Create Searchable PDF Files: Pick the Route by What You Have

To create a searchable PDF, start from what you hold. Paper goes through a scanner and then through text recognition. A scan that is already a PDF needs recognition only. Loose page images are joined into one file first. An office file is exported straight to PDF and needs no recognition at all.

Those four routes share an end point and very little else. The work, the time and the mistakes differ for each one, so sorting the job by its input saves you from running OCR on a file that never needed it, or from skipping it on one that did.

Finish every route the same way. Search the new file for the party names on the first page and for the federal court or state court named in the caption. Then search a Bates number from the middle of the file. Read the hit against the page image. An exhibit cover that shows its label on the picture and returns nothing is not finished.

Which starting point do you have: paper, a scan, loose images or an office file?

You have one of four starting points, and the file extension alone will not tell you which. Paper is physical pages. A scan is a PDF whose pages are pictures. Loose images are separate PNG, JPEG or TIFF files, one per page. An office file is a document you can still open in the program that wrote it.

Sort the job with three questions.

1. Is it on paper or on a screen? Paper is its own route, because the scanner settings decide how well everything afterwards works.
2. If it is on a screen, can you open it in a word processor, spreadsheet or presentation program and edit it? If so, it is an office file, and the characters are already there to be carried into a PDF.
3. If it is a picture of a page, is it one PDF or a folder of image files? One PDF is a scan. A folder of numbered images is the loose-image case, which is what a phone camera session or an export from an older imaging system leaves you with.

Starting points that look like something else

Some inputs disguise themselves. Check for these before you choose a route.

The last row matters more than it looks. When the editable source exists, the cleanest searchable PDF comes from that source, and scanning the printout throws away perfect text in order to guess at it again. Ask whether the source exists before you feed a stack into the scanner.

Creating a searchable PDF from each of the four starting points

Each starting point has its own route to a searchable PDF. Paper is scanned and then recognized. A scanned PDF is recognized as it stands. Loose images are put in order, combined into one PDF and then recognized. An office file is exported to PDF with its text intact, with recognition needed only for any pictures inside it.

From paper

Paper has to become a picture before it can become text. Some office scanners and their software recognize text during the scan and hand you a searchable file in one pass. Others produce a plain picture PDF that you convert afterwards. The post on going from paper to a searchable PDF at the scanner covers the device side, including where the file lands and what to test on the first sheet.

From a scanned PDF

A scanned PDF is the shortest route, since the pictures already exist. You upload or open a copy, let the software read each page, and save a PDF that carries the recognized words behind the pictures. The guide to making a scanned PDF searchable lists the steps to add a text layer to a scan, and that procedure is the one to follow here.

Schema Law is one tool for this route. It accepts a PDF, PNG or JPEG of up to 20 MB, reads the pages into text and tables, and offers Searchable PDF among its exports. The formats section of the homepage describes that export as your original file with a hidden text layer.

From loose images

A folder of page images is not yet a document. Put the files in page order, confirm none is missing, and combine them into one PDF so the document has one page count and one name. Then run recognition on the combined file. The guide to turning scanned images into one PDF covers ordering and combining. A TIFF needs converting to PDF, PNG or JPEG before a tool that takes only those formats will accept it.

From an office file

A document saved to PDF from the program that wrote it carries its characters into the PDF, so search works without OCR. The risk is an export setting that turns text into a picture, and Microsoft's instructions for saving Office files as PDF list one. The post on converting Word to a searchable PDF quotes that option and goes through the export paths.

What must a newly created searchable PDF pass before you call it done?

A newly created searchable PDF is done when it passes four checks: the text is there and reads correctly, the page count matches the source, the page pictures look like the original, and the file name tells it apart from that original. Selecting one word proves none of them.

Run the checks in this order, because each one is cheaper than the one after it.

• Page count. Count the pages of the source, on paper or in the original file. The searchable PDF must report the same number. A feeder that pulled two sheets at once or a split that dropped a part shows up here.
• Text on an early, middle and late page. Search for a word you can see on page one, on a page near the middle and on the last page. All three searches should land on the right page.
• A name and a number. Search for one surname and one figure exactly as printed. These are the items you will search for later, and one wrong character in either hides the hit.
• Picture. Put one page with a signature, stamp or handwriting next to the same page of the source. Nothing should be missing, cropped or redrawn.
• Name. The searchable copy has its own file name. The source keeps the name it arrived with.

How far to take the text check depends on what the file is for. A reading copy of a short letter needs the three searches and no more. A production set that people will rely on for "no hits" needs testing through the file, and the post on checking whether a PDF is searchable covers that across pages and across a folder. For the picture check, the post on whether OCR changes the original PDF explains what can differ and how to show that nothing did.

Whatever tool made the file, read important passages against the page. On the Schema Law homepage, the answer about scan quality says recognition is strongest on clean print and weaker on faint faxes, stamps over text and handwriting, and that nothing marks a passage as uncertain. Treat that as true of any text layer until your own checks say otherwise.

Creating searchable PDFs for one matter versus for the whole office

A single matter is a one-off job: convert the files you need, check them, and file them. An office-wide practice is a routine: a fixed point where every incoming document becomes searchable, a naming rule, and someone who owns the step. The same tool can serve both, but the decisions differ.

The one-off job

Take a fictional example. Marisol Okafor, a paralegal at a two-lawyer firm, receives 14 PDFs from a client in a lease dispute between Tarrant Mills LLC and Bexley Cold Storage, Inc. Eleven are scans, two were saved from Word, and one is a folder of 22 phone photos of a handwritten rent ledger.

She sorts before she converts:

1. The two Word PDFs already search. She tests each with one search and files them unchanged.
2. The eleven scans go through OCR as they are. She keeps the received files in a folder named received and saves the searchable copies beside it with _ocr added to each name.
3. The 22 photos are put in order by the page numbers written on the ledger, combined into one PDF, and then converted. She notes in the file that the ledger is handwritten and that search on it is unreliable.

The whole job is 14 files, and the only standing decision is the naming suffix. Nothing about it needs a firm policy.

The standing routine

An office that wants every document searchable has to decide where in the day the conversion happens. Three places are available: at the scanner when paper is scanned, at intake when a file is saved to the matter, or in a periodic sweep of folders. Each needs an owner, because a step that belongs to nobody is skipped on a busy day.

Questions to settle once, in writing:

• Which copy is the record, and where does the other one live?
• What suffix or folder marks a searchable copy?
• Who tests a file, and how much testing is expected for ordinary correspondence compared with evidence?
• What happens to documents that arrive already searchable?
• How many pages a month does the office convert, so the plan or licence fits?

Volume drives the last question. Schema Law's plans, for instance, are sized by pages a month and pages per document, and the pricing page lists the figures. For the daily paper side of a routine, see the post on scanning incoming mail to searchable PDF, and for a backlog folder, the post on bulk OCR for PDFs.

Who should create the searchable PDF: the sender, the scanner or you?

The best place to create the searchable PDF is as close to the source as possible. A sender who holds the original electronic file can give you true text. A scanner can add recognized text as the paper goes through. You, at the end of the chain, can only recognize text from whatever picture reached you.

The sender

When the document was written on a computer, the sender can save it to PDF with its text intact, and no recognition is involved. That file is better than any OCR result, since nothing was guessed. Whether a party in litigation must produce documents in a searchable form is a question for the rules, orders and agreements in your case, and this post does not answer it.

The scanner

When the document exists only on paper, the person scanning controls the quality of every later step. If your own office scans, set the device to produce searchable PDFs or route its output to a conversion step. If an outside vendor scans, say in the order what you expect. The post on ordering searchable PDFs from a scanning service has a specification to hand over.

You

When the file arrives as a picture and nobody upstream will redo it, the conversion is yours. That is where you stand with a document from a client, an opposing party or a records custodian when no better file is on offer. Keep the file as received, convert a copy, and check it.

If you also need to edit the language, a searchable PDF alone will not do it. The post on making scanned documents searchable and editable explains why two files serve better than one.

Slips that leave a newly created PDF only half searchable

Three slips leave a newly created PDF half searchable or leave you without a record of what you received: pages the recognition skipped, an export in the wrong format, and a searchable copy saved over the original. Each has a short test to run on the day the file is made.

Skipped pages

A file can search on some pages and stay silent on others. Sideways pages, very faint pages and pages that already held a scrap of text are the ones to suspect. The test is a search on a late page and on the worst-looking page. If recognition ran and whole pages still return nothing, the post on a PDF that is still not searchable after OCR works through the causes.

The wrong export

OCR tools offer several outputs, and only one of them is a searchable PDF. A Word or plain text export holds the recognized words but is a different document, with no page pictures. If you opened the result and the pages no longer look like the scan, you exported an editable format. Go back and choose the searchable PDF output.

A related slip happens with office files. Printing a document to a PDF printer with an image option switched on, or printing and rescanning a signed copy, produces a picture where text was available. Export from the program instead, and scan only the pages that changed on paper.

The overwritten original

Saving the searchable copy under the same name as the source leaves you with one file and no way to show what you received. Adding a text layer changes the file even when every page looks the same. Keep both, with different names, from the first save.

Text lost after creation

A PDF that was searchable on Monday can stop being searchable after someone prints it to PDF, flattens it or passes it through a system that hands back pictures. That is a separate problem from creating the file, and the post on why a PDF is no longer searchable covers the steps that strip a text layer.

A short record prevents repeat work. Note which route each file took, which pages you tested and which pages were poor. The next person to open the folder then knows what "searchable" means for that file.