Document contains existing invisible text layer. Should this existing data be used or should text recognition be run using the internal recognition engine?

Document contains no text layer. Recognition must be run before tables can be extracted.