Which AI is screening your job application? Part 1: What reads you first

This is Part 1 of a two-part series. It covers the machines between your application and a recruiter: what each one reads and what it does with it. Part 2 covers what the law makes employers tell you, what to do when you think the software got you wrong, and how to use AI on your side without hurting yourself.

My position, up front: AI did not create the gap between applicants with networks and applicants without them. It automated that gap. A referral still walks your name past every tool in this article and puts it in front of a person, which is why getting a referral when your network is thin is worth more of your time than another resume rewrite. Everyone else gets read by software first, and first-generation professionals are disproportionately everyone else.

How does AI screen a job application?

Your application first lands in an applicant tracking system (ATS), and more employers now run an AI layer on top of it that grades or ranks candidates against the job requirements before a recruiter looks. Around that core, AI appears at three other points: agents that search for candidates, chatbots that run first interviews, and software that scores recorded video interviews. If you are on the employer side, I compare these tools in my applicant tracking software guide.

Stage	Example tools	What the AI does	What you control
Sourcing	LinkedIn Hiring Assistant, Indeed Talent Scout	Searches profiles and resumes, builds shortlists, drafts outreach	A complete, searchable profile and resume
Resume ranking	Workday Recruiting with HiredScore	Grades applicants against the posted requirements	Matching the posting’s language and answering knockout questions accurately
Chatbot interview	Paradox’s Olivia	Screens and schedules by text	Complete answers and what personal data you type
Video interview	HireVue and similar	Records answers; some employers score the language	The content of your answers and accommodation requests

The chain exists because of volume. Greenhouse, an ATS used by thousands of employers, reports that applications per opening rose from an average of 116 in 2022 to 244 in 2025. Vendors sell their tools as fairer because they read every application. They do read every application, the way a bouncer reads every ID: quickly, against a list someone else wrote, with no interest in your story.

Does AI reject resumes automatically?

Usually not on its own; the common pattern is ranking, where AI grades every applicant and a recruiter reads from the top down, so a low grade works like a rejection you never see. Workday, which acquired HiredScore in 2024 and sells it inside Workday Recruiting, describes the product as AI-driven candidate grading, and its own product datasheet shows applicants sorted into A, B and C grades.

I will say plainly what the industry will not: a ranking is a rejection with better public relations. If 244 people apply and the recruiter interviews from the A pile, a C is a no that nobody had to say and nobody has to explain. Whether that kind of quiet no can be discrimination is now in front of a federal court, and I cover that case in Part 2.

The true automatic rejections come from knockout questions: yes-or-no screening questions about work authorization, a required license, a shift you must cover or a minimum credential. Answer one wrong and many systems drop the application before any ranking happens.

• Use the posting’s exact words for skills and titles you actually have. A grader matches requirements, and a synonym may not count.
• Read every knockout question twice. A rushed “no” to a misread question ends an application that would have earned an A grade.
• Keep plain sections, Experience, Education and Skills, so the parser files your history correctly.

For the full mechanics, read how applicant tracking systems screen candidates. An A grade only earns you a human reader, and that reader decides in seconds; here is what a recruiter sees in the first seven seconds.

What does LinkedIn’s Hiring Assistant look at?

LinkedIn’s Hiring Assistant reads your profile, your resume and your answers to screening questions, compares them to the recruiter’s qualifications, and can message you pre-screening questions by InMail. It is an AI agent, software that carries out a series of tasks on its own rather than answering one prompt, and it became globally available in English at the end of September 2025. LinkedIn says recruiters using it review 62% fewer profiles before reaching a shortlist, and that it can pre-screen candidates through InMail.

Read that 62% from the candidate’s side. It is not efficiency for you. It is human attention removed from the process. Your LinkedIn profile is no longer a networking page; it is the first round of the interview, run by software that never asks a follow-up question. Treat it that way.

• Write your headline and About section in the titles and skills recruiters search for, not your employer’s internal titles. LinkedIn says the agent saves recruiters 4+ hours per role, and none of that time goes to decoding a vague profile.
• Answer every screening question in full. A one-word answer gives the agent nothing to match.
• Treat a pre-screening InMail like a phone screen. Your replies are summarized for the recruiter.

For how recruiters search the platform directly, read how recruiters search LinkedIn.

How does Indeed’s Talent Scout find candidates?

Indeed’s Talent Scout is an AI agent for employers that searches millions of resumes and profiles, surfaces the strongest matches and drafts personalized outreach, so an employer can find your Indeed resume before you find its posting. Indeed launched it in September 2025 alongside Career Scout, an agent for job seekers, and it connects to Workday, Workable and isolved.

Here is the most useful thing any AI hiring tool has said so far. Indeed’s CEO told Fortune that Talent Scout advises employers to widen their candidate pool by raising pay, expanding location or relaxing requirements such as years of experience. When the employer’s own AI says the requirements are the problem, believe it. I make the same argument about degree requirements.

On your side: keep your Indeed resume public, current and written in the language of the jobs you want, and check any pay figure Career Scout gives you against the posted range. California employers with 15 or more employees must post one, and here is what to do when a posting leaves it out.

What is a chatbot interview?

A chatbot interview is a first-round screening run by conversational AI, usually by text message or web chat, that asks about availability, experience and eligibility and then schedules the next step. It is standard in restaurant, retail and warehouse hiring. The best-known assistant is Olivia, built by Paradox, which Workday bought in a $1.0 billion deal completed in October 2025. Workday said Paradox would schedule 32 million interviews that year, about one in ten in the United States.

I have no objection to a bot scheduling interviews. My objection is to employers treating the bot as someone else’s responsibility. In 2025, researchers found that McHire, the platform McDonald’s franchisees use, powered by Olivia, had an administrator account protected by the password “123456,” which exposed up to 64 million applicant records, chat logs included. Paradox fixed it the day it was reported. But no applicant chose Paradox. They chose the restaurant. In the applicant’s eyes that breach belongs to the employer, and it should sit in the employer’s risk register too.

• Answer what is asked, completely, and nothing more. Never type a Social Security number or bank details into a screening chat; employers collect those at hire, through payroll. In the McHire case, 64 million records sat behind one weak password.
• Screenshot your answers and the interview time the bot confirms.
• Ignore follow-up messages asking you to pay for training or equipment. Real employers do not charge applicants.

How are AI video interviews scored?

In a one-way, or asynchronous, video interview you record answers to on-screen questions with no one on the other end, and where software scores those recordings, it now scores mainly the words you say. HireVue, the best-known vendor, stopped using facial analysis in its assessments after its own research found visual analysis added little beyond the language. Many employers use the software only to record, and people watch the videos.

My view: the one-way video interview is the worst format in hiring. It asks for a performance with no audience, no follow-up and no chance to read the room, and it shifts the employer’s time cost onto the applicant. It rewards people who have rehearsed on camera before, which tracks privilege more than talent. If you have any leverage, ask for a live conversation instead.

• Lead each answer with the result, then the situation. An answer that opens with one specific number gives any scorer, human or software, something concrete to hold.
• Ask whether a person reviews the recordings and what the answers are scored on.
• If a disability makes a timed, recorded format a poor measure of you, request an accommodation, such as a live interview, before you record.

That is the machinery. Part 2 is about leverage: the notice laws, the Workday lawsuit that could make vendors answer for their grades, and how to use ChatGPT, Claude and Gemini without becoming the applicant the screen was built to catch.

McKinley holds a J.D. but is not a licensed attorney. Articles here are general information, not legal advice. For your specific situation, talk with an employment attorney.

Data current as of September 2026. Sources are linked where each figure appears.