ATS guides
Tables, icons, text boxes and rating bars
These are the features that make a CV look considered and quietly remove information from it. None of them are visible as problems when you open your own file, which is exactly why they survive to the parser. Appliora scores them under Format traps (6 points), and several of them do their real damage to other checks entirely.
One principle covers all of them
If the meaning is carried by a shape rather than by a character, it does not arrive.
A parser recovers characters. Everything else on the page - the rules, the boxes, the circles, the bars, the colour of a heading - is drawing, and drawing is discarded. That single sentence predicts every item below, and it is worth holding onto because it also predicts the ones this page does not list.
The corollary is the useful half: anything you can say in words survives perfectly. A skill level written as "advanced" arrives; the same level shown as four filled dots does not. A section boundary marked by the word Experience arrives; the same boundary marked only by a horizontal rule and a change of colour does not.
The catalogue
Ordered by how often they turn up on an otherwise good CV.
Rating bars, dots and stars
The worst offender, because it fails silently and it occupies the part of the CV a recruiter searches. Five circles next to "Excel" communicate a level to a human and leave the extracted text as the single word Excel, or as a run of bullet characters that fuse onto the skill name. Appliora's Concrete skills (8 points) check treats an entry of 4 characters or fewer as an orphaned fragment unless it is on a curated list of real abbreviations - SQL, AWS, C++, UX and so on - which is how the surviving half of a clipped skill gets caught.
Tables holding the body of the CV
Cell order is reconstructed, not stored. A two-column table of dates against roles usually survives. A table that holds your entire experience section, with a date column, a title column and a description column, is the two-column problem wearing different clothes, and it scrambles in the same way and for the same reason.
Text boxes and floating frames
A text box is positioned independently of the document flow. In a PDF that means a block of glyphs with coordinates that have no relationship to the surrounding text, so where it lands in the extraction is a guess. A profile summary in a floating box can arrive after your education section, or between two bullets of a job.
Icons in place of labels
An icon drawn as a vector extracts as nothing, which is harmless when there is a readable label next to it and destructive when the icon is the label. An icon taken from a symbol font extracts as whatever character the font maps it to, and that character then fuses onto the start of the value beside it - a private-use glyph attached to the front of your email address.
Non-embedded fonts
The file renders correctly on the machine that has the font and is re-rendered with a substitute everywhere else, which changes spacing and can break characters. Appliora reports fonts that the PDF relies on the reader having installed, and the fix is a single export setting.
Ligatures
A typographic nicety that replaces two letters with one character. It is invisible, it is correct typography, and it can make a word unmatchable by a search for that word. Worth knowing about mainly so you recognise it if a checker reports it.
Emoji and decorative symbols
They arrive intact, which is the problem. An emoji in a section heading stops that heading matching the heading dictionary, so a section you obviously have reads as missing, and Recognised sections (10 points) is the check that pays for it.
Where the damage actually lands
Format traps carries 6 points. Most of these cost more than that, somewhere else in the report.
| Trap | Charged to | Also damages |
|---|---|---|
| Rating bars and dots | Format traps | Concrete skills (8 points) - levels vanish and fragments appear |
| A table holding the experience section | Format traps | Reading order (14 points) and the 69-point cap |
| Contact details in a header or footer | Contact details | Nothing else - but the finding is critical, and it is half the check's 8 points |
| Icons replacing labels | Format traps | Contact details when the icon fuses onto the value |
| Emoji or symbols in headings | Format traps | Recognised sections (10 points) - the heading stops matching |
| Non-embedded fonts | Format traps | Accented characters (5 points) when substitution breaks accented characters |
Weights are the published ones for model version 4.5.
This is why a report can show a modest deduction on format traps and a large one on skills or sections: the trap is the cause, and the cost is charged where the information was lost. Reading the findings rather than the bars is how you see which is which.
What to write instead
Each of these keeps the look and returns the information to the text layer.
- Weak
- Excel ●●●●○
- Stronger
- Excel - advanced (pivot tables, Power Query, model maintenance)
- Why
- The level arrives as a word, and the parenthetical gives a recruiter something to search on. The dots communicated nothing to the machine and a vague claim to the human.
- Weak
- A three-column table holding dates, titles and descriptions
- Stronger
- A full-width entry: title and employer on one line, the date range on the same line, bullets underneath
- Why
- Same visual rhythm, one reading order, and the date stays attached to the role it belongs to - which is what Consistent dates is checking.
- Weak
- ✉ anna@example.com ☎ +48 600 000 000, in the page footer
- Stronger
- anna@example.com | +48 600 000 000, in the body of page one
- Why
- Several systems strip the header and footer band before parsing. Reachability is what Contact details scores, not presence.
- Weak
- A summary in a floating box beside the header
- Stronger
- A summary paragraph under a heading that says Summary
- Why
- The box lands wherever the extractor puts it. The heading is a structure marker the parser is actively looking for.
- Weak
- 🚀 My Journey
- Stronger
- Experience
- Why
- The emoji breaks the dictionary match and the phrase was never in the dictionary. Two separate reasons the section disappears.
Photos, and the check that does not charge for one
Personal details is worth 3 points, and a photo costs zero of them.
Appliora detects a portrait conservatively: a sufficiently large image, high on page one, in a document that has real text around it. The detection is deliberately cautious, because a false "you have a photo" on a CV that has none is worse than missing one. When it finds a photo it reports it as context rather than as a fault - the finding's own copy says a photo is normal in Poland and Czechia, and a check that says that cannot also take points.
What a photo does do is take space and push the layout. A portrait in a narrow left column is frequently the reason the contact details ended up in that column too, which is where the interleaving starts. Whether to include a photo at all is a market question rather than a parsing one.
Questions people ask
Are tables always a bad idea on a CV?
No. A simple two-column table used for dates against roles usually extracts in the order you meant. The risk grows with the number of columns and with how much of the CV lives inside the table, because cell order is a reconstruction rather than a fact, and a wide table is the same reading-order problem as a two-column page.
What happens to a skills rating bar with five dots?
Nothing arrives. Filled and empty circles are drawn shapes or decorative glyphs, so the extracted text next to a skill is either empty or a run of symbols. You have used a line of the page to communicate a level and communicated no level at all. Write the level as a word instead.
Do icons next to my email and phone cause problems?
Sometimes. An icon drawn as a vector shape extracts as nothing, which is harmless. An icon taken from a symbol font extracts as whatever character that font maps it to, which can fuse onto the start of your email address. The safe version is an icon that sits beside readable text rather than replacing a label.
Why would a ligature matter?
Because a ligature replaces two characters with one. The letters f and i in a word printed with a ligature can be stored as a single character that a keyword search for that word will not match. It is invisible on screen and it is the kind of thing only an extraction test reveals.
Is a photo a format trap?
Not in Appliora's model. A photo is detected and reported so you know it was found, and it costs zero points, because a photo is entirely conventional in Poland and Czechia and a check cannot say that and charge for it in the same breath. What a photo can do is push your contact details into a narrow column, which is a different check.
See what a parser reads from your CV
Upload your current file and get the extracted text plus a 0-100 score across 14 checks. No signup.