Resume Optimization in 2026: What a Recruiter Actually Reads

I have reviewed resumes for more than 1,000 hires across automotive, EV, AI, and autonomous vehicle companies. Mobility roles at Sony Honda Mobility, Fisker, and a dozen other companies at various stages of build and launch. The volume taught me something that most resume advice gets wrong: the bottleneck is almost never what you wrote. It is how it was structured before a human ever read it.

In 2026, most mobility companies run their applications through at least two filters before a recruiter opens the file. First an ATS parser extracts your data into structured fields. Then, increasingly, an AI screening layer scores or ranks you relative to the job description. Only after both of those does a human spend an average of six to eight seconds deciding whether to keep reading.

Here is what each layer is actually evaluating, and the structure that survives all three.

What ATS Parsers Actually Do

Applicant tracking systems do not read your resume. They extract it. The parser is looking for discrete fields: name, contact info, employer names, job titles, dates, education, and skills. It is trying to fill a structured database record from an unstructured document.

This means certain formats break parsers entirely. Multi-column layouts confuse left-to-right text extraction, so a two-column resume may have your dates read as your job title. Tables with nested cells get garbled. Headers and footers in Word documents sometimes disappear from the parsed record. Text inside graphics or icons is usually invisible.

The practical fix is simpler than most people expect. A single-column layout, a standard font, no text boxes, and dates in a consistent format (Month YYYY, not just 2022 or Q3 2021) will parse cleanly in virtually every system I have worked with across Greenhouse, Workday, Lever, and Ashby. The goal is not to make your resume beautiful. It is to make it legible to a machine that is reading it in the dark.

What AI Screening Actually Scores

The AI layer that has grown significantly since 2023 is not reading your resume for talent. It is scoring your resume against the job description for keyword and context alignment. This is a matching problem, and it is more semantic than literal.

Modern screening models do not just look for exact phrase matches. They understand that "ADAS" and "advanced driver assistance systems" are the same thing. They recognize that "battery management" and "BMS development" are related. What they penalize is vagueness. A bullet that says "contributed to the development of hardware systems" scores lower than "led validation testing for a 400V battery management system across seven vehicle programs."

The implication: for every role you apply to, your resume needs to reflect the language that role uses. Not fabricated qualifications. The same real experience, described in the words the company chose when they wrote the posting. If they use "NPI" and your resume says "new product introduction," rewrite it. The AI scores the match, not the meaning.

I have seen candidates with strong backgrounds score poorly on AI screeners because they used the language of their previous company instead of the language of the role they wanted. The ATS accepted the document. The screener ranked it fifteenth. Nobody read it.

What a Human Reads in Six Seconds

When a recruiter opens a resume after the machine filters, the first six to eight seconds land on three things: your most recent title and company, the date range you were there, and whether any of your bullet points start with something concrete. After that, they decide whether to keep going.

Most resumes I reviewed during the AFEELA launch were lost in the opening bullet. "Responsible for leading cross-functional teams to deliver engineering milestones" tells me nothing about scope, outcome, or capability. It does not answer the first question a recruiter is actually asking: what did you build, and at what scale?

The reframe that works: treat each bullet as a project debrief, not a job description. Start with the verb and the object. Add the constraint or context. End with the result or scale. "Redesigned supplier qualification process for a new powertrain program, cutting approval lead time from 14 weeks to 6" is something I can evaluate. It tells me what you touched, why it mattered, and what you delivered.

The Structure That Survives All Three

Given what each layer is doing, here is the structure that clears ATS parsing, scores well on AI screening, and holds a human's attention past the six-second mark.

Header: clean and parseable

Name, city and state (not full address), phone, email, LinkedIn URL, and optionally a portfolio or GitHub. Single column. No icons or graphics. Text only. Place this at the top of the document, not in a header field. Most parsers do not extract from document headers.

Summary: optional but high-value if brief

Two to three sentences maximum. Lead with your function and sector. Name one or two specific capabilities that are searchable and relevant to the roles you are pursuing. Do not open with "results-driven professional." Open with what you build, where you have built it, and at what stage of company. "Mechanical engineer with eight years in EV powertrain development, currently focused on cell-to-pack integration at the prototype stage" is far more useful than a generic value statement.

Experience: reverse chronological, project-debrief bullets

Each role needs: company, title, location (or Remote), and date range in Month YYYY format. Three to six bullets per role, each starting with a specific action verb and including scope or scale. The most recent two or three roles should be the most detailed. Earlier roles can have two to three bullets.

For engineers, include: what system or component, what phase (design, validation, production, launch), what constraints (weight, cost, timeline, regulatory), and what the outcome was. For program and project roles, include: team size or function scope, budget if material, and delivery outcome. For supply chain and manufacturing, include: commodity or process, volume, and the shift you drove.

Skills: keyword-dense, category-organized

List skills explicitly, even when they appear in your bullets. The AI screener often pulls from a dedicated skills section as well as bullets, and having both increases your match score on exact-phrase queries. Group by category: Hardware, Software, Methods, Tools, Certifications. Keep it scannable. Avoid long prose in this section.

Education: below experience, brief

Degree, institution, graduation year. If you have a relevant certification (SHRM-CP, PMP, FCA Level 3, Six Sigma Black Belt), list it here or in a separate certifications row. Do not put education at the top unless you graduated within the last two years and have limited work history.

One More Thing About Mobility Specifically

The companies on the Mobility Jobs board span at least five distinct stages: pre-seed prototype, Series B scaling, production launch, post-launch optimization, and public company steady-state. Each stage reads resumes differently.

A startup at the prototype stage is looking for someone who has done the thing before, in ambiguous conditions, with few resources. They are not reading for polish. They are reading for proof that you shipped something hard. A production-stage company needs process discipline and cross-functional coordination. An OEM doing an EV transition program is reading for specific system experience and the ability to work inside large organizations.

The single most common mistake I see is a candidate using the same resume across all of these. The structure should stay consistent. The emphasis, the vocabulary, and which bullets you lead with should shift to match the stage of the company you are targeting. Three to four targeted versions, not one generic document.

Your experience does not change. The frame you put around it does.