---
url: "https://youtube.com/watch?v=N1Fm4JYae1I&is=R0OA9PgMYMUM3Ebd"
title: "Write a Killer Résumé in the AI Era: 5 New Rules!"
source_kind: youtube
author: Jeff Su
captured: "2026-09-16T03:13:33+00:00"
comment_tree: false
topics: [other]
summary: Jeff Su shares five research-backed rules for writing resumes that pass both AI screening and human review, emphasizing simple formatting, tailored content, quantified impact, and proof of AI skills.
status: ok
---

# Write a Killer Résumé in the AI Era: 5 New Rules!

Channel: Jeff Su

## Transcript

[0s] Finding reliable resume advice online
[2s] was already hard enough. And now that
[3s] employers are using AI to screen job
[6s] applications, it's only fair to ask, "Do
[8s] resumes even matter anymore?" And so, I
[10s] spent weeks analyzing hiring reports and
[13s] academic studies from institutions like
[15s] MIT and Oxford, covering more than 4,000
[18s] hiring managers and nearly 2 million
[20s] applications. And bottom line up front,
[22s] I found that resumes still matter, but
[25s] the rules have changed. In this video,
[27s] I'll start off by sharing the five most
[29s] important findings and what each one
[31s] means for you. Then, I'll break down
[33s] exactly what you need to do to stand out
[35s] in the age of AI. Let's get started.
[37s] Kicking things off with the five key
[39s] findings. First, 87% of hiring managers
[42s] say their AI hiring software reads
[44s] simple, text-based resumes more
[46s] accurately than fancy, visual ones.
[49s] Second, tailored resumes saw 84% higher
[52s] interview rates, but resumes with the
[54s] most keywords actually performed worse.
[57s] Third, 59% of hiring managers see
[60s] candidates using AI as a good sign, but
[62s] 28% reject AI-heavy resumes that show
[66s] little to no effort. Fourth, resumes
[68s] that quantified a candidate's impact saw
[71s] 75% higher interview rates than those
[74s] that only described responsibilities.
[76s] And fifth, 60% of hiring managers want
[79s] candidates to prove their AI skills, not
[82s] just list them. I've linked the sources
[83s] I went through down below, along with
[85s] the prompts and bullet point examples I
[87s] showcased in this video. Now, before
[89s] diving into each finding, let's quickly
[91s] cover the two stages your resume has to
[92s] pass in this new era. Pass the AI and
[95s] pass the human. Stage one, pass the AI.
[98s] Basically, companies use AI to read,
[100s] rank, and organize applications. So,
[102s] your resume needs to be easy to process
[104s] and clearly show how your experience
[107s] matches the role. Stage two, pass the
[109s] human. So, if we squint really hard,
[112s] there is some good news in that only 6%
[115s] of hiring managers say AI is the
[117s] decision-maker when it comes to passing
[118s] or rejecting candidates. Meaning, humans
[121s] still play a central role in hiring. So,
[124s] your resume needs to take the human
[125s] reader into account as well. Put simply,
[128s] passing the AI gets your foot in the
[130s] door, and passing the human gets your
[132s] application moved forward. And the five
[134s] rules we're about to cover address both
[136s] parts of that equation. On that note,
[138s] rule number one, make sure the AI can
[140s] read your resume. For context, while 87%
[143s] of employers say their AI hiring
[144s] software prefers simple, text-based
[146s] resumes, an explosion of AI design tools
[149s] actively encourages job seekers to stand
[152s] out with graphics, icons, and large
[153s] images. Since it's much easier to sell
[156s] you a fancy resume template than to tell
[158s] you boring is best. Therefore, we should
[160s] be optimizing our resumes for
[162s] readability, not appearance. And
[165s] although there is no universal file size
[167s] limit across every hiring system, some
[169s] tools explicitly state they cannot parse
[172s] resumes larger than 2.5 megabytes. And
[175s] high-quality images are one common
[177s] reason files cross that limit. All this
[179s] to say, even if you have all the right
[181s] qualifications for a role, none of that
[183s] matters if the AI software can't even
[186s] read your resume properly. So, here's
[189s] what to do. First, keep the formatting
[191s] intentionally boring. Use one column and
[193s] conventional headings like summary,
[195s] experience, education, and skills. And
[197s] for God's sake, avoid skill bars like
[199s] it's the plague. The one caveat is to
[201s] follow local norms. For example, in
[203s] China, headshots are very common on
[206s] resumes, and many employers even ask for
[208s] a candidate's zodiac sign. True story.
[211s] Second, export your resume as a
[212s] selectable text PDF, ideally below 2.5
[216s] megabytes. A quick check is to open the
[218s] PDF and see if you can highlight and
[220s] copy the text. If you can't, the text
[222s] may be trapped inside an image. Moving
[224s] on to rule number two, make your fit
[226s] obvious. As a quick reminder, studies
[228s] found that tailored resumes saw 84%
[231s] higher interview rates, but resumes with
[233s] the most keywords actually performed
[235s] worse, which sounds contradictory until
[238s] you learn the difference between keyword
[239s] mapping and keyword stuffing. Keyword
[241s] mapping is when you use relevant
[243s] keywords from the job description to
[245s] describe work you've actually done,
[247s] which is not new advice, but AI makes
[250s] this process much faster. Keyword
[252s] stuffing is when job seekers cram in
[254s] every phrase from the job description,
[256s] whether or not their experience backs it
[258s] up, and AI makes this process much
[261s] faster as well. Taking a closer look at
[262s] the data, we see that across nearly 2
[265s] million applications, untailored resumes
[267s] had a 3.09% interview rate versus 5.71%
[271s] for tailored resumes, which is also
[273s] where the 84% increase from earlier
[275s] comes from. In that same study, we also
[277s] see resumes with the highest keyword
[279s] coverage receive 21% fewer interviews
[283s] compared to those with moderate
[284s] coverage. And so, in other words, using
[287s] AI to over-optimize can do more harm
[290s] than good. For example, imagine you're
[292s] applying for a customer success role. A
[294s] keyword-stuffed bullet might read,
[296s] "Delivered customer support, customer
[298s] service, and customer-focused
[299s] communication to improve the customer
[301s] experience." I wish I was kidding, but I
[303s] legitimately read something like that
[304s] during my time at Google. Whereas, a
[306s] thoughtfully mapped bullet would say,
[308s] "Reduced customer complaints by 31%
[311s] using AI to create an onboarding email
[313s] sequence from help center documents."
[315s] So, here's how to use AI to tailor your
[317s] resume faster without keyword stuffing.
[320s] First, give AI the job description and
[322s] your base resume. Ask it to identify the
[325s] problems the employer needs solved,
[327s] extract the skills and keywords that
[329s] matter most, and only then ask AI to
[332s] suggest stronger bullet points. Next,
[334s] review every suggestion and keep only
[336s] what you can prove. Meaning, if you
[338s] can't point to a real experience that
[339s] supports a keyword, remove it. So,
[341s] basically, the AI is smart enough to
[343s] figure out whether you're just adding
[345s] random keywords for no reason, or
[347s] including keywords intentionally. And
[349s] that completes stage one. The AI can now
[352s] process your resume and see why you
[354s] match the role. But, you still need to
[356s] convince the human to pass you, which
[358s] brings us to stage two. By the way, if
[360s] you want to get 1% better at AI every
[362s] week, sign up for my free newsletter to
[364s] receive one insanely actionable AI tip
[366s] you can apply in under two minutes. Link
[368s] down below. Next up, rule number three,
[371s] know where AI should stop. Remember, 59%
[374s] of hiring managers like to see
[376s] candidates using AI, but 28% reject
[380s] AI-heavy resumes that show little to no
[382s] effort. And the research shows why both
[385s] can be true. A randomized experiment
[387s] from MIT involving nearly half a million
[389s] job seekers found that using AI to help
[391s] with spelling, grammar, and wording
[394s] increased the probability of getting
[395s] hired by 8%. That said, a separate
[398s] experiment giving applicants access to
[400s] ChatGPT made their pitches more similar
[403s] and reduced evaluators' screening by up
[406s] to 9%, meaning it became harder to
[408s] identify who was actually more
[410s] qualified. I mean, just imagine if
[412s] you're the hiring manager and you read,
[414s] "Spearheaded cross-functional
[415s] initiatives to drive operational
[417s] excellence across 20 resumes." You'd
[420s] give up as well. Put simply, the
[421s] experiments show that thoughtful AI use
[424s] is fine, even encouraged, while lazy
[426s] prompts like, "Make my resume better."
[428s] produce the same generic BS like,
[430s] "Result-driven professional with a
[432s] proven track record of success." that
[434s] hiring managers are sick and tired of
[437s] reading. So, how do you use AI
[439s] thoughtfully? Here's a simple two-step
[441s] process. First, focus on brain dumping
[443s] your raw thoughts for every work
[445s] experience you have, and just make sure
[447s] you cover three things for each. The
[449s] final result, what you specifically
[451s] contributed, and how you achieved it.
[454s] And don't worry if the notes are a
[456s] complete mess because the goal at this
[458s] step is just to capture the facts, not
[460s] make them sound good. For example, your
[462s] brain dump might look something like
[464s] this. I compared the replies from sales
[466s] and support and there were so many
[469s] conflicting answers. Uh so, I wrote a
[471s] refund guide that they can both use and
[474s] uh once both teams used it, duplicate
[476s] tickets dropped by 50% and the customers
[478s] got consistent answers. Cool. Next,
[480s] upload the job description along with
[481s] your rough notes and ask AI to
[483s] strengthen the wording for each bullet
[485s] without changing any of the underlying
[487s] facts. Then, review every suggestion and
[490s] only keep the language you could explain
[491s] naturally in an interview. So, your
[494s] final bullet point might look something
[495s] like this. Decreased duplicate tickets
[497s] by 50% by creating a shared refund guide
[500s] for sales and support so customers
[502s] received consistent answers. To quickly
[504s] recap, hiring managers and recruiters
[506s] read enough resumes to spot the
[508s] difference between thoughtful AI use and
[511s] lazy So, do the hard thinking
[513s] first, then use AI to make that thinking
[515s] more clear and concise. Up next, rule
[518s] number four, prove your impact with
[519s] numbers. In a nutshell, the research
[521s] found resumes that quantified a
[523s] candidate's impact saw 75% higher
[527s] interview rates than those that only
[528s] described responsibilities. And this
[531s] makes perfect sense since two people can
[533s] have the exact same responsibilities and
[535s] yet produce completely different
[537s] results. Just imagine a hiring manager
[539s] reading, "Responsible for planning
[540s] social media campaigns." 10 times in a
[542s] row. And the 11th resume says, "Drove a
[545s] 30% year-over-year increase in
[547s] short-form views over two months." It's
[550s] not hard to guess which candidate stands
[552s] out. Now, none of this is new. Adding
[554s] metrics to your resume has been best
[556s] practice for decades. What is new is
[559s] that AI makes the two hardest parts much
[562s] easier. Figuring out which metrics
[564s] actually matter and turning your rough
[566s] notes into polished bullet points using
[568s] a proven framework. Here's what that
[570s] means in practice. First, upload your
[572s] resume and tell AI, "Help me identify
[575s] the most relevant metrics for each
[577s] experience on my resume. Ask me
[579s] clarifying questions if needed. This
[581s] matters because a lot of candidates
[582s] assume revenue is the only metric that
[585s] counts, when there are many many other
[587s] ways to measure impact, such as time
[589s] saved, speed, scale, accuracy, etc.
[592s] Next, after you've supplied numbers you
[594s] can verify, tell the AI, rewrite each
[596s] bullet using Google's XYZ formula.
[599s] Accomplish X as measured by Y by doing
[602s] Z. Do not add or change any facts or
[605s] numbers. Building on the earlier
[607s] example, a revised bullet might look
[609s] something like this. Drove a 30%
[611s] year-on-year increase in short-form
[612s] views within 2 months by testing
[614s] different video openings and repeating
[616s] the best performing formats. So, the
[618s] takeaway here is simple. Job seekers
[620s] used to skip metrics because they either
[622s] didn't know what to measure or didn't
[625s] have the time to rewrite every bullet.
[628s] But now AI removes both of those
[630s] barriers, turning one of the highest
[632s] impact resume improvements into one of
[634s] the easiest. And finally, rule number
[636s] five, prove your AI skills. So, it
[638s] shouldn't be surprising that 60% of
[640s] hiring managers want candidates to prove
[642s] their AI skills and not just list them,
[644s] right? And when we look at how they want
[646s] that proof, 26% prefer an interview or
[649s] task, 19% want work examples or
[652s] outcomes, and 15% look for
[655s] certifications or courses. On top of
[657s] that, an experiment from Oxford
[659s] University found that adding
[660s] role-relevant AI skills increased a
[662s] candidate's chance of being selected for
[665s] an interview by up to 15 percentage
[667s] points. Put simply, writing ChatGPT in
[669s] your skills section is not enough. You
[671s] need to actually demonstrate how you use
[674s] AI to solve a problem for the role
[676s] you're applying for. So, here's what to
[677s] do. First, whenever possible, make the
[680s] first bullet under each of your
[681s] experiences an AI achievement that is
[684s] directly relevant to the job
[685s] description. For example, cut weekly
[688s] customer feedback reporting from 2 hours
[690s] to 30 minutes by using Claude Code to
[692s] build a shared database with recurring
[694s] issues. If you don't have a relevant
[696s] workplace example yet, complete a small
[698s] project for the job you want and list it
[700s] under a separate project section in your
[702s] resume. For example, created a customer
[705s] insights report from 100 public product
[707s] reviews by using chat GPT work to
[709s] identify recurring complaints, checking
[711s] the themes, and recommending three
[713s] improvements. Next, make your proof easy
[716s] for the hiring manager to inspect by
[717s] linking directly to your portfolio.
[720s] Here's a real example for my friend Zara
[721s] Zhang who publishes all her AI projects
[723s] on GitHub. So, any hiring manager can
[725s] immediately see what she built and how
[728s] she used AI. And if you have no idea
[730s] what GitHub means, no problem. Simply
[732s] create a Google Doc and under each
[734s] project, explain the situation, how and
[736s] why you used AI, and what changed as a
[739s] result. Pro tip, unless you're a fresh
[740s] graduate, put your experience above your
[743s] education since 86% hiring managers
[746s] value relevant work experience over
[748s] formal education. All right, let's
[750s] quickly recap by closing the loop on the
[751s] two stages. Rules one and two help you
[753s] pass the AI by making your resume
[755s] readable and connecting your experience
[757s] to the role. Rules three, four, and five
[760s] help you pass the human by using AI
[762s] thoughtfully, quantifying your impact,
[765s] and proving your AI skills with real
[767s] work. Now, was every finding from this
[769s] video new and mind-blowing? No, not
[771s] really. But hopefully, now that you've
[773s] seen the data behind the advice, you can
[776s] follow them with more confidence,
[778s] especially now that AI is involved on
[780s] both sides of hiring. If you enjoyed
[782s] this, you might want to check out my AI
[783s] playlist next. See you on the next video
[785s] in the meantime.
[786s] Have a great one.