What We Learned Testing Resumes Across 77 ATS Platforms

We tested over 100 real job postings

When we built Career Quiver's resume generation system, we didn't just read documentation and call it done. We tested against real job postings — over 100 of them — across dozens of different ATS platforms. We submitted resumes, analyzed how they were parsed, and tracked what broke.

The results taught us things that no amount of reading "ATS tips" articles would have revealed. Here's what stood out.

The platforms that surprised us

Taleo and Jobvite are JavaScript black holes. These platforms render almost everything client-side. When you try to read the job posting programmatically — which is how we fetch the job description to analyze it — you get back a near-empty HTML shell. The actual content loads later through JavaScript.

This meant our standard approach of fetching the page and reading it didn't work for these platforms. We had to build a separate fallback path that uses AI to interpret the rendered content. It was a significant engineering effort for just two platforms, but they're popular enough that we couldn't skip them.

Some platforms are better than their reputation. Workday gets a lot of hate from job seekers — the application process is long and frustrating. But its resume parsing is actually decent. It handles standard formats well and its search functionality is reasonably sophisticated. The user experience is the problem, not the parsing.

Smaller platforms often parse better than enterprise ones. Platforms like Ashby, Lever, and Greenhouse — typically used by startups and mid-size companies — tend to have more modern parsers that handle varied resume formats gracefully. The legacy enterprise systems are where you see the most parsing failures.

What actually breaks resume parsing

After all our testing, the causes of parsing failures were surprisingly consistent across platforms:

• Tables and columns: Multi-column layouts using HTML tables or complex positioning break parsing on the majority of platforms. The parser reads across rows instead of down columns, turning your neatly organized resume into gibberish.
• Headers and footers: Content in document headers or footers is frequently ignored entirely. If your name and contact information only appear in the header, some platforms won't capture them.
• Graphics and icons: Star ratings, skill bars, progress circles, and decorative icons are invisible to parsers. Worse, they sometimes cause the parser to skip surrounding text.
• Non-standard section headings: Creative labels like "My Journey" instead of "Work Experience" or "Toolkit" instead of "Skills" confuse many parsers. They rely on recognizing standard section names to categorize your information.

We've covered the formatting specifics in our complete ATS formatting guide, but the testing is what convinced us these weren't edge cases — they affect the majority of decorated resumes.

The 72-90% accuracy range

Here's something we discovered that might surprise you: even for well-formatted resumes, individual job posting URLs were parsed with 72-90% accuracy across platforms. Not 100%. Not even close.

What does that mean? It means that even when your resume is formatted perfectly, some information may not be captured exactly right. A date might be misread. A job title might merge with a company name. A skill might be categorized under the wrong section.

This is why the content of your resume matters at least as much as the format. When a recruiter searches the ATS for candidates, they're searching by keywords. If your resume content matches the job description's language — even if the parser slightly mangled the formatting — you'll still surface in searches.

Career pages vs. individual job postings

One unexpected finding: the quality of our job analysis depended heavily on whether the URL pointed to an individual job posting or a company's general careers page.

Individual job posting URLs — the page for a specific open role — produced consistently high-quality results. The job title, requirements, responsibilities, and qualifications were cleanly structured and easy to parse.

Career index pages — the "browse all jobs" landing pages — were a different story. These pages list many jobs but contain little detail about any specific one. When we tried to analyze these, the results were poor because there wasn't enough job-specific information to work with.

This seems obvious in retrospect, but it shaped how we designed the product. We now validate that the URL points to a specific posting before running the analysis pipeline, and we provide clear guidance when a user submits a careers landing page instead.

Platform-specific quirks we documented

Every platform has its own personality. Here are a few of the specific behaviors we cataloged:

• USAJOBS uses a completely different resume format than private-sector ATS platforms. We wrote a separate guide for federal resumes because the rules are that different.
• iCIMS handles PDF uploads well but sometimes struggles with DOCX files that use advanced formatting features.
• Greenhouse is one of the more forgiving platforms — it handles varied formats without major issues, which makes sense given its focus on candidate experience.
• SuccessFactors has some of the most rigid parsing behavior we encountered. Standard formatting is essential.

What this means for how we build resumes

All of this testing feeds directly into how Career Quiver generates resumes. When we detect that you're applying through a specific platform, we use what we've learned about that platform's parsing behavior to make formatting decisions.

For platforms that struggle with columns, we use single-column layouts. For platforms with good parsing, we can use slightly more complex formatting that looks better to the human reader. For platforms with known quirks around section headings, we use the exact labels they expect.

This is the kind of specificity you can't get from a generic "ATS-friendly template." It requires knowing how each platform actually behaves — and the only way to know that is to test.

Testing never stops

This isn't a project we finished and moved on from. ATS platforms update their software regularly. New platforms enter the market. Existing platforms get acquired and merged — like when Paycor absorbed Newton Software.

We continue testing against live job postings to make sure our detection and formatting recommendations stay accurate. When something breaks, we fix it. When a new platform gains enough market share to matter, we add it.

The job search is already hard enough without having to guess whether your resume format works with a company's hiring software. That's the problem we're trying to eliminate — and testing is how we make sure we're actually doing it.