Searchable PDF: What the File Holds and How to Spot One
By Schema LawPublished Updated
A searchable PDF is a PDF whose pages carry real characters that a reader program can find, select and copy. A scan becomes one when OCR stores recognized text with the page picture. It differs from a plain scan by having that text, and from a word-processor PDF by where the text came from.
"Searchable" describes a capability, not a quality. A file earns the label the moment any text is present, whether or not the text is right. The sections below cover what is inside such a file, the two ways one is built, how text and picture are matched up, and what a recipient expects when asking for one.
What does a searchable PDF hold that a plain scan does not?
A searchable PDF holds characters. A plain scan holds one picture per page and nothing else, so a computer has no letters to match. A searchable PDF made from that scan keeps the picture and adds the page's words as text, which is what lets Ctrl+F return a result.
People also say "text searchable PDF" or, for a scan that has been through recognition, "OCR searchable PDF". All three describe the same property. What they leave open is where the characters came from.
- In a PDF saved from a word processor, the characters were typed. They are exact.
- In a searchable scanned document, the characters were recognized from the picture. They are a reading.
That difference in origin is the subject of the post on native PDF versus scanned PDF, which sets the kinds of file side by side. This post stays with the scan that has had text added.
A page of such a file has two parts that do separate jobs.
| Part | What it is | What it is for |
|---|---|---|
| Page picture | The scan, exactly as the scanner or camera captured it | Reading, printing, showing stamps and signatures |
| Text | Characters stored on the same page, usually invisible | Search, selection, copying, indexing |
Because the jobs are separate, each part can be good or bad on its own. A crisp picture can sit over careless text, and a murky fax can sit over text that happens to be accurate. How software writes the text into the file is explained under what makes a scanned PDF searchable.
Two builds of searchable PDF: picture with hidden text, or rebuilt page
Searchable PDFs made from scans come in two builds. One keeps the scanned picture as the visible page and hides the recognized text with it. The other discards the picture of the words and redraws the page in real fonts. Both are searchable. Only the first still shows you the paper.
| Picture with hidden text | Rebuilt page | |
|---|---|---|
| What you see | The scan | Fresh type set by the software |
| Where a misreading shows | Nowhere on screen | On the page itself |
| Stamps, initials, handwriting | Shown as scanned | Kept as picture fragments, or lost |
| File size | About the size of the scan | Often much smaller |
| Sharpness when zoomed | As soft as the scan | Crisp |
| Suited to | Exhibits, productions, the file copy | Internal reading copies |
Why the difference matters in legal work
With hidden text, a misread word never changes what a reader of the page sees. The scan says "Meridian Trust" even if the text behind it says "Meridian Tmst". The error costs you a search hit and nothing else.
With a rebuilt page, the software's reading becomes the page. If it read "Tmst", the page now prints "Tmst", in tidy type that looks authoritative. A judge or opposing counsel who opens that file is no longer looking at your document. For anything that must show the paper as it was, choose the build that keeps the picture.
Someone who asks for a searchable copy of an exhibit wants the party, the damages figure, and each citation available from the keyboard. Leave the picture in place as the proof of the paper. Text that misreads a citation produces a miss, and the printed amount remains visible on the page.
Some tools mix the two, redrawing body text and leaving pictures for signatures and logos. Treat a mixed page as rebuilt.
How to tell which build you have
- Zoom to 600% on a line of body text. Ragged, grainy edges mean a picture. Perfectly smooth edges mean redrawn type.
- Look at a signature or stamp. If it has a faint rectangular border or a different background shade from the text around it, the page was rebuilt around it.
- Compare the file size with the original scan. A file a tenth of the size has almost certainly had its pictures replaced or recompressed.
How is the hidden text of a searchable PDF lined up with the picture?
The hidden text of a searchable PDF is lined up by coordinates. Every piece of text is stored with a position and a size, and the software sets those so the invisible words sit where the visible ones are. How finely it does this, by word, by line or by page, decides where a search highlight lands.
When your reader finds a word, it highlights the place where the hidden text says the word is. It knows nothing about the picture. If the hidden word sits squarely under the printed one, the highlight looks right. If not, the highlight drifts.
| Placement | What the software stores | What you see on a search |
|---|---|---|
| Word by word | A position and width for each word | Highlight on the word |
| Line by line | A position for each line; words spaced by the hidden font | Highlight on the right line, a little left or right of the word |
| Page by page | The page's text in reading order, without matching each word | The right page; the highlight may be elsewhere on it |
Why a highlight lands off the word
A highlight lands off the word when the stored position and the picture disagree: the tool placed text by line or by page, the scan was tilted, or the page was cropped or rotated after recognition. The post on scanning a PDF for text goes through each cause under why a PDF search highlights the wrong spot.
An off-target highlight is a cosmetic problem as long as the right page comes up. It becomes a practical one if you plan to mark up or redact by searching, because tools that act on search results act on the hidden text's position.
Schema Law's searchable PDF export is of the page-by-page kind: your original file with the recognized text of each page written in reading order, not pinned under each word. A search takes you to the correct page, and you then find the word on it by eye. The export formats section of the homepage describes it as a hidden text layer "so search finds the page".
Searchable does not mean correct: the text behind the picture
A searchable PDF can be searched whether its hidden text is accurate or not. The label says characters are present. It makes no promise that they match the page. Since the text is invisible, a file with a poor reading looks identical to one with a perfect reading until you test it.
This is the trap in the word. A copier with a "searchable" setting, a free converter and a careful review vendor all produce files that pass the same quick check: type a common word, get a hit. The check proves that recognition ran. It says nothing about the surname on page 40.
Looking at the text you cannot see
You can bring the hidden text into view in under a minute. Select all the text on a page that matters, copy it, paste it into a blank document and read it beside the page. What you pasted is what search runs against. Look for three things: wrong characters in names and figures, lines that are missing, and lines in the wrong order.
For a file you intend to rely on, the fuller set of checks is in the section on how to confirm a PDF is really searchable.
What a bad text layer costs
| Problem in the hidden text | Effect |
|---|---|
| A name misread once | That page is missed by a search for the name |
| A digit misread in an amount | A search for the figure fails; a copied figure is wrong |
| A whole page with no text | The page is invisible to every search |
| Two columns interleaved | Phrase searches fail; copied passages are scrambled |
| Text from an older recognition pass still present | Duplicate or conflicting hits |
Searchable scanned documents arrive from many hands, and you seldom know what made the text. When a search is going to support a statement such as "the term does not appear", the hidden text has to be tested first. The post on why OCR output is full of errors explains where misreadings come from.
Is a text searchable PDF the same as an editable one?
No. A text searchable PDF lets you find and copy words. An editable document lets you change them and have the page reflow. A searchable scan is still a picture with text attached, and changing that text does not change the picture.
The confusion is understandable, since both involve "real text". But consider what happens if you open a hidden-text PDF in an editor and alter a word in the text layer. The scan still shows the old word. Search now finds the new one. You have made a file that says two different things, which is worse than either version alone.
| Property | Meaning | A searchable scan has it? |
|---|---|---|
| Searchable | A reader can find words | Yes |
| Selectable | You can highlight and copy text | Yes |
| Editable | You can revise wording and the layout adjusts | No |
| Fillable | It has form fields you can type into | No |
| Tagged | It carries structure for screen readers | Not from the text layer alone |
If you need to revise the language of a scanned agreement, export it to a word-processing format and work there. The guide to converting a scanned PDF to Word covers that path. Keep the searchable PDF as the record of what the paper said, and treat the Word file as a draft built from it.
A rebuilt-page PDF is closer to editable, because its words are real type. Even so, a PDF editor changes a line at a time and does not reflow a contract the way a word processor does.
When a recipient asks for the searchable PDF format
A recipient who asks for the searchable PDF format wants a file they can open in any reader and search straight away. Courts, clients and opposing counsel mean roughly the same thing by it, but each has its own written requirement, and that document controls, not the general idea.
Courts
Courts that address searchability in their electronic filing rules do so in different words and with different exceptions. Two examples show the range. California's Rule 2.256 on electronic filing says a document "must be text searchable when technologically feasible without impairment of the document's image." The Second Circuit's guidance on working with PDFs says a PDF filed in its system "must be text-searchable", and that one that is not will be returned to the filer.
Those are two courts. Yours may ask for more, less, or something different for exhibits. The post on searchable PDFs for court filing goes through the questions to put to your own court's rules. This is a description of documents, not advice on what to file.
Clients and opposing counsel
Outside court, the request is usually practical. A client's records department wants files its system can index. Opposing counsel wants to avoid running recognition on your production. In both cases, ask two things: whether they need one PDF per document, and whether they also expect separate text files.
Archives
Some recipients ask for an archival flavor of PDF meant to stay readable for decades. A searchable scan can be saved that way, with conditions. The post on what PDF/A is explains them.
Before you send one
- Test the hidden text on three pages, one of them poor.
- Confirm the picture is the scan and has not been redrawn.
- Check that redacted pages carry no text under the boxes.
- Check the file size against the recipient's limit.
- Keep the original scan, unaltered, under its own name.
If you do not have a tool for the conversion, the searchable PDF is one of the eight export formats on every Schema Law plan, including the free plan, which takes one document a day of up to three pages. The pricing page shows the limits on the larger plans.