Genius Scan OCR: Text Recognition on a Phone Scan

The OCR in Genius Scan is a feature called Text Recognition. The Grizzly Labs, which makes the app, says it reads the words in a scan on the phone, indexes them for search inside the app, and adds a text layer to PDFs you export. It is switched off by default and belongs to the paid plans.

So the short answer to "does it give me a searchable PDF" is yes, according to the vendor, once three conditions are met: the right plan, the setting turned on, and the document processed. This post takes each from The Grizzly Labs' own help center, pricing page and privacy policy.

Does Genius Scan have OCR?

Yes. The Grizzly Labs' help center has a section titled "Text Recognition / OCR", and says the feature is one "you might also know as OCR, short for Optical Character Recognition". It is not active when you install the app: the same page says Text Recognition is "disabled by default".

The vendor's page on Text Recognition in the Genius Scan help center lists what the feature allows:

• indexing and searching the content of your documents
• creating searchable PDFs, described this way: "when OCR is enabled, your exported PDFs will include a text layer which means that you will be able to copy their text"
• extracting details from business cards and receipts, including creating contacts and generating expense reports
• renaming a document based on content detected in the scan
• sharing the text of a document by exporting it as a text file

Turning it on

The steps on the vendor's page for enabling text recognition are the same on iOS and Android:

1. Open the Settings.
2. Tap Text Recognition.
3. Turn Text Recognition on.

The page then says: "Text Recognition will be applied to your new documents." It also carries a warning to have at least one active language in the Manage Languages menu, "or the Text Recognition feature won't be able to process your documents."

What the vendor says it is built for

On its page about improving the quality of Text Recognition, The Grizzly Labs writes that the feature "is designed to work on typed texts" and that "Handwritten texts might get poor results." Its tips are to activate the proper languages, crop the document close to the text, and use the Black and White filter in its Strong variation.

For a law office, that statement sets expectations. A printed notice or a typed contract is the intended input. A form filled in by hand, or a page with margin notes, is outside what the vendor says the feature is designed for.

Open the exported PDF on a computer and search for the contract date, a name on the notice, and the invoice total when one is in the stack. For an affidavit, search the venue line and the signature date, then read both against the page image. Treat a search by the document title inside the app as its own test.

Which Genius Scan plans include Text Recognition?

The paid ones. On the vendor's pricing page, the Text Recognition (OCR) row is ticked for Ultra and for Team and not for the free Basic plan. The help center also names an older plan, Genius Scan Plus, as including it. Check which plan your own installation is on before you count on the feature.

The vendor's pages use different plan names in different places, so here is what each one says.

Read side by side, these fit together: a new subscriber gets Text Recognition through Ultra or a Teams licence, and someone still on the legacy Plus plan has it too. The pricing page is the place to check current prices, which this post does not repeat.

Questions to settle before you rely on it

• Which plan is this phone on? The help center has a page called What plan do I have? for this.
• Whose subscription is it? A subscription bought on a personal account for a phone used on firm matters is a question for whoever manages firm software.
• Were older scans processed? The vendor says documents scanned before you upgraded "have not been processed". Its page on applying Text Recognition to old documents gives the fix: in Settings, choose Text Recognition, then choose Process old documents.

The help center's page on why Text Recognition isn't running gives three reasons: the plan, the default-off setting, and having no language activated. Those are the first three things to check when an exported PDF turns out to have no text.

What Genius Scan puts in the PDF it exports

According to the vendor, a document that has been processed with Text Recognition exports as a PDF with two layers: the picture of each page and a text layer over it. The same recognized text can be exported alone as a TXT file. A document that was not processed exports as pictures only.

The text layer

The help center page on searching scans with their content puts it this way: "With the Text Recognition feature, the PDFs you create will have a text layer in addition to the image layer." It continues that once a document has been processed with OCR, "it will contain a text layer when you export it as a PDF."

The Grizzly Labs' own blog post on the benefits of Text Recognition describes the mechanism: "When Genius Scan exports a document as PDF, it overlays an invisible text layer on every page". That is the standard build of a searchable PDF, and the guide to how to make a scanned PDF searchable explains how such a layer works and how to check one.

The order of events matters. Recognition happens when the document is scanned or edited, and the layer is written when you export. The vendor says that after you activate the feature, "every new document that you scan (or any page that you edit) will be re-processed with OCR." A scan made before you turned the setting on has nothing to write until you process it.

The TXT export

The page on exporting scans as text says: "If Text Recognition is applied to your document, Genius Scan can export it as Text." On iOS, the vendor's steps are to open the document, tap the Text Recognition button, expand the Text tab, and tap Share to copy or export the text. On Android, the Format choice in the share screen includes TXT.

A TXT file has the words and nothing else: no page image, no signature, no layout. It suits pasting a paragraph into a draft. It is not a copy of the document.

Export settings that affect the PDF

The vendor's page on adjusting export settings lists four resolution choices at export: Low, Medium, High and Actual. It warns that "Reducing the resolution will also reduce the quality of your scan." The same screen is where you choose JPEG, PDF or Text, and where a password can be set for a PDF. The plan comparison page also lists PDF/A export as an Ultra feature. The post on what PDF/A is explains when that format matters.

Where does Genius Scan run its text recognition?

On the phone, according to The Grizzly Labs. The pricing page describes the feature as "On-device text recognition", and the privacy overview in the help center says "all scanning operations are performed locally (on the user's device)". Documents leave the device when you choose an export, a backup or an account.

The vendor's statements, quoted

• The pricing page, under privacy: "Documents are processed and stored on your phone unless you explicitly decide to export them."
• The Privacy and Security Overview: "your documents are only stored on your device by default".
• The Genius Scan privacy policy: "Your documents are only stored on your device and are never sent to a third party unless you explicitly choose to." A later line in the policy reads: "They are also processed exclusively on your device (OCR, PDF generation, image processing...)."

When documents do leave the phone

The privacy policy lists the opt-in events that store documents outside the device: exporting to other apps, exporting to third-party services, and creating a Genius Scan account. On the last one it says: "your documents will be stored on our servers so that Genius Scan can synchronize them between your devices and back them up." It adds that the company "may provide additional services such as indexing your documents for better search or suggesting better titles".

Two more points from the policy are worth a lawyer's attention:

• Cloud exports go direct. For services such as Box, Dropbox and Google Drive, the policy says the interactions "happen directly between the mobile application and the cloud service, and your data does not transit through any server managed by The Grizzly Labs."
• An optional renaming feature sends text out. Under "Genius Renaming", the policy says: "We process the text of the first page of your scanned document to suggest document names. OpenAI may keep the text for up to 30 days for abuse monitoring." It states that the feature is optional.

This post reports what the vendor has published. It does not verify it, and it is not advice on whether the app fits your duties to a client. If you scan client documents on a phone, read the policy yourself, decide whether an account, cloud sync and the renaming feature should be on, and write the decision down. The post on whether it is safe to upload client documents to an OCR tool lists the questions to ask of any tool.

Searching your Genius Scan library by words inside a document

With Text Recognition on and a document processed, the app's search field finds documents by words inside them. Without it, the vendor says, "the search field in Genius Scan only searches in the title and tags of your documents."

The vendor's three steps, from its page on searching scans with their content, are:

1. Make sure Text Recognition is activated.
2. Make sure the document has been processed with Text Recognition. The page notes: "By default, OCR will only apply to new documents. You can go into the Genius Scan settings to process old documents."
3. Search for the document based on its content.

Its own example is finding an old receipt by typing a shop's name that appears only on the receipt and not in the title.

What in-app search is and is not

In-app search answers "which scan on this phone mentions the Hartwell account?" It is a finding aid for the person holding the phone. It does nothing for a colleague, and it stops being useful the moment the document is deleted from the app.

That is a different thing from the text layer in an exported PDF, which travels with the file and works in any reader that searches PDFs. A firm needs the second. The first is a convenience on the way there.

A quick check that search is working

Scan a printed page with a distinctive word in the body, a surname and a dollar figure. Give the document a title that contains none of them. Then search for each of the three. If the body word is found and the figure is not, you have learned how far to trust search for numbers on your own paper. If nothing is found, run through the vendor's three reasons: plan, setting, language.

A Genius Scan routine for pages collected outside the office

Set the app up before you leave, capture and check on site, export a processed PDF to an approved destination, and verify the file at your desk. The routine has nine steps, and it assumes Text Recognition is on.

Suppose Marisol Vega, an associate, is sent to a storage unit to copy a deceased client's papers for the executor, Mr. Lindqvist. She expects about forty pages of typed letters and account statements.

Before leaving

1. Confirm the plan and the setting. Settings, Text Recognition, on. Confirm a language is active.
2. Confirm the export destination. Use one your firm has approved, and sign in to it while you still have a good connection.
3. Decide on the optional features. Based on the privacy policy, decide whether the renaming feature and account sync should be off for this matter.

On site

1. Scan each document as its own multi-page file. A three-page letter is one document. A statement is another. The post on how to OCR a photo of a document covers light, glare and perspective.
2. Name each one as you go. Lindqvist estate - bank statement 2024-03 beats a date stamp when you are sorting forty files later.
3. Open the Text tab on one page of each document. If the text view is empty or nonsense, rescan while the paper is still in front of you.

Back at the office

1. Export each document as PDF at a resolution your firm has settled on, and open it on a computer.
2. Test the file. Search for a word you can see, then copy a line and paste it. The routine for confirming that a PDF is really searchable goes page by page.
3. Clear the phone. The vendor documents a setting to automatically delete a document after export, switched on per export destination, and warns that documents deleted this way cannot be recovered in the app. Whether you use it or delete by hand, a phone should not be the long-term home of estate papers.

When the phone is the wrong tool

Forty pages is a reasonable job for a phone. Four hundred is a job for a sheet-fed scanner, and the guide to digitizing law firm paper files covers the scanner settings for that. If you already use the scanner built into the iPhone, the post on scanning documents with Apple Notes explains what Apple does and does not say about its output.

Files that reach you already scanned by someone else are a different case again, since there is no page to photograph. A conversion service takes the file as it is. Schema Law accepts a PDF, PNG or JPEG of up to 20 MB and exports a searchable PDF or plain text, among other formats, and its four commitments on documents are stated on the site. Recognition there is strongest on clean print and weaker on handwriting, and nothing marks a passage as uncertain, so check names, dates and amounts against the page.