How to pass ATS filters in 2026

Where a resume can actually be stopped

An application passes through five stages before anyone offers you an interview, and only two of them produce a no. The software parses your file, applies whatever knockout rules the recruiter configured, grades you against the job criteria, waits in a database for recruiter searches, then gets skimmed by a person.

Knockout questions are the automatic no. They are the checkboxes you answer in the form: are you legally authorised to work here, will you need sponsorship, do you hold the licence, are you within commuting distance. Greenhouse lets recruiters attach an auto-reject rule to any yes/no, single-select or multi-select question, and that rule fires within seconds of submission. That is the source of the rejection email that lands before lunch.

The human skim is the other no, and it is where most rejections happen. Everything in between reorders a list. If you have applied to a hundred roles and heard nothing, the diagnostic in why you are not getting interviews separates a parsing problem from a targeting problem, which need different fixes.

Which ATS you are probably facing

Jobscan publishes the only regularly updated census of this market. In its sixth edition, dated 20 August 2026, it found a detectable ATS at 487 of the Fortune 500, or 97.4%, from a manual review of career pages plus more than a million resume scans across 13,130 companies. Workday runs over 40% of the Fortune 500 and SAP SuccessFactors around 12%.

Across the wider sample of 13,130 companies the shape changes completely: Greenhouse 19.0%, Workday 17.5%, Lever 15.6% and iCIMS 14.4%. Jobscan sells resume optimisation, so it has a commercial interest in a high ATS count, and its own older page gives different figures. Use it as an order of magnitude. The practical read is that the six systems in the table above cover most of what you will meet, and they behave similarly enough that one clean file works everywhere.

ATS friendly resume format in 2026: what breaks and what does not

Greenhouse accepts .doc, .docx, .pdf, .rtf and .txt files up to 100 MB and states no penalty for PDF (support page updated 19 August 2026). Workday portals accept PDF as well. The file extension is close to irrelevant. What matters is whether there is text inside.

• Image-only PDFs fail completely. A scan, a screenshot, a design export that flattens type, or a PDF printed from a LinkedIn profile gives the parser nothing. Open the file and drag your cursor across a line: if the text does not highlight, the parser reads an empty page.
• Columns and tables can still break some parsers. Textkernel, whose parsing engine it says is over 95% accurate on critical fields, has said two-column layouts can choke some parsers and that tables can come out as a column of dates, a column of titles and a column of companies with nothing linking them. Those statements date from 2021 and its newer LLM parser handles columns, so this is a risk with no upside rather than a certainty.
• Headers and footers are the classic silent loss. Contact details placed in a Word header survive visually and disappear from plenty of extractions.
• Skill bars and star ratings carry no text. Write the level out: Spanish, professional working proficiency.
• Standard section headings win. Work Experience, Education, Skills. A heading reading My journey maps to no field.

Textkernel also normalises skills and job titles to public taxonomies such as O*NET and ISCO, which is the technical reason a standard job title beats an internal one. Growth Ninja is a title no taxonomy contains. If you want to see the extracted text rather than guess, the ATS extraction preview shows what a parser pulls out of your file, and single-column resume examples show the structure that survives it.

How recruiters actually search the database

Once your resume is parsed and graded, it sits in a database that a recruiter queries. LinkedIn Recruiter supports AND, OR and NOT in capitals, parentheses and quoted phrases across the title, company, keyword, skills, location, school, industry and language fields, with plus and minus signs unsupported and stop words ignored in the keyword field. Greenhouse runs full-text search across resumes and returns snippets. A recruiter typing "product manager" AND (SQL OR Looker) is reading whatever text your file produced, nothing else.

Jobscan's analysis of 2.5 million applications found recruiters filtering on skills in 76.4% of searches, education in 59.7% and job title in 55.3%, and that a resume title matching the job ad went with a 10.6 times higher interview rate. Correlation, from a vendor, on its own users. It still matches what the taxonomies do, and the cheapest change on your resume is putting the posting's standard job title at the top when it describes what you did.

Matching is drifting away from exact strings. Greenhouse maps related terms to one skill, and LinkedIn's Hiring Assistant, an agent recruiters brief in natural language, went globally available in English in September 2025. Writing for meaning and writing for keywords are converging, which is the argument made in our roundup of AI adoption on both sides of the hiring table.

A checklist to run before you send

1. Select a sentence in your PDF with the cursor. If it highlights, the text is real.
2. Run the file through a parser and read the output rather than the layout.
3. Check the current job title, dates and employer names all came out attached to each other.
4. Put the posting's standard job title at the top of your most recent role when it is accurate.
5. Answer every knockout question deliberately, since that is the one automatic rejection.
6. Move anything from a header, footer or text box into the body.
7. Replace skill bars and star ratings with words.
8. Use Work Experience, Education and Skills as headings.
9. Spell out each acronym once, for example CRM (customer relationship management).
10. Delete any hidden text, white font or invisible instruction.

Two jobs sit next to this one and are covered properly elsewhere. Pulling the right terms out of a posting is a separate skill, set out in finding and placing resume keywords. Rewriting the document for a specific posting without inventing experience is a workflow with before and after examples, in tailoring your resume to a job description. This article stops at how the software reads and ranks what you send.

Frequently asked questions

Where Lumyhired fits

Everything above comes down to one question you cannot answer by looking at your resume: what text does a parser actually get out of it. The ATS preview shows that extracted text, the ATS resume checker flags the fields that came out empty or misfiled, and the tailoring step then rewrites the wording against a specific posting. Weekly plans start at €5 on the pricing page, cancel anytime, with a 3-day money-back window. If you are comparing scanners, we put ourselves against the best-known one in Lumyhired vs Jobscan, and the wider category is reviewed in the best ATS resume tools for 2026. For how this fits a whole search, see our guide to getting a job in 2026.