diff it against your master, predict 15 to 30 questions from its tailored lines, and log a hit rate for every round.*

- Canonical URL: https://www.refolk.ai/candidates/guides/sent-resume-question-sheet
- Pillar: Interviewing
- Format: Teardown
- Published: 2026-10-30
- Last reviewed: 2026-10-30
- Reading time: 19 min
- Keywords: how to prepare to talk about your resume in an interview, which resume version did I send, questions interviewers ask about your resume, interview prep from tailored resume, can't explain resume bullet in interview

## Key takeaways

- Prep from the file the employer holds, not your master: Workday documents no candidate download and no edits after submission, so your tracker's resume-version column is often the only reliable copy.
- In Greenhouse, interviewers with Basic permissions see your attached resume and their scorecard but not your parsed profile, so parse errors tend to come up at the recruiter screen and the wording of your file comes up at the panel.
- At 5 to 10 tailored line changes per application and three OPM-style probes per line, expect to rehearse 15 to 30 predicted questions about the resume itself.
- In Refolk's index, only 73 of 21,285 US Technical Recruiters (0.34%) list Greenhouse as a skill, so identify an employer's ATS from the apply URL, not from its recruiters' profiles.
- Hiring managers picked the AI-written resume out of four only 33.5% of the time against a 25% guess rate, so the real test of a stretched line is the follow-up probe, not a detector.
- Count a hit only when an asked question maps to one predicted line, because counting topical overlap inflates the hit rate and hides the lines that need fixing.

You tailored a resume for this posting, the invite just arrived, and the version the employer holds is not your master. This guide is for candidates preparing for a first call against that exact file. It follows one tailored version from invite to final round, including the decisions and the wrong turns, and ends with a question log that becomes the list of fixes for the next version you send.

## What does the employer actually hold when you interview?

The employer can hold four things about you: the file you uploaded, the profile its applicant tracking system parsed from that file, the scorecards interviewers fill in, and their notes. Different people see different pieces, so the same resume line can draw different questions in different rounds.

An applicant tracking system (ATS) is the software an employer uses to receive, store, and route applications. Parsing means pulling structured fields out of your document, such as employer, title, dates, and skills.

#### What an employer holds about one application

1. **Your sent file** - The PDF or DOCX attachment, worded exactly as you uploaded it
2. **Parsed profile** - Structured fields the ATS pulled from the file, sometimes only partly
3. **Interview kit and scorecard** - Instructions, questions, your resume, and a rating form for each interviewer
4. **Interview notes** - What interviewers wrote down, which UK applicants can ask to see

*Your file is one artifact among four, and each interviewer sees only some of them.*

Two systems document what each viewer gets:

- **Greenhouse.** The interview kit holds interview instructions, interview questions, the candidate's resume, and a scorecard. Interviewers with Basic permissions see public contact information, attachments (resume and cover letter), and their assigned scorecard, but "will not be able to view the candidate's profile." Basic users also "can't download interview kits, resumes, or other candidate data."
- **Workday.** Reviewers open an Attachments tab to see the resume. The profile shows structured fields: applicant profile, experience, education, and attachments.

| Round | Usually run by | What they look at | What tends to come up |
|---|---|---|---|
| Recruiter screen | Recruiter | Parsed profile and file | Parse errors: blank dates, missing employer |
| Panel (Greenhouse Basic) | Interviewer | Attachment, kit, scorecard | The wording of tailored lines |
| Workday review | Reviewer | Attachments plus profile fields | Both |

This table has limits. I did not check Lever or iCIMS. Public sources don't show whether recruiters see an AI-generated summary or "match highlights" by default.

AI is widespread in hiring, but nobody has published how often it writes questions about you in particular. A ResumeBuilder survey of 948 business leaders found that 82% of companies use AI to review resumes, and 76% planned to use it to ask interview questions. TestGorilla found that 21% of US organizations and 20% of UK organizations use generative AI to run at least first interviews. Public sources don't show how many employers generate questions from a specific candidate's resume. Prepare as though a person will read your file line by line, because at the panel stage one does.

## The worked case, and the first wrong turn

The case is a backend engineer who applied through an employer's `*.myworkdayjobs.com` site with a version tailored from their master resume. It has the faults that cause the most trouble in practice. The first one happened before any prep began: the candidate opened the master.

Huntr says 5 to 10 line-level changes per application is typical. This case is at the top of that range, with 10 changed lines. Three of them matter later:

- **Python in the skills list**, added to match the posting.
- **The current role dated "3/22 to present."**
- **One role written in the shape "Senior Product Manager, Google"**, with the title and the employer on one line.

**Wrong turn 1: prepping from the master.** The night before the recruiter screen, the candidate opened master.docx and started writing answers. It felt right because every line was familiar, and that familiarity is the trap. The 10 changed lines are the ones the candidate knows least, and they're the ones written to match the posting the panel is scoring against.

**How it was caught:** the master's last-modified date didn't match the date in the tracker's "Resume version" column. **Decision:** stop writing answers and recover the sent file first.

> **Rule:** Prep from the file they hold
>
> Every answer you rehearse must come from the exact file the employer received. If you can't produce that file, recover it before you write a single answer.

The tracker column did the work here. One practitioner template lists these fields per application: Company, Role, Source, Date applied, JD link, Resume version, Cover letter (Y/N), Referral (Y/N + name), Recruiter contact, Status, Next follow-up date, and Outcome / notes. One guide calls the version field "The column almost everyone skips." Huntr's tracker recommends storing "a unique resume and cover letter per job to easily access the correct documents during interviews." Sources disagree about how generous the free tiers of these trackers are, so check the current plan page before you rely on one.

If [Refolk](/candidates) tailored the version for this posting, open the version written for that posting, not the work history it was built from. The tailored lines are the ones you'll have to defend.

## Which resume version did I send, and how do I get it back?

Your own records are the most reliable copy. After that, try the portal, then the recruiter, and in the UK a subject access request. Identify the ATS first, because it determines what the portal can give back.

### Name the ATS from the apply URL

The apply URL tells you the system. A Workday application runs on the employer's own `*.myworkdayjobs.com` site. Don't try to work it out from the employer's recruiters, because almost none of them list the tool they use.

**0.34% - US Technical Recruiters in Refolk's index who list Greenhouse as a skill**

73 of 21,285 profiles, so a recruiter's profile tells you almost nothing about the employer's ATS.

| Cohort | Count | Share |
|---|---|---|
| US Technical Recruiters | 21,285 | 100% |
| ...listing Greenhouse | 73 | 0.34% |

Source: Refolk's index of professional profiles. The share is derived. The Greenhouse sample also contains noise, including a Pilates Instructor, so read the share as an order of magnitude, not a precise rate. The named employers in that small group include Snowflake, TikTok, and Meta.

### The recovery fork

Try these in order and stop at the first one that gives you the byte-identical file:

1. **Your tracker.** You're done if the Resume version entry points to a file dated the same day you applied.
2. **The portal.** Workday has "no documented download action for your own resume," and you "cannot change your application after submission." Once you submit, "you won't be able to edit it - including replacing your resume or cover letter."
3. **The recruiter.** One practitioner's fallback: "Ask the employer. They hold your application; a recruiter can usually send back what you submitted."
4. **A UK subject access request.** A subject access request (SAR) is a formal request for the personal data an organisation holds about you, and it "gives someone the right to obtain a copy of their personal information." Under the ICO employment practices code (now archived), applicants also "have a right to request access to their interview notes."

**Wrong turn 2: trying to download from the portal.** The candidate followed a guide's "Continue Application" path. That route is for applications still in progress, and even there "you aren't able to download them." The better route is Candidate Home, then My Applications, then the view of submitted materials, as the University of Washington's internal-applicant guide describes. In this case the file wasn't viewable there either. **Decision:** email the recruiter in the same message that confirms the interview slot.

**Request for the submitted resume**

```
Hi [Recruiter name],

Thanks for the invite. [Day and time] works for me.

One small request: could you send me the resume file I submitted for the [Role title] application on [date applied]? I tailored it for this role and want to make sure I'm preparing from the exact version you have.

Thanks,
[Your name]
```

*Send it in the reply that confirms your interview time, and name the role and the date you applied.*

> **Note:** SAR timing is a secondary-source figure
>
> A secondary source says UK SAR responses are due within one month and can be extended by up to two months. Check the ICO's own guidance before you count on that deadline.

## How can the parsed profile misread your resume?

The parser can blank out a date, merge a job title with the employer, or attach dates to the wrong job. These errors tend to come up at the recruiter screen as gap questions or "where did you work?" questions. I found no parser test with a published method and error rates, so the table below lists documented failure types, not how often each one happens.

| What you wrote | What the parser can do | Question it causes |
|---|---|---|
| "January 2022 - Present" | Start date kept, end date blank | Why did you leave? |
| "3/22 to present" or a graphic timeline | Empty date fields (Workday) | What happened in this gap? |
| "Senior Product Manager, Google" | One job-title field, no employer | Where was this role? |
| Two-column layout | Dates on the wrong employer, skills mixed into work history | Timeline confusion |

Greenhouse documents that a parse can be "partially successful (automatically filling some information, but skipping others)." A partial parse looks fine at a glance, and that's what makes it risky.

In the case, the candidate ran the sent file through a parser tool that shows raw JSON and checked that "Every job has start and end dates." Two errors came back:

- The current role had empty date fields, from "3/22 to present."
- The role written in the "Title, Company" shape had a blank employer field.

**Wrong turn 3: trying to fix the file.** The candidate's first instinct was to rewrite those two lines and upload again. Workday doesn't allow edits after submission, so that time was wasted. **Decision:** leave the file alone and prepare a spoken correction for each error, one sentence each, ready for the recruiter screen. For example: "My current role started in March 2022 and I'm still there. The system may have dropped the end date."

Sources disagree on when to do this check. Parser vendors put it before you submit ("Correct every pre-filled field."). This case does it afterwards because the application was already in. For your next application, run the parse check before you press submit.

## How do you turn tailored lines into questions interviewers ask?

Give each tailored line three probes: who was involved, what specific actions you took, and what the outcome was. At 5 to 10 changed lines, that's 15 to 30 predicted questions, which is the full rehearsal set for the resume part of the loop.

The probe sequence comes from the US Office of Personnel Management (OPM). Its structured-interview guide says to "establish the desired range of probing" before the interview and gives the example probes "Who was involved? What specific actions did you take? What was the outcome?" One limit: OPM writes its questions from job analysis, not from the resume, and I found no public-sector guide that says to derive probes from resume lines. This method borrows OPM's probes, not its reasoning.

### The three-way diff

Compare the sent file against two references: your master resume and your public LinkedIn profile. Tag every line that differs:

- **Added:** in the sent file, not in the master.
- **Reworded:** the same work, described in the posting's vocabulary.
- **Removed:** in the master, missing from the sent file.
- **Differs from public profile:** the sent file says something your LinkedIn doesn't.

The last tag matters if anyone opens your public profile next to the file. A mismatch you predicted is an easy question. One you didn't predict is a hard one.

### Map each line to a requirement

For each tagged line, write down the requirement in the posting it was meant to answer. One practitioner claims each requirement bullet "typically maps to one scorecard row." Greenhouse hasn't confirmed that, so use it as a working assumption. A line that maps to no requirement is usually a keyword stretch.

#### Which tailored lines to rehearse first

Horizontal axis runs from Low drift from master to High drift from master. Vertical axis runs from Low weight in posting to High weight in posting.

| Quadrant | What it means |
| --- | --- |
| Core proof | Rehearse all three probes, because this is your strongest evidence |
| Probe hotspot | Rehearse first, then name a project, a date, and an outcome or narrow the claim |
| Background | A one-sentence answer is enough |
| Keyword stretch | Answer honestly and cut it from the next version if you can't back it |

*The lines that drifted furthest from your master and carry the most weight in the posting draw the hardest probes.*

### The stretched tool claim

Python, added to the skills list for this posting, landed in the keyword-stretch quadrant. A parser treats Python the same "whether the candidate built production systems in it for a decade or listed it after taking a course." The candidate had taken the course. The probe "What specific actions did you take?" will expose that within one follow-up.

**Decision at this fork:** don't defend the line. Narrow it out loud ("I've used Python in coursework. My production work is in the language I shipped in.") and mark it to be cut from the next version.

Detection confidence is not the same as detection, and that's why the probe matters more than any tool.

| Measure | Value | Source |
|---|---|---|
| Say they can identify AI resumes | 80% | Resume Genius 2026 |
| Say they can tell immediately | 53% | Insight Global |
| Actually picked the AI resume out of 4 | 33.5% | TopResume 2025 |
| Random-guess baseline | 25% | Derived |

**33.5% - How often hiring managers picked the AI-written resume out of four**

Against a 25% chance rate, even though 80% in another survey said they can spot one.

Greenhouse's report found that 61% of US hiring managers use some software to detect AI in hiring. Whether or not a detector flags a line, an interviewer asking for actions and outcomes will find out whether you can back it.

> The probe is the detector. A line you can't attach a who, an action, and an outcome to will not survive it.

Refolk's fit score is useful at this stage: the requirements where your fit is thinnest are where the hotspot and stretch lines sit, so rehearse those first.

### Rehearse out loud

OPM advises having people "review questions and read them aloud." Answer each probe aloud, briefly enough that an interviewer could summarise it in a scorecard comment box. Greenhouse lets employers set scorecard visibility to "Never, Always, and Only after submitting their own," so a later interviewer may read what an earlier one wrote. Tell the same story the same way in every round.

## Step-by-step: from invite to a logged hit rate

This is the procedure the case followed, in order, with times taken from the dossier's estimates. The whole run fits in an afternoon if the recruiter sends the file back quickly.

#### The sent-resume question sheet procedure

1. **Name the employer's ATS** - Read the apply URL. A Workday application runs on the employer's own *.myworkdayjobs.com site. About 5 minutes.
2. **Recover the byte-identical sent file** - Check your tracker, then the portal's submitted-materials view, then ask the recruiter. Never re-export from the master. 10 to 30 minutes.
3. **Rebuild the parsed profile** - Run the sent file through a parser that shows raw JSON and confirm every job has start and end dates and its own employer field. About 15 minutes.
4. **Diff sent version, master, and public profile** - Tag every changed line as added, reworded, removed, or different from LinkedIn. About 30 minutes.
5. **Map each tailored line to a posting requirement** - Write the requirement each line answers. Lines that map to nothing are keyword stretches. About 30 minutes.
6. **Write three probes per tailored line** - Use who was involved, what specific actions you took, and what the outcome was. 60 to 90 minutes.
7. **Rehearse every answer out loud** - Answer each probe aloud, briefly enough to fit a scorecard comment box, and narrow any claim you can't back. About 60 minutes.
8. **Log predicted versus asked the same day** - List every question asked, mark a hit only when it maps to one predicted line, and compute the hit rate. About 15 minutes per round.
9. **Fix the next version you send** - Rewrite or cut the lines that drew missed or weak answers, and fix any layout that broke the parse. About 30 minutes.

## How this goes wrong

Most failures here look like good preparation. The candidate feels ready, but they rehearsed the wrong file, trusted a parse they never looked at, or scored their own log too generously. Each failure below is paired with the check that catches it.

| Failure | Why it looks right | Check |
|---|---|---|
| Prepping from the master | The lines are familiar | File hash or last-modified date against the tracker's Resume version column |
| Assuming the portal holds your copy | Workday shows the application | No documented download; "Continue Application" is for drafts |
| A garbled date becomes a "gap" question | The file reads fine to you | Null dates in the parsed JSON |
| A title swallowed the company name | The line looks normal on the page | A separate company field for every role |
| A skills-list stretch | It matched the posting | A project, a date, and an outcome for every listed tool |
| An inflated hit rate | Questions felt familiar | A hit counts only if it maps to one predicted line |
| Assuming the panel sees the parse | The recruiter mentioned it | Which role runs which round |

Three of these need more explanation.

**The garbled date.** A blank end date after "Present" reads as unemployment. The recruiter isn't being hostile when they ask about the gap. They're reading the field they were given. Your spoken correction has to be short and say where the error came from: what the file says, and what the system may have dropped.

**The panel and the parse.** In Greenhouse, Basic interviewers see attachments, not the profile. If you spend panel prep defending parse errors, you've prepared for the wrong audience. The panel will ask about your wording.

**The hit rate.** This is the failure that distorts everything that follows.

> **Watch out:** Topical overlap is not a hit
>
> "Tell me about your backend work" is not a hit just because you prepared something about backend work. Count a hit only when the question maps to one specific predicted line, or your log will report you were ready for questions you weren't.

An inflated hit rate does more damage than a low one. A low rate sends you back to fix lines. An inflated rate tells you the version is fine and you send it again.

## How do you log each round and fix the next version?

Log every round the same day: the questions you predicted, the questions you were asked, and which of the asked questions match one predicted line. Hit rate is matched questions divided by total questions asked. The questions you missed become the fix list for the next version.

**Per-round question log**

```
Role / company:
Resume version file name:
Round type (recruiter screen / panel / hiring manager / final):
Run by (recruiter, agency, or panel):

Predicted questions rehearsed:
Asked questions (write each one down):
1.
2.
3.

For each asked question:
- Maps to predicted line # ____ / no match
- Answer quality: solid / weak / couldn't answer

Hits (asked questions that map to one predicted line):
Total asked:
Hit rate = hits / total asked:

Misses to fix in next version:
Lines that drew weak answers:
Parse errors that came up:
```

*Copy one block per round. Fill in "Asked" from memory within the hour, before you look at your predictions.*

### The limit on the number

I'm not printing a hit rate for the worked case. No published baseline exists to compare it against, and an invented one would anchor yours. Compare your rate across your own rounds and your own versions. A useful hit rate is one that moves after you change a line.

### Who ran the round

Record who ran each round, because the recruiter and the panel are looking at different things. In Refolk's index, the US has 21,285 people titled Technical Recruiter and the UK has 709, about 30 times as many in the US. Several of the top employers in the US sample are staffing firms, including K2 Partnering Solutions, Experis, and Apex. I haven't verified this, but I'd expect some first calls to be run by an agency holding its own copy of your file. If an agency runs the screen, ask which version of your file it has.

### Fixing the next version

Turn the log into edits:

- **Missed questions** point to a line that invited a probe you didn't see coming. Rewrite it so the probe has an obvious answer.
- **Weak answers** point to a line that claims more than you can back. Narrow it or cut it.
- **Parse errors that came up** point to a layout problem. Fix date formats and title-employer lines, then run the parse check before you submit.

When Refolk tailors your next application, give it the lines the log flagged so the new version starts from what you can actually defend.

#### Before the first call

- [ ] The file I'm preparing from matches the Resume version entry and the application date in my tracker.
- [ ] I hold the sent file itself, not a fresh export from my master.
- [ ] Every role in the parsed output has a start date, an end date or present, and its own employer field.
- [ ] I have a one-sentence spoken correction for each parse error.
- [ ] Every changed line is tagged against my master and my public profile.
- [ ] Every listed tool has a project, a date, and an outcome I can name.
- [ ] Every tailored line has three probes, and I've answered each one out loud.
- [ ] I know which rounds the recruiter runs and which the panel runs.
- [ ] My per-round log is copied and ready to fill in after the call.

## Frequently asked questions

### Which resume version did I send if I didn't track it?

Check your sent email and downloads for a file dated the same day as your application, then look in the portal's submitted-materials view, then ask the recruiter. Workday documents no download action for candidates, and you can't change an application after you submit it, so the employer's copy is fixed. A recruiter can usually send back what you submitted. In the UK, a subject access request gives you a right to a copy of your personal information. Don't guess from your master.

### Can I download my resume from Workday after I apply?

Not through any documented route. Workday has no documented download action for your own resume, and the 'Continue Application' path is for applications you haven't submitted yet. Even there, one guide says you can't download the files. Try Candidate Home, then My Applications, and look for a way to view submitted materials. If you can't see the file, email the recruiter and ask for the file you submitted.

### What should I do if I can't explain a bullet on my resume in the interview?

Narrow the claim out loud rather than defend it. Say what you actually did, when, and what came of it, even if that's smaller than the line suggests. The probes that expose a stretched line are who was involved, what specific actions you took, and what the outcome was. Prepare those three answers for every tailored line before the call. Cut any line you can't answer from the next version you send.

### Do interviewers see my parsed ATS profile or my actual resume file?

It depends on the system and the interviewer's permissions. In Greenhouse, interviewers with Basic permissions see your attachments and their assigned scorecard but can't view your candidate profile. In Workday, reviewers open an Attachments tab and also see structured profile fields. So a parse error usually comes up with the recruiter, and the wording of your file comes up with the panel. Lever and iCIMS were not checked.

### How many questions should I prepare about my resume?

Plan for roughly 15 to 30. A tailored application typically changes 5 to 10 lines from the master, and each changed line deserves three probes: who was involved, what you did, and what the outcome was. That number is my own arithmetic from a tool vendor's typical-change figure and OPM's probe sequence, not a published benchmark. Your log of asked questions will tell you whether it's right for you.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/candidates/guides/sent-resume-question-sheet*