What searchable means inside a PDF. A scanned page is an image. To make it searchable, OCR recognizes the characters and the file stores them as real text, positioned over the words in the picture. PDF has a text mode for exactly this: text drawn in that mode is present for search, selection, and copying, but is not painted on the page. A reader finds the word, highlights roughly where it sits on the image, and the page still looks like the original scan.
What ScanRoute does at commit. OCR for the text layer runs when you commit, not while you plan the batch. For each page that will be written, ScanRoute checks whether the page already has text. If it does, the page is left alone. One known exception: text stored inside a reusable form object on the page is not inspected, so such a page can get a duplicate invisible layer; it looks the same, and a search may find a word twice. If the page has no text, the page is read with the Windows OCR language installed on the PC, and each recognized word is drawn in the invisible text mode, sized to the box OCR reported for it. The page image and its content are imported unchanged, and the layer is added on top. The setting is called Add a searchable text layer; it is on by default, and you can turn it off in Settings.
Which pages are skipped. Some pages get no layer, and each one is recorded per page in the TextLayer column of the CSV manifest. SkippedRotated means the PDF marks the page as rotated, which is common with landscape printouts. SkippedNonStandardBox means the page's visible area is set up differently from the standard page box. SkippedOcrUnavailable means no Windows OCR language was installed. SkippedUnsupportedText means none of the recognized words could be drawn with the embedded font, which can happen with scripts that font does not cover. OcrFailed means OCR returned an error on that page. In every case the page keeps its image and is written as scanned; it is simply not searchable.
Why upright matters. The layer has to line up with the picture. ScanRoute only adds it where the page geometry is the plain, unrotated case, rather than risk text that sits in the wrong place. A rotated page in the output looks right and reads right to a person; it just will not answer a search.
Checking the result. Open the manifest after a commit and look down the TextLayer column for skips. Then open one of the written PDFs and search for a word you can see on the page. We checked the invisible mode with the Windows PDF renderer and one other PDF library. Printing and other viewers were not tested, and a viewer that ignores that mode could show the text, so try a page in the reader your office uses before relying on it.
What it does not do. The layer is not editable text and does not turn a scan into a word-processing document. It does not touch your original scan, which ScanRoute only reads; the searchable pages are in the new PDFs written to the destination folder. Because ScanRoute is built to split and name combined scans, the text layer comes with that work rather than as a separate convert-a-file tool. And OCR can misread, so a search can miss a word the scan blurred.