How to write a resume that gets past AI screeners in 2026.

80% of resumes at mid-size and large companies are now read by an AI screener before they reach a human. Not just keyword-scanned — actually read, scored, and ranked. If you write your resume the way most career coaches still teach (long, narrative, dense paragraphs), you're optimizing for a reader that doesn't exist.

Here's the version that survives the screener and still reads well to the human who eventually opens it.

What the screener is doing

The AI is doing one thing: scoring how well your resume matches the job description, weighted toward signals it can extract reliably. Specific job titles, specific technologies, quantified accomplishments, dates, location. It's not doing literary criticism. It's not impressed by adjectives. It does not care about your "passion for driving impact".

The screener that scored highest in third-party evaluations in 2025 spends most of its compute on three sections: skills, recent experience, and the bullets under each role. Everything else gets one pass.

What to do, what not to do

Pre-AI resume habit

"Spearheaded cross-functional initiatives leveraging cutting-edge methodologies to drive transformational outcomes."

14 words. Zero specifics. The screener gives this almost no signal. The human reader gives it less.

AI-screener-friendly

"Led migration of 4 services from EC2 to ECS Fargate. Reduced infrastructure cost 38% ($14k/mo). Owned the runbook, on-call rotation, and rollback plan."

30 words. Three concrete numbers, two named technologies, the verb is what you did. The screener loves it; the human reads it twice.

The four-line rule

For every role on your resume, write four bullets. Not three, not seven. Four.

Each bullet must contain:

• A specific verb ("built", "shipped", "migrated", "owned")
• A specific noun (the system, the team, the deliverable)
• A specific number (latency, cost, time saved, % impact)
• A specific technology, where natural

The hardest of those is the number. Most candidates don't have great numbers. The fix is to ask better questions of your own work. "How much faster did the new flow run?" "How many customers used the feature in the first month?" "What was the bug-rate change?" Even small numbers help. A bullet with a number is read 3-4x more carefully than one without.

The skills section, in 2026

The skills section is no longer a place to list every technology you've ever touched. It's a place to mirror back the job description's must-haves so the screener confirms you have them.

Read the job description. Pull out the named technologies. Make sure each one you genuinely know is in your skills section, in roughly the same phrasing. Don't pad it with technologies you don't know — modern screeners cross-check skills against the experience section and penalize mismatches.

Pro move: copy the company's preferred phrasing exactly. If the JD says "Kubernetes (EKS)", write "Kubernetes (EKS)" — not just "Kubernetes". The screener handles this fine. Some recruiters' downstream tools don't.

The honest meta-move

If a job description has five things in the "must-have" list and your resume contains only two of them, you will not get past the screener. This is not a moral judgment about your fit. It's a math fact about how scoring works. Apply to roles where your resume already matches 80%, not roles where you'd be a stretch on paper.

Format that doesn't break

A few practical formatting rules screeners reward:

• One column, not two. Multi-column resumes break parsing in ways most candidates never notice.
• Standard section headers: "Experience", "Skills", "Education". The screeners look for these by name. Creative headers ("My Adventures") get skipped.
• Plain text date formats: "Jan 2023 - May 2025". Avoid "1/23 - 5/25" or anything stylized.
• PDF, not Word. Same file, but PDF preserves layout consistently across screeners.
• No graphics, no charts, no icons next to bullets. They look great to a human and confuse the parser.

The 30-minute audit

Take your current resume. Open the job description for a role you'd genuinely apply for. Count two numbers:

1. How many of the job's named technologies appear, in matching phrasing, on your resume?
2. How many of your bullets have a specific number?

If either number is under 5, that's your weekend project. The screener doesn't need you to lie. It needs you to be specific. Most resumes lose at the screener stage not because the candidate wasn't qualified, but because the candidate wasn't specific.

Fix the specificity. The interview rate roughly doubles in the next two weeks. Boringly, that's the whole game.