The expression “resume for ATS” usually suggests that there is a single robot, with a secret formula, evaluating all candidates in the same way. In 2026, this idea is a poor starting point. Different platforms read different information, allow different filters and can use matching criteria configured by the company itself. What works consistently is not trying to fool an algorithm, but rather making your history easier to read and presenting evidence relevant to the position.
The comparison between Gupy, Greenhouse and Workday helps to separate four layers that are often mixed: file parsing, filters or eliminatory requirements, matching or ordering and human evaluation. Understanding these layers changes the way you prepare your resume because it shows why a technically perfect file still may not advance — and why an “ATS score” obtained outside the company is not a reliable prediction of your actual position.
ATS is not a single step
A recruitment system can receive the resume, extract data from the document, store form responses, apply eliminatory questions, allow word searches, compare qualifications with requirements and organize the queue for review. Not all of these functions exist on all platforms, nor are they configured in the same way by all companies.
At Gupy, documentation published in 2026 states that the AI agents used in sorting analyze fields of skills and experience compared to requirements defined by the company. Gupy itself also separates this ordering from eliminatory criteria configured by the recruiter. This means that, in an application of this type, filling out the structured fields well may be more important than trying to “optimize” a PDF in isolation.
On Greenhouse Recruiting, the resume can be analyzed to fill in profile fields, and the platform also offers search, filters and a Talent Matching feature. In this feature, the
hiring defines and considers criteria for the vacancy; the system compares the resume and application responses against these criteria and presents compatibility categories. The documentation makes it clear that the resource is assistive and that recruiters and managers remain responsible for the decision.
In the Workday ecosystem, HiredScore can transform resumes into analyzed profiles, compare qualifications to the job description, and assign scores for prioritization. In August 2026, the product documentation noted an opt-in update to Spotlight Fit & Gap that now evaluates qualifications individually with a language model-based approach. Customers who do not enable the change can continue with the previous logic. Even within the same product family, therefore, version and configuration matter.
1. Parsing: first the system needs to understand what you wrote
Parsing is the transformation of the document into data that the system can organize. The objective is not to “grade” the resume. It involves recognizing information such as name, contacts, positions, companies, dates, training and skills, depending on the platform.
This difference is important because a lot of ATS advice confuses technical readability with ranking. A resume can have great content and still be interpreted incompletely if the visual structure hides important information. It can also be perfectly read and still not meet the job requirements.
The Greenhouse documentation shows a concrete example. The parser may fail or extract only part of the data when the resume uses photos, graphics, word art, complex tables, headers, footers, text boxes, or multiple columns. Contact information placed in a header, footer or text box also appears among risk cases. The problem is not that every ATS on the market prohibits these elements, but rather that sophisticated structures increase the possibility of inconsistent interpretation.
Therefore, a safe recommendation is to keep the main resume simple: one column, recognizable sections, real text instead of images, clearly written job titles and companies, and easy-to-identify dates. Design doesn't need to disappear, but it shouldn't compete with readability.
PDF or DOCX: why there is no universal format
The phrase “ATS only accepts DOCX” is as problematic as “always send PDF”. Greenhouse accepts DOC, DOCX, PDF, RTF and TXT files for applications. MyGreenhouse recommends PDF for best quick profile results, but this recommendation pertains to that specific flow. Workday HiredScore also documents DOC, DOCX, PDF, RTF, and TXT support for attached resumes.
The practical decision is simple: follow the format requested by the vacancy. When there is no instruction, use a common and readable file, preferably with selectable text and without relying on complex graphical resources. If the application automatically extracts data, review what you filled out before submitting. Greenhouse itself warns that parsing can make errors and recommends checking the fields after importing.
2. Filters and eliminatory requirements are not the same thing as “resume score”
Another common confusion is to imagine that every initial failure means that the resume received a low grade from the AI. In many processes, objective requirements can be applied before or independently of any matching.
Gupy explains that the company can define mandatory criteria when registering. If the person does not meet a requirement marked as eliminatory, failure may occur automatically. The documentation also differentiates this mechanism from the actions of AI agents: the requirements are defined by the recruiter and the automation verifies the answers provided.
In practice, this changes the focus of preparation. If the vacancy requires a license, availability, location, language, training or other objective condition, no repetition of keywords compensates for an incompatible answer. Likewise, it makes no sense to attribute to a supposed “resume-reading robot” an elimination that came from a specific question on the form.
Therefore, before adjusting CV terms, read the mandatory requirements and respond to the form accurately. The form is part of the application. On some platforms, it may carry information that the attached file does not replace.
3. Matching and ordering: words help, but not as a trick
Once the data is available and the basic requirements have been considered, some platforms can help the recruiter prioritize applications. This is where the obsession with keywords arises, but the current documentation shows a more sophisticated scenario than the idea of “copying the job to climb the rankings”.
In Greenhouse Talent Matching, the team sets up a calibration with criteria and weights. The platform uses these definitions to calculate the strength of the match and shows the recruiter which experiences, sectors and skills contributed to the result. The documentation also states that the resume may highlight exact matches and similar terms.
In Workday, matching features can consider exact and similar skills. On HiredScore, scores depend on how much your resume and qualifications match what was specified for the request. In the August 2026 Fit & Gap update, each applicable qualification can be evaluated individually, with explanations of what was met and what was not.
At Gupy, agents analyze skills and experiences in relation to the job requirements. The company itself recommends clear and authentic descriptions and rejects the idea of “cheating” the ordering with exaggerated keywords.
The practical conclusion is not to abandon keywords. It's using them as alignment language, not as filler. If the job asks for “Power BI” and you actually use Power BI, write that down. If you ask for “project management” and this skill is part of your experience, make it explicit in the context in which it was used. What doesn't help is repeating terms without evidence, copying blocks from the job description or inserting skills you don't have.
Myth: copying the exact words always increases the score
There is not always a numerical score, and not every platform relies solely on literal coincidence. Greenhouse documents exact and similar terms within Talent Matching. Workday also describes exact and similar skill matching. Gupy works with fields of skills and experience and with requirements defined by the company.
Using job vocabulary is still useful when it accurately describes what you did, because it reduces ambiguity for systems and people. But replacing all the synonyms on the resume with a literal copy of the advertisement does not guarantee better ordering. In systems that consider similarity, context or weighted criteria, the relationship between experience and requirement is more important than repeating a sequence of words.
4. Human assessment continues to be part of the process
The fear of an ATS often turns the candidacy into a contest against a machine. The analyzed platforms themselves describe another role for their AI capabilities: supporting prioritization and review.
Gupy states that its agents organize applications and that the final decision remains with the recruiting company. Greenhouse says Talent Matching does not automatically advance or reject candidates and that recruiters and managers stand by the decision. Workday also describes its capabilities as supporting the process and maintaining human oversight; In HiredScore documentation, candidates with lower scores may still be considered suitable after review.
This point has direct consequences for the curriculum. A text artificially inflated to maximize coincidences may appear optimized in an external tool, but it will still be subject to human reading, interviewing and coherence checking. A good CV needs to survive both readings: the structured one, made by the system, and the contextual one, made by those who hire.
What is a safe recommendation in practically any ATS
There is a core of good practices that remains useful even when the platform changes. The resume should use selectable text, predictable structure, and clear language. Name, contact, positions, companies, dates, training and relevant skills must appear in easy-to-identify places. Central information should not rely solely on images, icons or decorative elements.
It is also safe to adapt the content to the vacancy without distorting the trajectory. This means highlighting relevant experiences, using recognizable industry terminology, and making explicit skills that the ad actually asks for when you have them. Adaptation does not need to recreate the entire curriculum; It often consists of changing priority, order and level of detail.
Another robust practice is to review the profile within the platform itself. If the system fills in fields from the file, check whether the title, company, dates and training were interpreted correctly. If the platform requires structured fields, complete them even if the same information is already in the PDF.
Finally, answer eliminatory questions carefully. A well-written resume does not correct an objective response that is incompatible with the position, and trying to circumvent a mandatory requirement creates a bigger problem when the application reaches human review.
Three myths that deserve to be retired in 2026
“There is a file format that passes any ATS”
There is no universal rule demonstrated by the platforms themselves. Greenhouse and Workday HiredScore support several common formats. Some flows recommend PDF; others can handle
well with DOCX; and structured profile-based systems may give more weight to completed fields than to the attachment. Vacancy instruction must have priority.
“The more keywords I repeat, the higher I rank”
Current documentation does not support this rule as a universal strategy. Systems can consider similar terms, skills, experience, weighted requirements and structured fields. Repetition without context is no substitute for evidence of experience and can even make the resume worse for human review.
“My score on an ATS website predicts my position in the company”
An external evaluator can be helpful in reviewing clarity, competency coverage, and readability, but does not necessarily know the actual process setup. At Greenhouse, for example, the company defines and considers the calibration criteria. At Gupy, criteria and priorities are defined in the context of the vacancy. In Workday HiredScore, the logic depends on the request requirements and may vary depending on enabled features and versions.
Therefore, an external score should be treated as a diagnosis of a specific model, not as a prediction of the actual queue. If a tool says “82%”, that number only has meaning within that tool’s methodology. It is not an automatic translation of the grade, category, or position that another ATS will use.
How to adapt your CV without falling for hacks
Start with the vacancy, not with the imaginary algorithm. Identify which skills, experiences and requirements appear as central. Then, make sure your resume shows concrete evidence of these skills. When there is a real match, use terms close to those in the ad to reduce ambiguity.
Then, simplify the structure. If your resume was created as a visual piece, keep a more straightforward version for online applications. Avoid placing essential information in headers, footers, text boxes, infographics or complex columns. The goal is to make reading predictable without turning the document into a block of text without hierarchy.
Also review the quality of the descriptions. “Responsible for reporting” communicates less than “created and maintained sales reports in Power BI for weekly goal tracking”, as long as that is true. The second example offers activity, tool and context. This type of specificity helps both the matching and the person doing the screening.
When there are measurable results, include them with context. Numbers are not required in all experiments, but they can make the contribution more verifiable. Avoid making up percentages just to appear results-oriented.
Finally, treat the form as part of the resume. On platforms that use structured fields, an experience described only in the attachment may not have the same weight as information filled in the correct field. Read what the application asks for and review before completing.
What to do when you don't know which ATS the company uses
The best strategy is to optimize for interoperability. Use a simple version of your resume, with clear sections, selectable text, straightforward language, and essential information in the body.
document. Follow the upload instructions exactly, complete all required fields and avoid assuming that the company's process works like Gupy, Greenhouse, Workday or any other well-known platform.
If the vacancy allows you to attach your resume and also fill out a profile, do both consistently. If the system extracts data from the file, review it. If there are requirements questions, answer them accurately. If the job description uses specific nomenclature for a skill you actually possess, make this correspondence explicit.
This approach may seem less enticing than a list of “10 hacks to beat ATS,” but it is technically more defensible. It does not depend on discovering a secret algorithm and remains valid when the company changes weights, filters, version or tool.
The resume that works for ATS also needs to work for people
In 2026, preparing a resume for recruitment systems is not writing for a robot. It is to reduce information losses throughout a flow that may involve automatic reading, structured fields, mandatory requirements, matching and human review.
The most useful rule is to separate these layers. Parsing requires readability. Filters require attention to objective requirements. Matching requires true alignment between experience and vacancy. Human assessment calls for clarity, coherence and evidence.
When these four parts are well attended to, the curriculum no longer depends on tricks. It becomes easier to interpret by different systems and, at the same time, more convincing for those who make the decision to continue the conversation.
