What is actually in the file
For years Telugu DTP was done in fonts such as Anu, Priyaanka, Kranthi and Brahma, which put Telugu shapes in the places an English font keeps its letters. The document stores those codes, and the font turns each code into a Telugu shape on the page. A PDF made from it carries the codes and the font together, so it prints and displays perfectly.
What it does not carry is which Telugu letter each code stands for. Copying hands over the codes, and the editor you paste into draws them in an ordinary font, where they are ü, ÷ and ≤. The PDF usually has a map that is meant to turn codes back into characters, called ToUnicode, and the Anu fonts in the application above have one. It maps every code to a Latin character, because that is what the font says it is. A check that only asks whether the map exists passes exactly the files that fail.
Why installing a Telugu font does not help
The damage is in the copied text, not in how it is displayed. Pasting into a document set in Gautami, Nirmala UI or Noto Sans Telugu changes nothing, because the pasted characters really are Latin letters and symbols, not Telugu letters shown in the wrong font. Changing the font of the pasted text does not bring the Telugu back either. The text has to be converted, code by code, into Unicode Telugu.
How to tell for sure
The analyser above lists every font in the PDF and reads the codes each one draws. For the application in the sample it reports “6 pages · Telugu · legacy font detected”, names Brahma, Kranthi, Priyaanka and PriyaankaBold as Anu Telugu, and says text extraction fails for 100% of the text. The English address on the same page is set in Arial and Times and copies out correctly, which is why only some lines break.
It identifies the family from the codes rather than trusting the font name, because names are unreliable. A 154-page Telugu novel names several of its fonts TT1D00O00, TT1D39O00 and TTE2371118O00, and the analyser identifies all three as Anu fonts from what they draw.
Getting the Telugu back
Legacy PDF to DOCX reads each code through a table built for the Anu family and writes real Unicode Telugu into a Word document: the heading above comes back as సమాచార హక్కు చట్టం. It reads the file’s codes rather than the printed page, so print quality does not matter, but the table is not perfect and the result is worth reading before it is relied on. If the analyser says some pages have no text layer, those pages are pictures of text, and Telugu OCR is the tool for them.