Junk characters and decorative bullets, and why they glue to your words
The short answer
Unreadable characters is an 8 point finding in our published rule set. Decorative bullet glyphs is 3. They are separate checks with different weights, and they have one cause between them: a symbol in your file that the software has no letter for.
When that happens the character does not politely disappear. It comes through as a replacement mark, and because bullets sit at the start of a line, the junk tends to fuse to the first word of the sentence it was introducing. Your best achievement arrives with a piece of debris welded to the front of it.
What a replacement character actually means
When extracted text comes back with replacement marks in it, the file's font encoding does not map back to real letters.
The plain version: a document does not store letters, it stores numbers, plus a table that says which number means which letter. When that table is missing, wrong, or built for a font the reader does not have, the numbers arrive with nothing to turn them into. The reader substitutes a placeholder and moves on.
It usually happens for one of a small number of reasons. A font was embedded in a way that lost its mapping. A file was converted between formats by a tool that did not carry the encoding across. Text was pasted in from a web page or a PDF and brought foreign characters along with it. A design tool exported with settings that outlined some glyphs and not others.
The important part is that none of these are visible to you. Your machine has the font, or a close enough substitute, so it renders the letters correctly. The failure only appears on the other side, in the extracted text.
Why bullets are the usual suspect
Decorative bullet glyphs get their own check because of where they sit and what people choose.
Stars, arrows, check marks, diamonds, small squares, hollow circles. They are chosen precisely because they are distinctive, which is another way of saying they are less common, which is another way of saying they are less likely to be in a given parser's character map.
When one fails, the damage lands in the worst possible position. A bullet is the first thing on the line. The junk character attaches itself to the leading word of your achievement, so a bullet that should read as a clear statement of what you did arrives with a broken glyph stuck to the front. A reader skims past it. A parser may treat the fused result as a single unknown word.
There is a second, quieter cost. A parser often uses bullet characters as a signal that it is looking at a list. When the bullet is a symbol it does not recognise, it may not register the list as a list at all, and your carefully separated achievements collapse into one undifferentiated block of text.
The test that finds both
Select all the text in your exported resume, copy it, and paste it into a blank plain text document.
That paste is close to what a parser receives, and both of these problems are visible in it immediately. Replacement characters show up as obvious placeholder marks. Failed bullets show up as junk at the start of lines, or as lines that have lost their separation entirely.
Read the whole thing rather than skimming the top. Encoding problems are often local to a section, usually the one that was pasted in from somewhere else, so the first half of a document can be perfect while the second half is full of debris.
The fix
For bullets, use the standard round bullet. It is in every character map that has ever existed and it costs you nothing that matters. If you want visual differentiation between levels of a list, indentation does that job without introducing a character anyone has to recognise.
If a distinctive bullet is genuinely part of a design you care about, keep it for the version you hand to humans and use the plain one for the version that goes through application forms. That is the same split that makes sense for heavily designed layouts generally, and we make the argument for it in two column resumes.
For replacement characters, find the section they cluster in and retype it rather than trying to repair it. Pasted content brings its encoding with it, and clearing the formatting on a paste often leaves the underlying characters untouched. Retyping a paragraph is faster than diagnosing a font mapping, and it is certain.
Then re-export and run the copy and paste test again on the new file. Do not assume the fix worked. The reason this check exists is that the problem is invisible from the inside, and that stays true after you have edited it.
The third small check in this group
There is one more parse check in the same size range, and it belongs here because it comes from the same place: type below 9 points, worth 4 points.
Sub nine point type is rarely a parsing failure by itself. It is a signal. It usually means two pages of content have been squeezed into one page of space, and it reads poorly for the human who sees the resume after the software does.
The reason it belongs alongside the other two is that all three tend to arrive together. A page that has been compressed to fit is a page where the type has been shrunk, the margins have been pushed out, the bullets have been swapped for something tighter, and content has been pushed into boxes and tables to save room. Those are separate checks worth 8 points each, and we cover them in tables and text boxes in a resume. The compression is the common cause underneath several findings at once.
If you are fighting for space, the answer is almost always to cut a role or shorten some bullets rather than to shrink the type. A shorter resume that parses is worth more than a dense one that does not.
Keeping this in proportion
Three points and eight points are small numbers. They are worth fixing because they are cheap to fix, not because they are the main event.
The main event is further up: parse is worth 35 of the 100 points, and the two checks that can take all 35 on their own are a locked file and a file with no text layer, both covered in the two resume files that score zero. After that, evidence is worth 25, and a single check inside it, numbers in your bullets, is worth 15 by itself.
Swapping a star for a round bullet takes thirty seconds and you should do it. Just do it after you have confirmed the file can be read at all, and after you have put numbers in the bullets those symbols are introducing. The full ranked order is in what actually breaks a resume in an ATS, and if you want the specific findings for your own file, the free ATS checker runs every published check in your browser without uploading anything. Whether you keep one resume or several versions of it is a related question, answered in one resume or several versions.
Small findings are the easiest to miss by eye and the easiest for software to spot, which is what the free ATS checker from iapplyai.app is for. It shows each one against the exact line that triggered it, so a stray glyph is a location rather than a hunt.
