What breaks in document AI

Wrong or unverifiable citations

10 of 10 quotes

• “If a table continues across a page break, its chunk gets the page number of the table's **last** page.”

• “It very, very routinely points at articles in its RAG "citations" that simply don't support the (modestly likely to be correct) assertions.”

• “offsets break when someone re-uploads an edited version of the same document. Every citation silently points a few lines off.”

• “So on \~30% of the grounded quotes the text is correct but the citation points at the wrong chunk (cites p73, the line is in p72).”

• “The blue highlight box is displaced to the **page header** area instead of the actual cited table rows.”

• Hit this one too? We are building for it. Join the waitlist.

• “we keep shipping RAG responses that look grounded, cite real chunks, and are still subtly wrong.”

• “Wrong / inconsistent page citations — The model often cites pages that don’t actually contain the claimed information, or the right sidebar shows different pages than what the model referenced in the answer.”

• “Help with local RAG pipeline – poor retrieval quality, wrong page numbers”

• “The system is unable to display the original citation, although the 'display original text' option has been properly configured.”

• “the output of agent can not directly jump into the origin pdf file. but it works in previous version”

• Hit this one too? We are building for it. Join the waitlist.