Skip to content

Prompts · Resume

Resume prompts for ChatGPT and other AI, tested

10 resume prompts, each run on GPT-6 Luna, GPT-6 Sol, Claude Sonnet 5, and Gemini 3.8 Flash: 30 of 40 replies passed our checks. Below, each prompt with its replies side by side: whether each passed, what it counted as on Pro and what it cost to run. 4 of the 10 prompts split the models.

A real reply from these prompts' runs

GPT-6 Luna

Here's a job ad and my resume summary. List the 6 most important skills or keywords from the ad that my summary doesn't mention, most important first, as a numbered list. Only use words from the ad. Job ad: Marketing Analyst. You'll build weekly dashboards in Looker, run A/B tests on email campaigns, and turn customer data into clear recommendations. Must have: SQL, Excel, experience with A/B testing. Nice to have: Python, HubSpot. You'll work closely with the product and sales teams. My summary: Marketing coordinator with 3 years' experience running email campaigns and social media, building reports in Excel, and presenting results to the sales team.

GPT-6 Luna · passed our check · September 28, 2026

1. SQL
2. experience with A/B testing
3. Looker
4. weekly dashboards
5. customer data
6. clear recommendations

Ask your own question: sign up free for 5 messages on 14 of the models, no card.

Start free

Test runs checked . Updated .

Short answer

Claude Sonnet 5 passed the most of these prompts: 9 of 10, in our runs on September 28, 2026. The clearest split: “Duties into achievements, no invented numbers”, passed by Claude Sonnet 5 and Gemini 3.8 Flash and failed by GPT-6 Luna and GPT-6 Sol.

What these prompts are for

Prompts for the resume you already have: turning duties into achievements, finding the keywords a job ad wants, writing a summary, cutting to your best bullets, explaining a gap, opening a cover letter, checking for applicant tracking problems, preparing interview answers, writing a LinkedIn headline and grouping your skills.

People search for these as ChatGPT prompts; we ran them on two GPT-6 models, Claude Sonnet 5 and Gemini 3.8 Flash, and checked above all that nothing was invented.

The prompts at a glance

Every prompt on every model, as our check scored the reply. Tap a prompt to jump to it and read the replies.

Every prompt on every model: passed or failed
PromptGPT-6 LunaGPT-6 SolClaude Sonnet 5Gemini 3.8 Flash
Duties into achievements, no invented numbersFailedFailedPassedPassed
Keywords the job ad wantsPassedPassedPassedPassed
A three-sentence summary from your factsPassedFailedFailedFailed
Cut five bullets to the best threePassedPassedPassedPassed
Explain a career gap in one sentencePassedPassedPassedPassed
A cover letter's first paragraphFailedFailedPassedFailed
Check a resume for ATS problemsPassedPassedPassedPassed
Interview questions from your bulletsPassedPassedPassedPassed
LinkedIn headlines under 120 charactersFailedPassedPassedFailed
Group your skills and drop the misfitsPassedPassedPassedPassed
Passed7 of 107 of 109 of 107 of 10
Each model on these prompts
ModelPassedEach reply on ProCost per replyTime per reply
Claude Sonnet 5Anthropic9 of 101 message on Pro$0.00323.3 s
Gemini 3.8 FlashGoogle7 of 101 message on Pro$0.00041.5 s
GPT-6 SolOpenAI7 of 101 message on Pro$0.00122.5 s
GPT-6 LunaOpenAI7 of 101 message on Pro$0.00001.9 s
Passed: of the prompts each model answered, how many replies passed their check. Each reply on Pro: what one of these messages counts as on llmwise's Pro plan. Cost: what OpenRouter charged us per reply, on average; in llmwise you pay per message, not per token.

Where the models split

The same prompt, a pass on one model and a fail on another: what failed, in the check's words and the grader's.

  • Duties into achievements, no invented numbers

    Passed: Claude Sonnet 5 and Gemini 3.8 Flash. Failed: GPT-6 Luna and GPT-6 Sol.

    Neither GPT-6 model invented a number, but both wrote bullets that describe activities rather than results, and GPT-6 Sol left no [add number] gaps where a figure would help.

    • Why GPT-6 Luna failed: Graded 3.7 of 5 on average (lowest 3).

      The grader: “Every bullet opens with a strong verb, but only the last one states a measurable result; the others describe activities, and there are no invented numbers, though only one placeholder is used where others would help (e.g., newsletter reach, customers onboarded), and 'Guided' slightly overstates 'hel”
    • Why GPT-6 Sol failed: Graded 3.7 of 5 on average (lowest 3); but missing "[add number]".

      The grader: “The reply opens every bullet with a reasonable verb and invents no figures, but its results are vague activity descriptions rather than outcomes, and it adds no [add number] placeholders where metrics would clearly help (e.g., ticket reduction, onboarding volume).”
  • LinkedIn headlines under 120 characters

    Passed: GPT-6 Sol and Claude Sonnet 5. Failed: GPT-6 Luna and Gemini 3.8 Flash.

    Gemini 3.8 Flash and GPT-6 Luna each wrote a headline with "MSc Health Informatics" and no "student", which reads as a finished degree. GPT-6 Sol and Claude Sonnet 5 marked it as in progress every time.

    • Why GPT-6 Luna failed: Graded 3.7 of 5 on average (lowest 3).

      The grader: “All five headlines are under 120 characters (the longest is about 111), but options 2 and 5 list 'MSc Health Informatics' without 'student', which could read as a finished degree, and four of the five reuse the same pipe-separated nurse/ICU/MSc/informatics pattern, with only option 3 taking a differ”
    • Why Gemini 3.8 Flash failed: Graded 3.7 of 5 on average (lowest 3).

      The grader: “All five headlines are under 120 characters, but #3 lists "MSc Health Informatics" with no student or candidate qualifier, so it reads as a finished degree; the angles are only partly varied, since #1 and #5 are near-identical and all five use the same pipe-separated structure.”
  • A three-sentence summary from your facts

    Passed: GPT-6 Luna. Failed: GPT-6 Sol, Claude Sonnet 5, and Gemini 3.8 Flash.

    The cheapest model wrote the only summary that passed: GPT-6 Luna's. Claude Sonnet 5 turned 5 teachers into "school-wide" adoption, and GPT-6 Sol and Gemini 3.8 Flash listed the facts without saying why a teacher suits instructional design.

    • Why GPT-6 Sol failed: Graded 3.7 of 5 on average (lowest 2).

      The grader: “The reply sticks to the listed facts in three sentences, but it only lists them and never presents the teaching experience as a strength for instructional design, and the first sentence is clunky and redundant.”
    • Why Claude Sonnet 5 failed: Graded 3.7 of 5 on average (lowest 3).

      The grader: “The reply sticks mostly to the given facts, but 'school-wide' overstates adoption by 5 teachers and 'instructional methods' is added to training that covered only digital tools; it ties teaching experience to instructional design reasonably well in three sentences, with minor redundancy such as 'Reg”
    • Why Gemini 3.8 Flash failed: Graded 3.7 of 5 on average (lowest 3).

      The grader: “No tools or methods are invented, though 'bi-annual' could be misread as every two years; the reply names the transition but lists teaching facts rather than tying them to instructional design strengths; it has three sentences, but the first repeats 'instructional design' clumsily.”
  • A cover letter's first paragraph

    Passed: Claude Sonnet 5. Failed: GPT-6 Luna, GPT-6 Sol, and Gemini 3.8 Flash.

    Only Claude Sonnet 5 opened with the surveyor's site visits. The other three began with a stock line ("I'm excited to apply", "I am writing to express my interest"), which the grader scored low as a hook.

    • Why GPT-6 Luna failed: Graded 3.3 of 5 on average (lowest 2).

      The grader: “The reply builds around offline mode and surveying site visits without inventing facts, though phrases like "practical, thoughtful feature" drift toward generic praise, and the stock "I'm excited to apply" opening gives no real hook.”
    • Why GPT-6 Sol failed: Graded 3.7 of 5 on average (lowest 2).

      The grader: “The reply centres on offline mode and site visits without flattery or invented facts, but the opening is generic and the payoff is flat.”
    • Why Gemini 3.8 Flash failed: Graded 3.3 of 5 on average (lowest 2).

      The grader: “The reply centres on offline mode and site visits, though it stays vague with lines like 'dependable functionality is critical'; it invents little but leans on stock phrases like 'passion for the product'; and it opens with the generic 'I am writing to express my interest', which gives the reader li”

The 10 prompts, with every reply

Copy a prompt, replace the parts in [brackets] with your own, and send it to any model. Below each, the replies we got with our example filled in.

1. Duties into achievements, no invented numbers

Hard · 2 of 4 passed

Models love to add impressive figures you never gave them; a placeholder keeps every number on your resume one you can defend.

The prompt: replace the parts in [brackets]

Rewrite these resume bullets so each starts with a strong verb and shows a result. Don't invent numbers: where a number would help, write [add number] instead. [paste your bullets]

Open in llmwise
How we filled it in and checked the replies
Rewrite these resume bullets so each starts with a strong verb and shows a result. Don't invent numbers: where a number would help, write [add number] instead.

- Responsible for the company newsletter
- Helped with onboarding new customers
- In charge of the social media accounts
- Worked on reducing support tickets
  • Verbs and results: Each bullet starts with a strong verb and points to a result.
  • No invented numbers: No figures that aren't in the original; placeholders instead.
  • Honest: Doesn't inflate what the person did.
  • Automatic rules: exactly 4 bullet points; includes “[add number]”.
  • GPT-6 Luna

    Failed

    Graded 3.7 of 5 on average (lowest 3).

    - **Produced** the company newsletter, keeping employees informed about company updates and initiatives.
    - **Guided** new customers through onboarding, helping them get started with the product.
    - **Managed** social media accounts, maintaining a consistent presence across channels.
    - **Improved** support processes to reduce customer support tickets by **[add number]**.

    1 message on Pro · under $0.0001 to run · 1.4 s · 74 tokens out

  • GPT-6 Sol

    Failed

    Graded 3.7 of 5 on average (lowest 3); but missing "[add number]".

    - Produced the company newsletter, keeping employees informed about company updates.
    - Supported new-customer onboarding, helping customers get started.
    - Managed the company’s social media accounts, maintaining a consistent brand presence.
    - Advanced efforts to reduce support tickets, addressing issues that generated repeat requests.

    1 message on Pro · $0.0018 to run · 4.7 s · 281 tokens out

  • Claude Sonnet 5

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    Here are the rewritten bullets:
    
    - **Wrote and distributed** the company newsletter, growing subscriber engagement by [add number]%
    - **Streamlined** onboarding process for new customers, reducing time-to-first-value by [add number] days
    - **Managed** social media accounts, increasing follower engagement by [add number]%
    - **Reduced** support tickets by [add number]% through [briefly note method, e.g., improved documentation/proactive troubleshooting]

    1 message on Pro · $0.0027 to run · 3.5 s · 156 tokens out

  • Gemini 3.8 Flash

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    * **Authored** and distributed the company newsletter to [add number] subscribers, driving an open rate of [add number]%.
    * **Facilitated** onboarding for [add number] new customers, reducing time-to-value by [add number]% and improving initial adoption.
    * **Managed** [add number] social media channels and grew audience engagement by [add number]% through a targeted content strategy.
    * **Resolved** recurring customer issues by creating self-service resources, decreasing support ticket volume by [add number]%.

    1 message on Pro · $0.0004 to run · 1.6 s · 111 tokens out

2. Keywords the job ad wants

Everyday · 4 of 4 passed

Comparing the ad with your own text, and only the ad's words, shows exactly what a screener will look for and not find.

The prompt: replace the parts in [brackets]

Here's a job ad and my resume summary. List the 6 most important skills or keywords from the ad that my summary doesn't mention, most important first, as a numbered list. Only use words from the ad. Job ad: [paste the job ad] My summary: [paste your summary]

Open in llmwise
How we filled it in and checked the replies
Here's a job ad and my resume summary. List the 6 most important skills or keywords from the ad that my summary doesn't mention, most important first, as a numbered list. Only use words from the ad.

Job ad:
Marketing Analyst. You'll build weekly dashboards in Looker, run A/B tests on email campaigns, and turn customer data into clear recommendations. Must have: SQL, Excel, experience with A/B testing. Nice to have: Python, HubSpot. You'll work closely with the product and sales teams.

My summary:
Marketing coordinator with 3 years' experience running email campaigns and social media, building reports in Excel, and presenting results to the sales team.
  • From the ad: Every item appears in the ad.
  • Actually missing: None is already in the summary (not Excel, email campaigns or the sales team).
  • Order: The must-haves come before the nice-to-haves.
  • Automatic rules: exactly 6 bullet points; includes “SQL”; includes “A/B”; includes “Looker”.
  • GPT-6 Luna

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. SQL
    2. experience with A/B testing
    3. Looker
    4. weekly dashboards
    5. customer data
    6. clear recommendations

    1 message on Pro · $0.0001 to run · 3.2 s · 141 tokens out

  • GPT-6 Sol

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. SQL
    2. A/B testing
    3. Looker
    4. dashboards
    5. customer data
    6. recommendations

    1 message on Pro · $0.0017 to run · 2.8 s · 250 tokens out

  • Claude Sonnet 5

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. SQL
    2. Looker
    3. A/B testing
    4. Python
    5. HubSpot
    6. Product

    1 message on Pro · $0.0018 to run · 2.2 s · 39 tokens out

  • Gemini 3.8 Flash

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. SQL
    2. A/B testing
    3. Looker
    4. Python
    5. HubSpot
    6. Dashboards

    1 message on Pro · $0.0002 to run · 0.9 s · 28 tokens out

3. A three-sentence summary from your facts

Everyday · 1 of 4 passed

"Using only the facts below" stops the model adding tools and titles you don't have.

The prompt: replace the parts in [brackets]

Write a 3-sentence resume summary for [who you are], using only the facts below. [your facts]

Open in llmwise
How we filled it in and checked the replies
Write a 3-sentence resume summary for a teacher moving into instructional design, using only the facts below.

8 years teaching high-school science
Designed a flipped-classroom unit used by 5 teachers in my school
Built courses in Google Classroom and Canva
Completed a certificate in instructional design (2026)
Led staff training on digital tools twice a year
  • Only the facts: Nothing invented, such as tools or methods not listed.
  • Frames the switch: Presents the teaching experience as a strength for instructional design.
  • Concise: Three tight sentences.
  • Automatic rules: at most 3 sentences; includes “8 years”.
  • GPT-6 Luna

    Passed

    Graded 4.0 of 5 on average (lowest 3).

    High-school science teacher with 8 years of experience designing engaging learning experiences and building courses in Google Classroom and Canva. Created a flipped-classroom unit adopted by five teachers at the school and led staff training on digital tools twice a year. Completed an instructional design certificate in 2026 and is transitioning into instructional design.

    1 message on Pro · under $0.0001 to run · 1.3 s · 67 tokens out

  • GPT-6 Sol

    Failed

    Graded 3.7 of 5 on average (lowest 2).

    High-school science teacher with 8 years of experience transitioning into instructional design and a certificate in instructional design (2026). Designed a flipped-classroom unit adopted by 5 teachers and built courses in Google Classroom and Canva. Led staff training on digital tools twice a year.

    1 message on Pro · $0.0007 to run · 1.1 s · 58 tokens out

  • Claude Sonnet 5

    Failed

    Graded 3.7 of 5 on average (lowest 3).

    Dedicated high-school science teacher with 8 years of classroom experience, transitioning into instructional design after completing a certificate in the field (2026). Designed a flipped-classroom unit adopted by 5 teachers school-wide and built courses using Google Classroom and Canva, demonstrating strong instructional content development skills. Regularly led biannual staff training sessions on digital tools, reflecting a proven ability to support educator adoption of new technology and instructional methods.

    1 message on Pro · $0.0027 to run · 3.2 s · 154 tokens out

  • Gemini 3.8 Flash

    Failed

    Graded 3.7 of 5 on average (lowest 3).

    High school science educator with 8 years of teaching experience and a 2026 instructional design certificate transitioning into instructional design. Proven background in digital learning development, including creating courses in Google Classroom and Canva, as well as a flipped-classroom unit adopted by 5 colleagues. Skilled in professional development, demonstrated by leading bi-annual staff training sessions on digital tools.

    1 message on Pro · $0.0003 to run · 1.3 s · 74 tokens out

4. Cut five bullets to the best three

Everyday · 4 of 4 passed

Choosing is harder than writing; asking for the strongest three for one role gets the model to judge, not just trim.

The prompt: replace the parts in [brackets]

Cut these bullets to the 3 strongest for a [the role you want] application, each under 20 words. Keep their facts. [paste your bullets]

Open in llmwise
How we filled it in and checked the replies
Cut these bullets to the 3 strongest for a project manager application, each under 20 words. Keep their facts.

- Coordinated a 12-person team to launch a new booking system two weeks early
- Attended weekly meetings with stakeholders
- Cut vendor costs 15% by renegotiating three contracts
- Answered emails from customers
- Created a risk log that the whole department now uses
  • The strongest three: Keeps the booking-system launch, the 15% vendor saving and the risk log.
  • Under 20 words: Each bullet is under 20 words.
  • Keeps the facts: The numbers and facts are unchanged.
  • Automatic rules: exactly 3 bullet points; includes “12”; includes “15%”; includes “risk log”.
  • GPT-6 Luna

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    - Coordinated a 12-person team to launch a new booking system two weeks early.
    - Reduced vendor costs 15% by renegotiating three contracts.
    - Created a risk log now used across the department.

    1 message on Pro · under $0.0001 to run · 1.1 s · 47 tokens out

  • GPT-6 Sol

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    - Led a 12-person team to launch a new booking system two weeks early.
    - Reduced vendor costs 15% by renegotiating three contracts.
    - Created a risk log now used across the department.

    1 message on Pro · $0.0007 to run · 1.4 s · 46 tokens out

  • Claude Sonnet 5

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    - Coordinated a 12-person team to launch a new booking system two weeks early
    - Cut vendor costs 15% by renegotiating three contracts
    - Created a risk log now adopted department-wide

    1 message on Pro · $0.0018 to run · 1.8 s · 66 tokens out

  • Gemini 3.8 Flash

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    * Coordinated a 12-person team to launch a new booking system two weeks ahead of schedule.
    * Cut vendor costs by 15% through the renegotiation of three key contracts.
    * Designed and implemented a department-wide risk log to standardize mitigation tracking.

    1 message on Pro · $0.0003 to run · 1.1 s · 57 tokens out

5. Explain a career gap in one sentence

Everyday · 4 of 4 passed

One plain sentence is all a gap needs; "no apology" is the instruction that keeps it confident.

The prompt: replace the parts in [brackets]

Write one sentence for my resume explaining a [how long] gap, when I was [what you were doing]. Honest and brief, with no apology.

Open in llmwise
How we filled it in and checked the replies
Write one sentence for my resume explaining a 14-month gap, when I was caring for a parent after a stroke, and took an online data course near the end. Honest and brief, with no apology.
  • Honest: States the caring role plainly.
  • No apology: No apologetic or defensive wording.
  • Looks ahead: Mentions the data course.
  • Automatic rules: at most 35 words; at most 1 sentence; doesn't use “sorry”; doesn't use “apolog”; doesn't use “unfortunately”.
  • GPT-6 Luna

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    Took a 14-month career break to care for a parent recovering from a stroke and completed an online data course toward the end.

    1 message on Pro · under $0.0001 to run · 1.1 s · 31 tokens out

  • GPT-6 Sol

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    Took a 14-month career break to care for a parent following a stroke and completed an online data course toward the end of that period.

    1 message on Pro · $0.0005 to run · 1.0 s · 33 tokens out

  • Claude Sonnet 5

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    "Took a 14-month leave from paid work to care for a parent recovering from a stroke, completing an online data analysis certification near the end of this period."

    1 message on Pro · $0.0015 to run · 2.7 s · 48 tokens out

  • Gemini 3.8 Flash

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    **Career Break** | *[Start Date] – [End Date]*  
    Stepped away to manage full-time family caregiving and rehabilitation following a parent's stroke, concluding the period with advanced coursework in data analytics.

    1 message on Pro · $0.0002 to run · 1.2 s · 47 tokens out

6. A cover letter's first paragraph

Everyday · 1 of 4 passed

One real detail about the company does more than any adjective; the prompt makes the model build the opening around it.

The prompt: replace the parts in [brackets]

Write the first paragraph of a cover letter, under [length], for [the role] at [the company]. Mention [something specific] and why it makes me want to work there. Use only the facts I gave.

Open in llmwise
How we filled it in and checked the replies
Write the first paragraph of a cover letter, under 90 words, for a junior product designer role at Fieldnote, the note-taking app. Mention their offline mode, which I rely on for site visits as a surveyor and why it makes me want to work there. Use only the facts I gave.
  • Specific: Built around the offline mode and the surveying site visits.
  • Genuine: No generic flattery or invented experience.
  • Hook: Makes the reader want to read on.
  • Automatic rules: at most 90 words; includes “Fieldnote”; includes “offline”.
  • GPT-6 Luna

    Failed

    Graded 3.3 of 5 on average (lowest 2).

    I’m excited to apply for the junior product designer role at Fieldnote. As a surveyor, I rely on Fieldnote’s offline mode during site visits, so I’ve experienced firsthand how valuable it is to have a note-taking app that works when I need it. That practical, thoughtful feature is one reason I’d love to contribute to the product.

    1 message on Pro · under $0.0001 to run · 3.9 s · 75 tokens out

  • GPT-6 Sol

    Failed

    Graded 3.7 of 5 on average (lowest 2).

    Dear Fieldnote team, I’m excited to apply for the junior product designer role. As a surveyor, I rely on Fieldnote’s offline mode during site visits. That experience makes me want to work on the app: I know firsthand how valuable it is to be able to take notes when I’m offline.

    1 message on Pro · $0.0007 to run · 1.3 s · 67 tokens out

  • Claude Sonnet 5

    Passed

    Graded 4.0 of 5 on average (lowest 3).

    As a surveyor, I spend most of my workdays in the field, often without reliable signal, which is why Fieldnote's offline mode has become an essential part of my toolkit for capturing notes during site visits. Using it firsthand has shown me how thoughtful design can solve real, practical problems for people working outside a typical office setting. That experience is exactly what draws me to apply for the Junior Product Designer role — I want to help build tools that work as hard as the people relying on them.

    1 message on Pro · $0.0026 to run · 2.7 s · 152 tokens out

  • Gemini 3.8 Flash

    Failed

    Graded 3.3 of 5 on average (lowest 2).

    I am writing to express my interest in the Junior Product Designer role at Fieldnote. As a surveyor, I rely on Fieldnote’s offline mode during site visits, where dependable functionality is critical. Experiencing firsthand how effectively the app supports users in the field inspired me to contribute directly to its design, and I would love the opportunity to bring my perspective and passion for the product to your team.

    1 message on Pro · $0.0003 to run · 1.2 s · 80 tokens out

7. Check a resume for ATS problems

Hard · 4 of 4 passed

"Only problems that are actually in it" stops a generic checklist; you want the problems in your file.

The prompt: replace the parts in [brackets]

Check this resume text for things that often trip up applicant tracking systems. List only problems that are actually in it, each with a fix. [paste your resume text]

Open in llmwise
How we filled it in and checked the replies
Check this resume text for things that often trip up applicant tracking systems. List only problems that are actually in it, each with a fix.

JORDAN LEE | [photo] | jordan.lee@email.com

~ My Journey ~
2021–now: Senior Barista → Shift Lead, Bean There Café (see table below)
[TABLE: Skills | Level; Latte art | ★★★★☆; Scheduling | ★★★☆☆]

Stuff I'm Good At: people, coffee, rotas

References available on request
  • Finds the real problems: The photo, the table, the star ratings and the non-standard headings.
  • Nothing invented: Doesn't flag problems that aren't there, such as fonts or columns not shown.
  • Fixes: A concrete fix for each.
  • GPT-6 Luna

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    - **Photo included:** ATS parsers may not handle images. Remove `[photo]`.
    - **Table used for skills:** Tables can scramble or omit text. Replace it with a plain-text skills list, such as `Skills: Latte art, scheduling`.
    - **Decorative heading:** `~ My Journey ~` may parse poorly or fail to identify the section. Use a standard heading like `Experience`.
    - **Vague skills wording:** `Stuff I'm Good At: people, coffee, rotas` uses informal phrasing and broad terms. Use a standard `Skills` heading and specific terms, such as `Customer service, coffee preparation, staff scheduling`.
    - **“See table below” reference:** The experience entry points to content in a separate table. Put relevant details directly under the role instead.

    1 message on Pro · $0.0001 to run · 2.0 s · 165 tokens out

  • GPT-6 Sol

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    - **Photo placeholder:** `[photo]` can disrupt parsing and adds no searchable qualification. **Fix:** Remove it.
    - **Nonstandard section headings:** “My Journey” and “Stuff I’m Good At” may not be recognized as experience and skills sections. **Fix:** Use “Work Experience” and “Skills.”
    - **Two roles compressed into one line:** `Senior Barista → Shift Lead` doesn’t show when each role began. **Fix:** List each title separately with its own dates under Bean There Café.
    - **Ambiguous date:** `2021–now` gives no start month and uses “now” instead of a standard end date. **Fix:** Use `Month 2021–Present` if the month is known.
    - **Table-style skills and star ratings:** The table notation and `★★★★☆` ratings may parse poorly and provide weak searchable text. **Fix:** Use a plain-text skills list without ratings.
    - **Vague skills:** “people, coffee, rotas” may miss terms employers search for. **Fix:** Replace them with accurate, specific terms such as “customer service,” “espresso preparation,” and “staff scheduling.”

    1 message on Pro · $0.0025 to run · 5.4 s · 409 tokens out

  • Claude Sonnet 5

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    Here are the ATS problems actually present in this resume:
    
    1. **Photo included** — `[photo]` — ATS parsers can choke on embedded images, and many systems flag/strip them. **Fix:** Remove the photo entirely; it also invites bias issues.
    
    2. **Table for skills** — `[TABLE: Skills | Level...]` — Tables often parse incorrectly or get scrambled/dropped by ATS. **Fix:** List skills as plain bullet points instead (e.g., "Latte art (advanced), Scheduling (intermediate)").
    
    3. **Non-standard section headers** — "~ My Journey ~" and "Stuff I'm Good At" — ATS looks for standard headers like "Experience" and "Skills" to categorize content correctly. **Fix:** Rename to "Work Experience" and "Skills."
    
    4. **Star ratings for skill levels** — ★★★★☆ — Symbols like this are often unreadable/meaningless to ATS text parsers. **Fix:** Use plain text proficiency levels (e.g., "Advanced," "Intermediate").
    
    5. **Vague, non-keyword job title/skills phrasing** — "Stuff I'm Good At: people, coffee, rotas" — ATS matches specific keywords from job postings; casual phrasing won't match. **Fix:** Use concrete terms like "Customer Service, Barista Skills, Staff Scheduling."
    
    6. **Combined title with arrow** — "Senior Barista → Shift Lead" — the arrow symbol may not parse and obscures two distinct roles/dates. **Fix:** List as two separate entries or write "Senior Barista, promoted to Shift Lead" with plain text.
    
    7. **"References available on request"** — not a parsing error per se, but wastes space ATS could use for keywords, and adds no value. **Fix:** Remove it.
    
    8. **Missing phone number and location** — only email is listed. **Fix:** Add phone number and city/state, as many ATS fields expect them for contact parsing.

    1 message on Pro · $0.0080 to run · 7.1 s · 673 tokens out

  • Gemini 3.8 Flash

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    Here are the ATS issues found in the text and how to fix them:
    
    1. **Embedded Photo (`[photo]`)**
       * **Problem:** ATS parsers cannot read images and may either crash, reject the file, or misalign the surrounding text.
       * **Fix:** Remove the photo entirely. Resumes for ATS should contain text only.
    
    2. **Unconventional Section Headings (`~ My Journey ~`, `Stuff I'm Good At:`)**
       * **Problem:** ATS algorithms rely on standard keywords to categorize your experience and skills. Decorative symbols (`~`) and non-standard phrasing confuse the parser.
       * **Fix:** Use standard headings like **Work Experience** (or *Professional Experience*) and **Skills**.
    
    3. **Tables and Special Characters (`[TABLE: ...]`, `★`)**
       * **Problem:** Parsers frequently read tables out of order, scramble text across columns, and cannot interpret graphical ratings (stars, rating bars).
       * **Fix:** Remove the table and star ratings. List your skills as simple bullet points or a comma-separated list of standard text terms.
    
    4. **Vague and Non-Standard Keywords (`people, coffee, rotas`)**
       * **Problem:** ATS matches your resume against specific job description keywords. Casual phrases like "people" and "rotas" will be skipped by keyword searches.
       * **Fix:** Use recognized industry terms, such as: *Customer Service*, *Espresso & Coffee Preparation*, and *Staff Scheduling*.
    
    5. **Cluttered Role Progression in a Single Line (`Senior Barista → Shift Lead`)**
       * **Problem:** Combining two distinct titles with an arrow symbol (`→`) makes it hard for the parser to determine your actual job title, dates, and career level.
       * **Fix:** Separate the roles clearly:
         * **Shift Lead**, Bean There Café (*Date – Present*)
         * **Senior Barista**, Bean There Café (*Date – Date*)
    
    6. **Outdated Reference Line (`References available on request`)**
       * **Problem:** Wastes scannable space and offers no keyword value; modern ATS algorithms and recruiters expect references only later in the interview stage.
       * **Fix:** Delete this line completely.

    1 message on Pro · $0.0011 to run · 3.2 s · 478 tokens out

8. Interview questions from your bullets

Everyday · 4 of 4 passed

Questions drawn from your own bullets are the ones you'll actually get; practise answering them out loud.

The prompt: replace the parts in [brackets]

Based on my resume bullets below, write [how many] questions a hiring manager is likely to ask me, each tied to a bullet, as a numbered list. [paste your bullets]

Open in llmwise
How we filled it in and checked the replies
Based on my resume bullets below, write 5 questions a hiring manager is likely to ask me, each tied to a bullet, as a numbered list.

- Cut monthly cloud costs 22% by moving batch jobs to spot instances
- Led the migration of 40 services to Kubernetes over 6 months
- Mentored 3 junior engineers, two promoted within a year
- Wrote the on-call runbook used by 4 teams
  • Tied to bullets: Each question is clearly about one of the bullets.
  • Probing: Asks how, why or about trade-offs, not just "tell me about".
  • Realistic: Questions a hiring manager would really ask.
  • Automatic rules: exactly 5 bullet points.
  • GPT-6 Luna

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. **Cut monthly cloud costs 22% by moving batch jobs to spot instances:** How did you identify which jobs were good candidates, and how did you handle interruptions or reliability risks?
    2. **Led the migration of 40 services to Kubernetes over 6 months:** How did you plan and sequence the migration, and what was the biggest challenge?
    3. **Mentored 3 junior engineers, two promoted within a year:** How did you support their growth, and what contributions did you make to their promotions?
    4. **Wrote the on-call runbook used by 4 teams:** What problems did the runbook address, and how did you make sure teams adopted and maintained it?
    5. **Cut monthly cloud costs 22% by moving batch jobs to spot instances:** How did you measure the savings and ensure the change didn’t negatively affect job completion times or service quality?

    1 message on Pro · $0.0001 to run · 1.8 s · 183 tokens out

  • GPT-6 Sol

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    1. **Cloud costs:** How did you decide which batch jobs could run on spot instances, and how did you handle interruptions?
    2. **Cloud costs:** How did you measure the 22% savings, and were there any trade-offs in reliability or performance?
    3. **Kubernetes migration:** How did you plan and coordinate the migration of 40 services over six months?
    4. **Mentorship:** What did you do to help the junior engineers grow, and how did you assess their readiness for promotion?
    5. **On-call runbook:** How did you make sure the runbook worked for all four teams, and what impact did it have on incident response?

    1 message on Pro · $0.0012 to run · 2.4 s · 162 tokens out

  • Claude Sonnet 5

    Passed

    Graded 4.0 of 5 on average (lowest 3).

    1. You cut monthly cloud costs by 22% using spot instances for batch jobs — how did you handle the risk of spot instance interruptions, and what fallback mechanisms did you build in?
    
    2. Migrating 40 services to Kubernetes in 6 months is a significant undertaking — how did you prioritize which services to migrate first, and what was the biggest technical or organizational obstacle you hit along the way?
    
    3. You mentored 3 junior engineers, two of whom were promoted within a year — what did your mentoring approach look like, and how did you measure or support their growth?
    
    4. You wrote the on-call runbook adopted by 4 teams — what gap did you notice that led you to create it, and how did you drive adoption across teams that weren't originally your own?
    
    5. Across these achievements — cost optimization, large-scale migration, mentorship, and operational documentation — which one are you most proud of, and why?

    1 message on Pro · $0.0049 to run · 4.7 s · 365 tokens out

  • Gemini 3.8 Flash

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. **(Cut monthly cloud costs 22% by moving batch jobs to spot instances):** How did you handle spot instance interruptions to ensure your batch jobs completed reliably without data loss or SLA breaches?
    2. **(Led the migration of 40 services to Kubernetes over 6 months):** How did you prioritize which services to migrate first, and what strategy did you use to manage downtime and rollbacks during the transition?
    3. **(Led the migration of 40 services to Kubernetes over 6 months):** With such an aggressive 6-month timeline across 40 services, how did you handle team resistance and upskill developers who were unfamiliar with Kubernetes?
    4. **(Mentored 3 junior engineers, two promoted within a year):** What specific framework or mentoring approach did you use to accelerate their technical growth and prepare them for promotion so quickly?
    5. **(Wrote the on-call runbook used by 4 teams):** How did you identify the critical failure modes to document, and what process did you put in place to ensure the runbook stayed accurate and up to date across all four teams?

    1 message on Pro · $0.0006 to run · 2.0 s · 237 tokens out

9. LinkedIn headlines under 120 characters

Hard · 2 of 4 passed

A character limit is where headlines fail; so is claiming a job you're still aiming for.

The prompt: replace the parts in [brackets]

Write [how many] LinkedIn headline options for me, each under 120 characters, as a numbered list. About me: [about you]

Open in llmwise
How we filled it in and checked the replies
Write 5 LinkedIn headline options for me, each under 120 characters, as a numbered list.

About me: Registered nurse, 6 years in ICU, now doing a part-time MSc in health informatics and looking for a clinical informatics role.
  • Under 120 characters: Every headline is under 120 characters.
  • Accurate: Doesn't claim a finished MSc or an informatics job title held now.
  • Varied: Five different angles.
  • Automatic rules: exactly 5 bullet points.
  • GPT-6 Luna

    Failed

    Graded 3.7 of 5 on average (lowest 3).

    1. Registered Nurse | 6 Years in ICU | MSc Health Informatics Student | Aspiring Clinical Informatics Professional
    2. ICU Nurse Transitioning to Clinical Informatics | 6 Years of Critical Care Experience | MSc Health Informatics
    3. Registered Nurse Bringing 6 Years of ICU Experience to Clinical Informatics | Part-Time MSc Student
    4. Clinical Informatics Candidate | ICU Registered Nurse | Health Informatics MSc Student
    5. Critical Care Nurse | 6 Years in ICU | MSc Health Informatics | Seeking Clinical Informatics Opportunities

    1 message on Pro · under $0.0001 to run · 1.6 s · 112 tokens out

  • GPT-6 Sol

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    1. Registered Nurse | 6 Years in ICU | MSc Health Informatics Student | Aspiring Clinical Informaticist
    2. ICU Nurse Transitioning to Clinical Informatics | Part-Time MSc in Health Informatics
    3. RN with 6 Years of ICU Experience | Bridging Bedside Care and Health Informatics
    4. Critical Care RN | Health Informatics MSc Student | Seeking a Clinical Informatics Role
    5. Registered Nurse | ICU Experience + Health Informatics Training | Focused on Better Clinical Workflows

    1 message on Pro · $0.0009 to run · 1.7 s · 107 tokens out

  • Claude Sonnet 5

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    1. ICU RN (6 yrs) | MSc Health Informatics Candidate | Bridging Bedside Care & Clinical Data Systems
    2. Critical Care Nurse → Clinical Informatics | Turning Frontline Experience into Data-Driven Insights
    3. RN, ICU Veteran | Health Informatics MSc Student | Passionate About EHR Optimization & Patient Safety
    4. Blending 6 Years of ICU Nursing with Health Informatics to Improve Clinical Workflows
    5. ICU Nurse | MSc Health Informatics (Part-Time) | Seeking Clinical Informatics Opportunities

    1 message on Pro · $0.0033 to run · 3.4 s · 221 tokens out

  • Gemini 3.8 Flash

    Failed

    Graded 3.7 of 5 on average (lowest 3).

    1. Critical Care RN | MSc Health Informatics Student | Aspiring Clinical Informatics Specialist
    2. ICU Nurse (6 Yrs) | Health Informatics MSc Candidate | Bridging Bedside Care & Digital Health
    3. Registered Nurse | MSc Health Informatics | Passionate About EHR Optimization & Clinical Workflow
    4. Experienced ICU RN transitioning to Clinical Informatics | MSc Candidate | Healthcare Technology
    5. Critical Care Nurse | Health Informatics MSc Student | Seeking Clinical Informatics Roles

    1 message on Pro · $0.0003 to run · 1.2 s · 90 tokens out

10. Group your skills and drop the misfits

Everyday · 4 of 4 passed

Asking it to say what it dropped lets you overrule it; the grouping does the rest.

The prompt: replace the parts in [brackets]

Turn my skills into a resume skills section for a [the role you want], grouped into 3 categories. Drop the ones that don't fit the role and list what you dropped at the end. Skills: [your skills]

Open in llmwise
How we filled it in and checked the replies
Turn my skills into a resume skills section for a data analyst, grouped into 3 categories. Drop the ones that don't fit the role and list what you dropped at the end.

Skills: SQL, Excel, forklift licence, Tableau, Python (pandas), public speaking, first aid, A/B testing, Spanish (fluent), latte art, statistics, Power BI
  • Right drops: Drops the forklift licence, first aid and latte art.
  • Sensible groups: Three clear categories that suit a data analyst.
  • Says what it dropped: Lists the dropped skills at the end.
  • GPT-6 Luna

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    ### Technical Tools
    SQL · Excel · Python (pandas) · Tableau · Power BI
    
    ### Analytics
    Statistics · A/B testing
    
    ### Communication
    Public speaking
    
    **Dropped:** Forklift licence, first aid, Spanish (fluent), latte art

    1 message on Pro · under $0.0001 to run · 1.3 s · 55 tokens out

  • GPT-6 Sol

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    ## Skills
    
    - **Data Analysis & Experimentation:** SQL, Python (pandas), statistics, A/B testing
    - **Reporting & Visualization:** Excel, Tableau, Power BI
    - **Communication:** Public speaking, fluent Spanish
    
    **Dropped:** Forklift licence, first aid, latte art.

    1 message on Pro · $0.0011 to run · 2.9 s · 134 tokens out

  • Claude Sonnet 5

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    **Skills**
    
    **Data Analysis & Statistics**
    - Python (Pandas)
    - Statistics
    - A/B Testing
    
    **Tools & Technologies**
    - SQL
    - Excel
    - Tableau
    - Power BI
    
    **Communication & Languages**
    - Public Speaking
    - Spanish (Fluent)
    
    **Dropped (not relevant to data analyst role):**
    - Forklift licence
    - First aid
    - Latte art

    1 message on Pro · $0.0026 to run · 2.0 s · 146 tokens out

  • Gemini 3.8 Flash

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    ### **Skills**
    
    * **Data Analysis & Statistics:** SQL, Python (pandas), Excel, Statistics, A/B Testing
    * **Business Intelligence & Visualization:** Tableau, Power BI
    * **Communication & Languages:** Public Speaking, Spanish (Fluent)
    
    ***
    
    **Dropped Skills (Not relevant to a Data Analyst role):**
    * Forklift licence
    * First aid
    * Latte art

    1 message on Pro · $0.0003 to run · 1.3 s · 81 tokens out

Getting more from these prompts

  • Tell the model not to invent numbers, and to leave a placeholder instead; then fill it with a figure you can defend.

  • Paste the job ad itself: the ad's own words are what a screener searches for.

  • Ask for the strongest three bullets, not a shorter version of all of them.

  • Read every line before you send it: a claim you can't explain costs more in an interview than it gains on paper.

How we ran and checked them

Each prompt was sent the way llmwise sends a message: the app's own system prompt, each model's own settings, and Pro's reply size limit (8,000 tokens), through OpenRouter. Every reply is shown as it came.

Rubric (graded). The grader model scores the reply from 1 to 5 on each published criterion. It passes with an average of 4 or more and no criterion under 3, and only if it also meets the prompt's automatic rules (length, words it must or mustn't use).

The grader is Claude Opus 5.5 at low reasoning effort; its own replies are graded by GPT-6 Astra, so no model grades itself. Its prompt is on our test runs page.

More tested prompts

Questions

Will AI make things up on my resume?

Several prompts here tell the model not to invent anything: no numbers you didn't give, no titles you don't hold. The replies show which models keep to that. Every fact on your resume should be one you can explain in an interview.

Is this a resume builder?

No. These are prompts for editing the resume you have: rewriting bullets, tailoring to an ad, trimming, summarizing and preparing for the interview. They don't build a resume from nothing.

Which AI is best for resume prompts?

We ran them on GPT-6 Luna, GPT-6 Sol, Claude Sonnet 5, and Gemini 3.8 Flash. Claude Sonnet 5 passed the most: 9 of 10. The same prompts work in any chat app, ChatGPT included.

Claude, GPT, Gemini, DeepSeek, Grok, Kimi, and GLM, in one chat.

See what a message costs before you send it. Free is 5 messages to try; sign in with an email link, no password or card.