The Application Answer Bank, Written at 255, 600, and 1,200 Characters

This guide shows you how to build an application answer bank. That's one document of pre-written, pre-checked answers to the free-text questions application forms keep asking, in each length a form is likely to allow. It's for anyone sending out a batch of applications who is tired of starting from an empty box. After one working session of about three hours, you'll have 8 to 10 prompt families, each written at 255, about 600, and about 1,200 characters. Every sentence will trace to a resume line or a record, and every detail that changes per employer will sit in a marked slot.

What is an application answer bank?

An application answer bank is a single document of reusable answers to recurring application-form questions. The answers are grouped by prompt family, written at three fixed lengths, and split into a fixed core plus slots you fill for each employer. It turns "describe your experience with X" from a writing job into a fill-in job.

Three terms come up throughout:

• Prompt family: a set of prompts worded differently that all want the same underlying answer. "Why do you want to work here?" and "What draws you to our mission?" belong to one family.
• Length tier: one version of an answer written to a fixed character ceiling, with spaces counted. This bank uses 255, about 600, and about 1,200.
• Employer slot: a marked gap, such as [COMPANY-SPECIFIC], that you fill fresh for each application. Everything outside the slots is the fixed core, which you write once and check once.

How this differs from the sample answers that rank

Search for "why do you want to work here application answer" and you'll find pages like Novoresume's list of 79 sample answers and the Himalayas formula page. They give one-off answers to one prompt, and they start from the company. That's fine for a single application. Across a batch, though, you end up writing the same paragraph from scratch every time, and nothing checks that a figure matches your resume.

This method starts from your own records and gives you something you keep. It's also different from a behavioural story bank, which prepares stories you tell out loud in interviews. This bank is written down, kept under length limits, and pasted.

Why the effort pays off

Ashby's Application Questions report looked at 4.8M submitted applications over four years. It found that the number of questions per application nearly doubled in that time.

The same report says that "even with significantly more questions, the average rate of questions answered is relatively stable." In other words, people who finish applications are answering more questions than before. Ashby also reports that applications per hire have tripled, with roles now getting more than 300 applications per hire on average. So there's more competition and more writing per form. If you send any volume at all, that's the case for building the bank.

The limits of that evidence:

• Ashby's data covers only submitted applications. It doesn't count people who started a form and gave up.
• An earlier Ashby report described its sample as mostly US tech companies with 1 to 1,500 employees.
• Nobody has published a comparable figure for Workday or iCIMS enterprise employers. If you apply mainly to large non-tech employers, count the long-form questions in your own recent applications before you trust the 46.8%.

Which character limits do application forms actually enforce?

Only one limit has been reproduced in testing: 255 characters on Greenhouse single-line text boxes, which cut longer answers off without warning. Workday limits are set by each employer and aren't documented. The 600 and 1,200 tiers in this guide are my design choices, not published standards.

Greenhouse: 255, silent, and not shown in advance

Greenhouse's official Job Board API lists the field type input_text and says to render it "with an input of type text." It doesn't publish a length limit. Practitioners found one anyway. In one public test, an answer of 512 characters went in and the page kept the first 255, cut mid-sentence, with no message. The cause was that the box carries maxLength="255", and the job board API doesn't mention it. A second project confirmed that the public API identified these questions as input_text but didn't expose their length limits. Greenhouse's own docs list the first and last name fields as "max 255 chars, no URLs," so 255 is a ceiling the platform uses elsewhere too.

That's the practical problem. Neither you nor any tool can read the limit before you paste. The form accepts the cut-off text and looks like it submitted correctly. A 255-character version written in advance is the only safe default.

Workday: limits set per employer, saving unreliable

Workday limits are set by each employer when it configures the form, and nobody has published them. Practitioners report unannounced character limits, no autosave, and forms that clear themselves if your laptop goes to sleep mid-sentence. Sources disagree here: one guide says Workday does autosave, "but not reliably across all configurations," while another says some tenants have no autosave at all. Plan for the worse case.

Why three tiers and not one

The tiers nest, so you only write once. You write the long version, then cut it down, and the shortest core survives every cut. If a field states a limit between two tiers, such as 1,000 characters, trim the next tier up to fit instead of writing a new answer. The 600 and 1,200 figures are a structure for doing that trimming, not a claim about what forms enforce.

To check a limit yourself, paste your 255 version and confirm the last word is still there. You can also open your browser's developer tools and look at the field's maxlength attribute. Record whatever you find in the bank against that employer.

Which applicant tracking system will you meet most?

Outside tech, it's most likely Workday. In Refolk's index of professional profiles, 533 US recruiter and talent acquisition profiles list Workday as a skill, against 121 that list Greenhouse. An applicant tracking system (ATS) is the software an employer uses to publish postings and collect applications, and the form you fill in belongs to it.

Read that table with care. These are skills recruiters chose to put on their profiles, not a count of installed systems. Still, the samples clearly differ. Other top employers in the Greenhouse sample include Paraform, Zearn, and YETI, a mostly tech and consumer mix. The Workday sample includes C&S Wholesale Grocers and Green Dot alongside Boeing and a hospital system. Don't read anything into the UK: Refolk's index shows only 2 UK recruiter profiles listing Greenhouse, against 121 in the US. With n=2, that comparison tells you nothing.

What this means for the bank:

• Applying to tech startups and scale-ups: silent truncation at 255 is the main risk, so the 255 tier matters most.
• Applying to enterprise, healthcare, public sector, or retail: losing typed text and running into undisclosed limits are the main risks, so drafting outside the form matters most.
• Applying to both: keep all three tiers and the habit of drafting outside the form. Together they cover both systems.

Which prompt families should the bank cover?

Build your families from prompts you've actually seen, not from a published list, because nobody has published a ranked frequency count of application prompts. Thirty or more raw prompts from your own past and planned applications will usually sort into 8 to 10 families.

Here's a starting set to sort against. Change it, merge families, or split them to match what your list actually contains:

1. Why this company. Carries the most slots of any family. The core states your relevant record, and the slots give the reason.
2. Why this role. The core is the line through your history that leads to this job. The slot holds the role's stated priority.
3. Describe your experience with X. This is one sub-family per core skill you're hired for, since X changes. Each gets its own three tiers.
4. Something not on your resume. The core holds one sourced fact your resume leaves out. There's often no employer slot at all.
5. A project or accomplishment you're proud of. Mostly core. Use the slot to link it to the posting.
6. Why you're looking or leaving. Mostly core, kept short and neutral.
7. Anything else you'd like us to know. A small cover note. Core plus one company slot.
8. Salary expectations. A short fact, not a narrative.
9. Notice period and start date. A short fact.
10. Work authorisation. A short fact.

The last three are short-fact families. Write each as one exact line and skip the tiers. One practitioner who writes about Workday keeps stock answers for exactly these (work authorisation, salary range, notice period) along with "why this company." The tiered structure is for the narrative families, 1 to 7.

If a prompt won't fit any family, don't make a new family for one prompt. Park it in a "misc" list. If the same kind of prompt turns up again in later applications, promote it to a family during your next update.

The build session: eight steps in about three hours

The bank takes eight steps and about three hours, in this order: gather records, collect prompts, group them, write the master, compress, mark slots, verify, use. Starting from your records instead of from the company is the deliberate break from template-first advice. It's what makes every claim hold up when an interviewer asks about it.

Step 1: the source register

Your source register is the numbered list every answer points back to. Good inputs are your resume, any brag document or work log, and performance reviews. Julia Evans's essay "Get your work recognized: write a brag document" describes the brag-document habit. Gergely Orosz publishes a work log template that serves the same purpose. A brag doc kept up properly reaches roughly 5 to 10 pages by year-end. That's more raw material than you need, so take only lines with a concrete outcome.

If Refolk wrote your resume from your own history, those lines are already pulled out of your records, and you can number them straight away.

R1 | Resume | <Role, Employer, Dates> | <the bullet exactly as it appears>
R2 | Work log | <Month> | <what you did, the figure, and how you know it>
R3 | Review | <Reviewer role, review cycle> | <the quoted line>
R4 | Record | <document or dashboard name> | <the metric and where it lives>

Steps 2 and 3: prompts and families

Paste prompts exactly as written, including the field's label. Wording tells you a lot about what the employer wants. For Greenhouse-hosted boards, the public Job Board API returns the application form's questions, so if you're comfortable with a URL request you can pull them straight from a posting. Keep in mind that the API won't tell you their length limits. For everything else, copy the prompts by hand as you meet them.

Step 4: writing the master

Write the 1,200 master as if this were the only answer you'd ever send: one core claim, two pieces of proof, the reason this employer, and a closing line linking you to the role. Put a tag at the end of each sentence, such as (R4). A sentence you can't tag is either a slot or something to cut. Opinions and motivations count as tagged if a record supports them. "I want to work on payments infrastructure" is supported if R2 shows you've done payments work.

Step 5: compressing

Cut from the bottom up. From 1,200 to 600, drop the second proof and the closing line, and merge the slots into one sentence. From 600 to 255, keep the core claim with its single best figure plus one slot. Count characters with spaces included. Don't count the tags, since you'll remove them before pasting. Aim for about 230 so a long slot fill still fits under 255.

Step 6: marking slots

Slots are where the reader can tell whether your answer was written for them. Use a few names consistently across the bank:

• [COMPANY-SPECIFIC]: something true of this employer and no other, taken from their posting, product, or published material.
• [ROLE PRIORITY]: the problem the posting says this hire will solve.
• [TEAM OR PRODUCT]: the specific team, product, or customer you'd work on.

FAMILY: Why this company
PROMPTS MAPPED: <paste each raw wording you grouped here>
SOURCES USED: R4, R9, R12

1,200 MASTER (target 1,100 to 1,200 characters, spaces counted, tags not counted)
<Core claim sentence.> (R4)
<Proof sentence with the figure.> (R9)
<Second proof sentence.> (R12)
What draws me to you specifically is [COMPANY-SPECIFIC].
Because you are prioritising [ROLE PRIORITY], <closing link back to R4 or R9>.

600 VERSION (target 550 to 600)
<Core claim.> (R4) <One proof with figure.> (R9) [COMPANY-SPECIFIC], which matters for [ROLE PRIORITY].

255 CORE (hard ceiling 255, aim for about 230)
<Core claim with one figure.> (R9) [COMPANY-SPECIFIC]

LIMITS SEEN: <employer | field | limit observed | how you checked>

Step 7: verifying

Go through each family and check every tag against the register. Then paste each 255 version into any field you know has a 255 cap, or into a character counter, and confirm the last word is still there. This takes ten minutes, and it's the step that catches the problems you otherwise wouldn't see until an interview.

How does an answer bank go wrong?

An answer bank fails in seven predictable ways, and most of them are invisible when you submit. The form will happily accept truncated, empty, or untraceable text. Each failure below comes with what it looks like and the check that catches it.

1. Truncation at 255. The answer is cut mid-sentence and no error appears, so the form looks fine. Check: re-read the field after pasting, and keep the claim in the first 200 characters.
2. Lost Workday text. The back button, a refresh, or a session timeout wipes anything you hadn't saved. Within a stage, your typing stays local until you press save. Check: always paste from the bank, never compose in the form, and press save between stages.
3. "Answered" but empty. Ashby's AI-assisted review only checks whether the question was answered, and flags blank or off-topic answers. Ashby says its AI never ranks applicants or gives them numerical ratings, and a human makes the decision. Ashby also reports that 50 to 60% of candidates eventually get screened by a person. So a generic answer gets past the machine and fails with the human. Check: every block you submit contains at least one filled slot.
4. Untraceable claim. A figure in the answer doesn't match your resume. Ashby's guidance to candidates says plainly that answers "will be verified during the interview process." Check: every sentence carries a tag that resolves in the register.
5. Leftover template text. A bracket like [Company Name] makes it into the submitted answer. The template-first pages that rank teach exactly this format, which is why it's so common. Check: search the pasted text for "[" before submitting.
6. Generic tells. As one analysis puts it, "what triggers the recruiter's radar is the absence of specific details, not the algorithmic fingerprint." Check: delete the slot content and ask whether what's left would work for any employer. If it would, rewrite it.
7. Assuming one system's limits apply everywhere. Workday limits differ from one employer to the next, and a limit you saw on one form tells you nothing about the next. Check: count characters before every paste, and record limits per employer.

Whether a one-line answer, or an answer identical to other applicants', is treated differently by reviewers isn't publicly established. I wouldn't assume either a penalty or a free pass. The slot check covers both cases.

The AI-detection question, sized honestly

People worry that reused or AI-assisted text gets caught. The evidence is weaker than either side claims:

• A secondary report of a blind test with 1,000 recruiters says 82% couldn't correctly identify all the AI-written letters. I couldn't get the primary source, so treat this as unconfirmed.
• A claimed TopResume survey says 67% of hiring managers believe they can spot AI-written cover letters. That's people's opinion of their own ability, also unverified at the source, and it contradicts the blind test.
• AI detectors flagged an average of 61.3% of TOEFL essays written by non-native English speakers as AI-generated. If English isn't your first language, a detector can flag your own writing.
• A Robert Half survey found 65% of hiring managers say AI-enhanced applications make it harder to verify candidates' skills.

Taken together: verification is what worries reviewers, and missing detail is what makes them suspicious. Answers built from your own numbered records protect you better than any trick for evading detection, and that matters most for non-native writers, who are the people detectors flag wrongly.

Filling a live application in under ten minutes

On a live form, you pick the family, choose the tier that fits, fill the slots from the posting in your document, strip the tags, count, paste, re-read, and save. Never write inside the form. Workday can throw away unsaved typing, and Greenhouse can cut it without telling you.

The per-form routine:

• Identify the family. Match the prompt's wording to a family. If it combines two, such as "why this role and what you'd bring," take the 255 core from each family and join them.
• Pick the tier. If the field is a single-line box, use 255. If it's a text area with a stated limit, use the largest tier under that limit and trim if needed. If there's no stated limit, use 600 unless the prompt clearly asks for depth.
• Fill the slots in your document. Pull [COMPANY-SPECIFIC] and [ROLE PRIORITY] from the posting itself. This is the same reading you do when tailoring a resume. If you're already using Refolk to tailor your resume to each posting and score your fit, the requirements it finds there are your [ROLE PRIORITY] fill, so you aren't reading the posting twice.
• Strip and count. Remove every (R#) tag, search for "[", and count characters.
• Paste, re-read, save. Read the field from the first word to the last, then press save before you move to the next stage.

Log every limit you find against the employer and the field. After a dozen applications, that log is the only record of real limits you'll have, because neither ATS publishes them.

Keeping the bank current

Update the bank whenever your source records change and whenever a new prompt doesn't fit an existing family. A quarterly pass works well, and it matches the cadence brag documents are usually kept on. The bank is only as reliable as its register. A stale figure in R4 spreads to every answer that cites it.

The maintenance pass, in order:

1. Update the register first. Add new outcomes from your work log as new R numbers. If a figure has changed, retire the old line and add a new one instead of editing it in place, so you can see which answers still cite the old number.
2. Re-tag affected answers. Search the bank for each retired R number and rewrite those sentences against the new line. Recompress all three tiers. Don't patch only the long one.
3. Promote misc prompts. Any prompt that has come up again since the last pass becomes a family with its own three tiers.
4. Review the limits log. If an employer's limit has changed or a new pattern shows up, note it. Workday configurations are set per employer and can change without notice.
5. Re-run the matrix test. Read each fixed core and put every sentence in a quadrant. Anything that has drifted into "honest but generic" needs a sharper slot, and anything in "plausible fiction" needs a source or has to go.

Where the evidence is thin, check it yourself instead of trusting anyone's number, mine included. The 46.8% long-form figure describes submitted applications at mostly US tech companies. The 4.4x Workday-to-Greenhouse ratio counts skills recruiters list on their own profiles. Only the 255 limit has been reproduced. Your own limits log and your own count of long-form questions are the figures that actually describe the forms you'll fill in.