You spent a whole evening on that resume. The clean icons next to your contact info, the elegant skill bars showing your Excel mastery, the little chart of your sales growth. It looks like the resume of someone who has their act together. And then the applications go out, and nothing comes back, and you cannot figure out why.
Here is the hard truth, delivered gently: the software that reads your resume before any human does cannot see any of it. Not the icons, not the bars, not the chart. This article walks through exactly what happens to each visual element at parse time, and then shows you how to translate every single one into text without losing what it was trying to say.
What parsing does to a designed resume
An applicant tracking system does not look at your resume. It extracts from it. Parsing software opens your file, pulls out the text characters, and assembles them into a structured candidate record: contact fields, work history, skills, education. That record, not your PDF, is what gets searched, filtered, and ranked. The full pipeline is laid out in what an ATS actually reads from your resume.
Text extraction has a defining limitation: it can only extract text. A vector icon is drawing instructions. A skill bar is a colored rectangle. A chart is shapes and maybe a few stray axis labels. None of these contain sentences. So the parser does not misread your graphics. It does something quieter and worse: it skips them entirely, and the information they carried simply does not exist in your record.
No error message. No warning. Your beautiful resume produces a sparse, gap-riddled record, and you never find out.
Element by element: what survives and what vanishes
Let's walk through the common visuals one at a time.
Icons (phone, email, location, LinkedIn)
The icon itself vanishes. If the text next to it is real text, that survives. The risk is the layouts where icons and contact details sit inside a designed header, sidebar, or text box; depending on the template, the parser may read them out of order or not at all. A phone number the parser cannot attach to a phone field can mean a recruiter who cannot call you. Sidebars and text boxes have their own failure modes, covered in do resume tables and columns break ATS.
Skill bars and star ratings
The bar is a rectangle; the parser gets nothing from it. If the skill name is real text, the name survives with no level attached, which is the good case. In some templates the whole skills block is one exported graphic, and then even the names are gone: your entire skills section, the part keyword matching depends on most, contributes zero matches.
And here is the twist: even for human readers, skill bars underperform. A recruiter sees "Python: 4 out of 5 stars" and thinks, according to whom? Self-graded ratings carry no evidence. "Python: 4 years, built the team's reporting pipeline" beats four stars with both audiences, and it is checkable, which is the whole point of an honest resume.
Charts and infographics
A chart of your sales growth extracts as, at best, a few floating numbers with no context. The actual story, the thing you made the chart to say, is gone. Meanwhile one plain sentence carries all of it: "Grew territory sales from $400K to $1.1M in two years." Every chart on a resume is a sentence wearing a costume.
Logos, headshots, and decorative elements
Company logos parse as nothing, so if a logo replaces the company name in text, that employer may drop out of your work history. Decorative lines and shapes are harmless when they are purely decorative, and harmful when the template uses them to structure content. Photos are their own topic with bias implications on top of the parsing problem.
Text saved as an image
The catastrophic case. Some design tools and export paths flatten text into pixels. The resume looks identical on screen, and the parser receives a blank page. If your file came out of Photoshop, Canva's image export, or a scanner, check it immediately. There is a dedicated guide to whether ATS can read a scanned or image resume, and the short answer is: assume no.
One reassurance while we are here: plain colored text is not a graphic. Color survives extraction fine, though it has its own considerations for readability, covered in does color on a resume hurt with ATS.
The translation guide: every visual, said in words
Here is the part that should actually make you feel better. Everything a graphic communicates can be said in text, and the text version is usually stronger because it can carry evidence. Nothing true gets lost in this translation.
- Icon row of contact details becomes labeled text: "Email: you@example.com | Phone: 555-0100 | City, State | linkedin.com/in/you". Boring, bulletproof, and it parses into exactly the right fields.
- Skill bar becomes an honest phrase: "Excel (advanced: pivot tables, Power Query, 6 years)". You just replaced a self-graded rectangle with verifiable specifics. Note the honesty upgrade: a bar let you feel like 5/5 without committing to anything. Words make you say what you actually know, which is exactly what you want, because you will be asked about it in the interview.
- Star-rated language skills become standard levels: "Spanish (professional working proficiency), French (conversational)".
- Chart becomes a bullet with the real numbers: "Increased qualified leads 60% year over year (from 250 to 400 per quarter)".
- Logo wall of employers or tools becomes a plain list of names, which is what the keyword matcher was looking for all along.
Run this translation and you lose exactly one thing: decoration. You keep every fact, you gain keyword matches, and you gain defensibility.
Do you need to abandon design entirely?
No. You need to put it in the right lane.
- For any application that goes through an online portal: use the plain-text-structured version. Single column, standard headings, real text everywhere. This is the version the bots build your record from.
- For humans you reach directly: a tasteful designed version is fine when you email a hiring manager, hand a resume over at a career fair, or bring copies to an interview. Humans can see your skill bars, even if they should still be sentences.
Two rules keep this honest and safe. First, the two versions must contain identical facts; only the presentation differs. Second, when in doubt about which version a destination needs, send the plain one. A plain resume in front of a human loses you nothing. A designed resume inside a parser can lose you everything.
See exactly which parts of your resume went blank
You do not have to guess which of your visuals survived. The free Bounce scan at careerbounce.io shows you your resume the way the bots receive it: the raw extracted text, with every icon, bar, and chart reduced to whatever it actually left behind, which is often nothing. It is free, it runs entirely on your device, and your resume never leaves your machine.
Upload the designed version and look at the gaps. Then translate each gap into honest text using the guide above. You will end up with a resume that is less decorated and more convincing, one where every claim is written in words you can stand behind when a human finally asks about it. That trade wins every time.