---
title: "OCR PDF Online - Make Scanned PDF Searchable - CartHacker"
description: "Read the text off a scanned PDF with Optical Character Recognition (OCR) and weld it back on as an invisible, selectable layer, so the page looks identical and Ctrl+F starts working. Runs in your browser, no upload."
url: "https://carthacker.com/pdf-tools/ocr-pdf/"
image: "https://carthacker.com/wp-content/themes/carthacker-core/assets/og/pdf-ocr-pdf.jpg?v=1791719545"
publisher: "CartHacker (carthacker.com)"
language: "en"
---

# Make a Scanned PDF Searchable: Read the text off a scan and weld it back on as an invisible, selectable layer

Drop the scanned document into the emerald zone below and press the button. **Optical Character Recognition, which is what the three letters OCR stand for, reads the printed characters out of a page image**, so every page here is rasterized, straightened, read, and given an invisible text layer positioned exactly over the ink. The page still looks pixel for pixel identical and Ctrl+F starts working. Printed and typed characters are what it reads; handwriting is not. Pages that already carry real text are left completely alone, because a guess is worse than the text that was already there. **Nothing is uploaded and no language pack is downloaded.** The whole recognizer is 60 KB.

- Files never leave your device
- No installation, no watermark
- Works offline once the page is open

OCR PDF online, Live

- **Privacy:** 100% private
- **Your plan:** Free to use
- **You get:** Searchable PDF

Your files are read by JavaScript running on your own machine. There is no upload, no queue, and no copy on our server to delete later.

The recognizer above works right now. **The guide below is unusually blunt about what it is good at and what it is not** - the kind of document it reads well, the kind it will produce nonsense from, the four things that decide how accurate any run will be, how the invisible text layer is aligned so that selecting a word highlights the right word, and why the confidence number on the result screen is the most useful thing on it.

## How to Make a Scanned PDF Searchable With OCR

A minute or two for a short document, longer for a long one, because every page has to be rasterized and read. This **OCR tool** runs inside the page you are reading, on your own processor, and **it does not download a language pack** - which is the reason it starts working immediately instead of after a 15 MB wait.

1. **Add the scan.** Drop the PDF into the zone above. A photograph of a document works too. The bigger and cleaner the original, the better everything downstream gets - a 300 dpi scan reads dramatically better than a 150 dpi one, and no amount of processing recovers detail that was never captured. Machine-printed and typed pages are what this reads; handwriting is not.
2. **Pick the reading resolution.** Leave it on 300 dpi unless the print is unusually small, in which case 400 helps. **Higher is slower and only helps if the text is genuinely fine** - rendering a page at 400 dpi that was scanned at 150 adds time and no information at all.
3. **Leave the skip switch on.** Most documents people bring to a recognizer are partly digital already. **Any page that already carries real text is left exactly as it is**, because recognized text is a considered guess and text that is already in the file is the real thing. The result screen reports how many pages were skipped for that reason.
4. **Run it and read the confidence.** Press Run OCR. The result screen reports how many words were read and the average confidence across them. **Above about seventy percent the layer is genuinely useful for search. Below it, the scan is fighting the recognizer** and the notes on screen say exactly that instead of reporting a success.

The download looks exactly like the scan you put in. That is the point: the page image is untouched and the text sits behind it, invisible, doing nothing except answering searches and letting you select a paragraph.

## What This Recognizer Is Good At, and What It Is Not

Stated first, before you spend three minutes on a document it cannot read. **This is a template matching recognizer tuned for clean, machine-printed pages.** That is a real limitation and a real strength. Accuracy is decided by four things: the resolution the page was scanned at, the contrast between the ink and the paper, how far the page is rotated or skewed on the glass, and how cleanly the original was printed in the first place. No honest tool quotes you a percentage before it has seen your file, so what follows is a list of real documents and what each one gets.

| Document | Expect |
| --- | --- |
| A laser-printed letter or contract scanned at 300 dpi | **Very good.** This is the case it was built for and it reads it accurately. |
| An invoice, statement or form from an office printer | **Very good.** Regular typefaces at sensible sizes, cleanly separated. |
| A phone photograph of a document in decent light | **Good.** The lighting is flattened and the page is straightened before anything is read. |
| A fax, a photocopy of a photocopy, or a 150 dpi scan | **Mixed.** Some words come through, some do not. Check the confidence number. |
| A newspaper, a magazine spread, a dense multi-column layout | **Mixed.** Columns are read in a sensible order but the format fights the segmentation. |
| Handwriting, cursive, or a signature | **Nothing useful.** Handwriting recognition is a different discipline and this tool does not pretend to do it. |
| Heavily stylized display type, script faces, gothic | **Poor.** The reference alphabet is ordinary text faces. |

**The confidence number on the result screen is not decoration.** It is the average margin by which each character beat its runner-up, weighted by length. Above seventy percent the search layer is dependable. Below it the notes say so plainly, and you should read the extracted text before relying on it.

### Why there is no 15 MB engine behind this

The usual approach is a WebAssembly build of a full desktop recognition engine, meaning desktop code compiled to run inside a browser tab, plus a trained language pack. Together those run to fifteen megabytes or more. That is a real cost, and on this site it would be paid by **everybody who lands on this page from a search result**, most of whom are here to read.

So this is our own engine, sixty kilobytes, loaded on the first file that touches the drop zone and never before. The trade is stated here and not buried: **a heavyweight engine reads harder documents than this one does.** On clean office paper, which is what people actually scan, the gap is far narrower than the size difference suggests. On a fax or a photocopy of a photocopy, it is not.

## The Engineering Behind Reading Text Off a Scan

Seven stages, and most of the accuracy comes from the four before any character is identified. Recognition is mostly preparation.

1. **Flatten the lighting.** A scan is evenly lit and a photograph never is. A heavily blurred copy of the page is, by definition, the paper without the text on it, so every pixel is divided by its own local paper level. A page in the shadow of a hand and a page under a lamp end up the same.
2. **Separate ink from paper.** The threshold is chosen by Otsu's method, which finds the split that best divides the two populations, and it is taken from the **middle of the flat region** in the histogram rather than the first best answer. A page of text is about one percent ink, and taking the first maximum puts the cut on the shoulder of the ink itself - which silently eats thin stems and makes whole letters vanish.
3. **Find the letters.** Every connected blob of ink is collected, eight-connected so that a diagonal hairline in a serif stays one stroke. Specks with no neighbors are dropped as sensor noise.
4. **Straighten the page.** The skew is measured by fitting a line through the **bottoms of the letters**, which are the baseline, rather than by searching for the angle that makes a projection sharpest. Descenders and punctuation are excluded from the fit, because a line ending in "typography" would otherwise tilt the whole page.
5. **Find the lines and the words.** Components are grouped by vertical overlap, and each line gets its baseline, x-height and cap-height measured. Words are split on gaps measured relative to that line rather than against a fixed number of pixels, so a change of font size or resolution needs no setting.
6. **Identify each character.** Every glyph is normalized into a 16 by 16 grid sampled by area, which is what makes the same letter at 9 points and at 30 points produce nearly the same vector. That is compared against a reference alphabet **rendered on your own machine** from the faces already installed on it.
7. **Write the invisible layer.** Text render mode 3 draws nothing at all. The words are placed, sized and spaced exactly where the ink is, and then not painted.

### The part that separates an o from an O

Normalizing a shape into a fixed box throws away exactly the information that distinguishes a lower case o from a capital O - they have the same 16 by 16 grid to four decimal places. So the classifier does not only compare shapes. It also compares **where the glyph sits relative to the line and how tall it is**, measured in multiples of that line's own x-height, plus its aspect ratio, its ink density, and how many enclosed holes it has.

The hole count is unglamorous and it separates whole groups of letters at a stroke: B and 8 have two, a b d e g o p q 0 4 6 9 A D O P Q R have one, everything else has none. Getting that from a flood fill of the background is far more dependable on a noisy scan than any attempt to analyze strokes.

- **Approach:** Template matching against an alphabet rendered locally
- **Reference faces:** Eight, spanning the sans, serif and monospaced faces on your machine
- **Features:** A 16x16 area-sampled grid, height, baseline position, aspect, density, hole count
- **Deskew:** Least squares through glyph baselines, applied only if the page is crooked
- **Output:** Invisible text, render mode 3, positioned over the ink
- **Page image:** Untouched. The scan is not re-encoded, re-compressed or altered.

## Why the Page Looks Exactly the Same Afterwards

This is the property that makes the whole exercise worth doing, and it is the one people get wrong most often. **Recognition here does not replace your scan with typed text.** It adds a layer of invisible text underneath the picture, positioned so precisely that selecting a word highlights the word you can see.

So the document still looks like a scan, because it is one. The signatures are still signatures, the letterhead is still the letterhead, the coffee ring is still there. What changes is that Ctrl+F works, text selects and copies, and any system that indexes documents can now index this one.

**The horizontal scale is computed per word, not assumed.** Helvetica at 11 points is not the width of whatever the document was actually printed in, so each word is stretched or squeezed until its invisible box is exactly as wide as its visible one. Without that, a long line drifts and by the right hand margin your selection is a whole word out of step.

### What this makes possible

- **Finding the document at all.** Every search index on every operating system reads PDF text and none of them read pictures. A folder of scans is invisible until this is done.
- **Finding something inside it.** Ctrl+F for an invoice number across a 200 page scanned bundle.
- **Copying a paragraph** instead of retyping it.
- **Accessibility.** A screen reader has nothing to say about an image of a page. After this it reads the document aloud.
- **Feeding it to something else.** [Convert to Word](https://carthacker.com/pdf-tools/pdf-to-word/) or [to Markdown](https://carthacker.com/pdf-tools/pdf-to-markdown/) only works if there is text to convert.

## How to Get Substantially Better Recognition

Almost all of the accuracy is decided before the file reaches this page. Five things, in order of how much they matter.

1. **Scan at 300 dpi.** The single biggest factor by a wide margin. At 150 dpi a 10 point character is about 20 pixels tall and the fine distinctions are simply not there. At 300 it is 40 pixels and everything downstream gets easier. **No amount of processing recovers detail that was never captured.**
2. **Scan in grayscale, not color and not pure black and white.** Grayscale keeps the soft edges the engine uses to normalize each glyph. A scanner's own black and white mode throws that away with a threshold chosen by firmware that cannot see the whole page.
3. **Keep the page straight.** The engine measures and corrects skew, but correcting it costs a little sharpness on every stroke. A page that was straight to begin with never pays that.
4. **Light it evenly if you are photographing it.** The lighting is flattened automatically, but a hard shadow across a third of the page still loses contrast. Daylight from the side beats a flash from the front every time.
5. **Do not compress it hard.** A heavily compressed JPEG scan has ringing around every letter, and that ringing is noise the recognizer has to work through. If your scanner offers a quality setting, use a high one.

And one that is not about the scan at all: **check whether the document already has text.** A great many "scanned" PDFs are ordinary digital documents that somebody described that way because their viewer would not let them select anything. Try Ctrl+F on the original first. The tool checks this itself and skips those pages, but knowing beforehand saves you the run.

## Zero-Trust Processing: The Scan Never Leaves Your Machine

Think about what people actually scan. Contracts, medical letters, bank statements, passports, tax returns, identity documents. **The category of document that most needs recognition is the category that most obviously should not be uploaded to a stranger.**

Every mainstream online recognizer is a server product: you upload the scan, a queue processes it, and you download the result. The document sits on a disk you cannot inspect for as long as their retention policy says, which you are asked to take on trust.

Here every stage runs in this tab: turning the page into pixels, splitting ink from paper, straightening the page, grouping the blobs into lines and words, identifying each character, and writing the file back out. There is no upload endpoint because there is no server side to the operation.

- **Scan upload:** None. Read from your disk by the browser.
- **Recognized text:** Stays in the tab and goes into your download. Nowhere else.
- **Language packs:** None downloaded. The reference alphabet is rendered locally.
- **Server storage:** Nothing. No copy is made.
- **Works offline:** Yes. Load the page, disconnect, recognize.

**Verify it in ten seconds.** Load this page, turn off your network, and run the recognizer on a scan. It works end to end. No upload based service can pass that test, and no privacy policy substitutes for it.

## The Limits, Stated Before You Hit Them

Every tool in this suite carries this section. This one has more to say than most, because a recognizer that oversells itself wastes your time in a way most tools cannot.

- **It does not read handwriting.** Not cursive, not hand-printed block capitals, not a signature, not a filled-in form. **Only machine-printed or typed characters are recognized reliably.** A handwritten page produces nonsense, and the confidence number on the result screen will say so before you trust any of it.
- **It reads the Latin alphabet.** English, and the Western European languages that share the character set. Cyrillic, Greek, Arabic, Hebrew, Chinese, Japanese and Korean are outside what it does.
- **Characters the standard PDF fonts cannot draw are dropped from the layer** rather than substituted. A wrong character in a search layer is worse than a missing one - the gap fails to match, the wrong character matches the wrong thing.
- **It does not rebuild the layout.** The output is a scan with searchable text behind it, not a reflowable document. If you need an editable file, run this first and then [convert to Word](https://carthacker.com/pdf-tools/pdf-to-word/), and expect to do some tidying.
- **It does not read tables as tables.** Cells come through as words in roughly the right places, which is enough to search and not enough to reconstruct a spreadsheet.
- **A very long document takes a while.** Every page is rasterized at 300 dpi and read. Two hundred pages is a few minutes on a laptop and longer on a phone.
- **An encrypted document cannot be read at all.** Encryption is not damage and it is not a scanning fault. Until the file is opened with its password the page content is unreadable to every piece of software, this one included. Run [Unlock PDF](https://carthacker.com/pdf-tools/unlock-pdf/) first, then come back and recognize it.

**Never run this before redacting a scan.** The layer it adds is invisible and fully selectable, and if the page contains something that has to be removed, recognition creates a searchable copy of exactly that. [Redact first](https://carthacker.com/pdf-tools/redact-pdf/), verify, then make it searchable.

## Where This Fits in a Real Working Day

Recognition is almost always the step that makes a pile of scans usable, and the order it comes in matters.

| What you are doing | The order that works |
| --- | --- |
| Filing a backlog of paper | [Photograph or scan](https://carthacker.com/pdf-tools/scan-to-pdf/) it, run recognition, then archive. A scanned folder with no text layer is a folder your search cannot see. |
| Something in the scan has to be removed | [Redact first](https://carthacker.com/pdf-tools/redact-pdf/), then recognize. Never the other way round. |
| You need the text as an editable document | Recognize here, then [convert to Word](https://carthacker.com/pdf-tools/pdf-to-word/). The converter needs text to convert. |
| A long-term archive | Recognize, then [convert to PDF/A](https://carthacker.com/pdf-tools/pdf-to-pdfa/). Archival formats want a text layer as much as they want embedded fonts. |
| A scanned contract that needs signing | Recognize first so the file is searchable, then [sign it](https://carthacker.com/pdf-tools/sign-pdf/). Doing it afterwards makes the recognizer read your own signature as text. |
| Several scans going into one archive | Recognize each one, then [merge](https://carthacker.com/pdf-tools/merge-pdf/). Recognizing the merged file works too and takes exactly as long. |

One habit for anybody filing paper regularly: **recognize at the moment of scanning, not later.** The scan you do not process today is the document you cannot find in eighteen months, and by then nobody remembers which of the four hundred files it was. The rest of [the PDF suite](https://carthacker.com/pdf-tools/) runs on the same architecture as this page - nothing uploaded, nothing stored, everything on your own machine.

## Frequently Asked Questions

### How do I make a scanned PDF searchable?

Drop the scan into the zone at the top of this page and press Run OCR, which stands for Optical Character Recognition. Each page is rasterized, straightened and read, and an invisible text layer is placed exactly over the ink. **The page looks identical and Ctrl+F starts working.** Nothing is uploaded and no language pack is downloaded. No account is needed for a first run: anonymous visitors get one operation per 12 hours counted by network address, a free account gets two, and CartHacker Premium removes the counter.

### Is my scanned document uploaded to a server?

No. **The file is read from your disk by the browser, processed in memory, and written back to your downloads folder.** The category of document people most need to recognize - contracts, medical letters, statements, identity documents - is the category that most obviously should not be handed to a stranger. Verify by disconnecting from the internet after this page loads.

### Will the recognized text replace my scan?

No, and that is the point. **The page image is untouched and the text sits behind it, invisible.** The signatures are still signatures, the letterhead is still the letterhead. What changes is that search works, text selects and copies, and a screen reader can read the document aloud.

### How accurate is the recognition?

It depends on the scan, and any tool quoting you a single accuracy figure has not seen your file. Four things decide it: the resolution the page was scanned at, the contrast between the ink and the paper, how far the page is rotated or skewed, and how cleanly the original was printed. On a laser-printed page scanned straight at 300 dpi the result is genuinely good. On a fax, a photocopy of a photocopy or a 150 dpi scan it is mixed. **On handwriting it produces nothing useful and the tool says so instead of pretending.** The result screen reports an average confidence across every character read, and above about seventy percent the search layer is dependable.

### Can it read handwriting?

No. **Only machine-printed or typed characters are recognized reliably.** This is a template matching recognizer: it compares each shape on the page against an alphabet of real printed letters, and handwriting has no stable shapes to match. Cursive, hand-printed capitals and signatures all produce nonsense, which the confidence number will make obvious. No free browser based tool reads handwriting reliably, and the ones that claim to are describing something else.

### Which languages does it support?

The Latin alphabet: English and the Western European languages that share the character set. **Cyrillic, Greek, Arabic, Hebrew and the East Asian scripts are outside what it reads.** Characters the standard PDF fonts cannot draw are dropped from the layer rather than substituted, because a wrong character in a search index is worse than a missing one.

### Why does it skip some pages?

Because they already have real text. **A large share of documents people bring to a recognizer are ordinary digital PDFs** that somebody called scanned because their viewer would not let them select anything. Recognizing those would replace a perfect text layer with an approximation of one. The result screen reports how many pages were skipped.

### What resolution should I choose?

300 dpi for almost everything. Use 400 only if the print is genuinely small, and 200 if you are on a phone and the document is long. **Rendering at 400 dpi a page that was scanned at 150 adds time and no information** - the detail either exists in the original or it does not.

### Why is there no big engine to download?

Because the usual approach costs fifteen megabytes or more in engine and language data, paid by everybody who lands on this page from a search result. **This recognizer is our own, sixty kilobytes, loaded on the first file you drop and never before.** The trade is honest: a heavyweight engine reads harder documents. For clean office documents the gap is much smaller than the size difference suggests.

### Should I run OCR before or after redacting a scan?

**After. Always after.** The layer this adds is invisible and fully selectable, so recognizing a page before removing something from it creates a searchable copy of exactly what has to go. [Redact first](https://carthacker.com/pdf-tools/redact-pdf/), verify with copy and paste, then come back here.

### Will the selection line up with the words I can see?

Yes, and it takes real work to get right. **The horizontal scale is computed per word** so that each invisible word is stretched or squeezed until its box is exactly as wide as the visible ink. Without that a long line drifts and by the right margin the selection is a whole word out of step.

### Does it fix a crooked scan?

It measures the skew by fitting a line through the bottoms of the letters, which are the baseline, and corrects it before reading - but **only when the page is genuinely crooked.** Resampling a straight page costs a little sharpness on every stroke and gains nothing, so a page that measures level is read as it is.

### Can I convert the scan into an editable Word document?

Run this first, then [convert to Word](https://carthacker.com/pdf-tools/pdf-to-word/). The converter needs text to convert, and a scan has none until this step. Expect to tidy the result: this adds searchable text over an image, it does not rebuild the original layout as reflowable paragraphs.

### How long does a long document take?

Every page is rasterized at the chosen resolution and read, so it scales with page count. A twenty page document is quick, two hundred pages is a few minutes on a laptop and noticeably longer on a phone. The progress readout names the page it is on so you can see it moving.

## The rest of the PDF suite

All of these run the same way: your file is read by JavaScript on your own machine, the work happens on your processor, and the result is handed back to you as a download that never existed anywhere else.

Organize PDF

- [Merge PDF](https://carthacker.com/pdf-tools/merge-pdf/)
- [Split PDF](https://carthacker.com/pdf-tools/split-pdf/)
- [Remove Pages](https://carthacker.com/pdf-tools/remove-pages/)
- [Extract Pages](https://carthacker.com/pdf-tools/extract-pages/)
- [Organize PDF](https://carthacker.com/pdf-tools/organize-pdf/)
- [Scan to PDF](https://carthacker.com/pdf-tools/scan-to-pdf/)

Convert to PDF

- [JPG to PDF](https://carthacker.com/pdf-tools/jpg-to-pdf/)
- [PNG to PDF](https://carthacker.com/pdf-tools/png-to-pdf/)
- [WebP to PDF](https://carthacker.com/pdf-tools/webp-to-pdf/)
- [AVIF to PDF](https://carthacker.com/pdf-tools/avif-to-pdf/)
- [GIF to PDF](https://carthacker.com/pdf-tools/gif-to-pdf/)
- [BMP to PDF](https://carthacker.com/pdf-tools/bmp-to-pdf/)
- [Word to PDF](https://carthacker.com/pdf-tools/word-to-pdf/)
- [PowerPoint to PDF](https://carthacker.com/pdf-tools/powerpoint-to-pdf/)
- [Excel to PDF](https://carthacker.com/pdf-tools/excel-to-pdf/)
- [HTML to PDF](https://carthacker.com/pdf-tools/html-to-pdf/)

Convert from PDF

- [PDF to JPG](https://carthacker.com/pdf-tools/pdf-to-jpg/)
- [PDF to PNG](https://carthacker.com/pdf-tools/pdf-to-png/)
- [PDF to WebP](https://carthacker.com/pdf-tools/pdf-to-webp/)
- [PDF to AVIF](https://carthacker.com/pdf-tools/pdf-to-avif/)
- [PDF to GIF](https://carthacker.com/pdf-tools/pdf-to-gif/)
- [PDF to BMP](https://carthacker.com/pdf-tools/pdf-to-bmp/)
- [PDF to Word](https://carthacker.com/pdf-tools/pdf-to-word/)
- [PDF to PowerPoint](https://carthacker.com/pdf-tools/pdf-to-powerpoint/)
- [PDF to Excel](https://carthacker.com/pdf-tools/pdf-to-excel/)
- [PDF to HTML](https://carthacker.com/pdf-tools/pdf-to-html/)
- [PDF to PDF/A](https://carthacker.com/pdf-tools/pdf-to-pdfa/)
- [PDF to Markdown](https://carthacker.com/pdf-tools/pdf-to-markdown/)

Edit PDF

- [Rotate PDF](https://carthacker.com/pdf-tools/rotate-pdf/)
- [Add Page Numbers](https://carthacker.com/pdf-tools/add-page-numbers/)
- [Add Watermark](https://carthacker.com/pdf-tools/add-watermark/)
- [Crop PDF](https://carthacker.com/pdf-tools/crop-pdf/)
- [Edit PDF](https://carthacker.com/pdf-tools/edit-pdf/)

PDF Security

- [Unlock PDF](https://carthacker.com/pdf-tools/unlock-pdf/)
- [Protect PDF](https://carthacker.com/pdf-tools/protect-pdf/)
- [Sign PDF](https://carthacker.com/pdf-tools/sign-pdf/)
- [Redact PDF](https://carthacker.com/pdf-tools/redact-pdf/)

Optimize PDF

- [Repair PDF](https://carthacker.com/pdf-tools/repair-pdf/)
- OCR PDF

Free CartHacker tools

## Free speed test and trust badges for Shopify and WooCommerce stores

Find the apps that slow your store down, by name and in milliseconds, then answer shoppers' doubts with trust badges under Add to cart. **Both are free, with no card and no access to your store admin.**

Store Speed Test

### Which app is slowing your store down?

Paste any page and get its three Core Web Vitals, **the apps costing time, like Klaviyo or Judge.me**, and the Shopify or WooCommerce fix.

- 1 free test a day, no sign-up
- 5 a day on a free account
- Google's Lighthouse engine

Trust Seals & Badges

### Do shoppers hesitate at Add to cart?

Show the promises that settle doubts, from free shipping to a money-back guarantee, **right under the Add to cart button**. One line in your theme.

- 8 badges in 2 styles, free
- Premium: 50 badges, 30 languages
- View counts from the first shopper

[Build my trust badges](https://carthacker.com/trust-badges/#builder) Design them without signing up.

Also free

[QR Code Generator](https://carthacker.com/qr-code-generator/)

Change the link after printing and count the scans. **5 tracked codes on a free account**, static codes unlimited.

- Logo and colors
- SVG for print
- Wi-Fi and business cards

---

Source: [OCR PDF Online - Make Scanned PDF Searchable - CartHacker](https://carthacker.com/pdf-tools/ocr-pdf/), CartHacker.

## Free tools from CartHacker

- [Store Speed Test](https://carthacker.com/store-speed-test/): Check your store's Core Web Vitals for free.
- [Core Web Vitals Monitoring](https://carthacker.com/core-web-vitals-monitoring/): Monitor your store's Core Web Vitals every day, for free.
- [Trust Seals & Badges](https://carthacker.com/trust-badges/): Add trust badges under your Add to cart button.
- [QR Code Generator](https://carthacker.com/qr-code-generator/): Make a QR code you can still change after printing.
- [PDF Tools](https://carthacker.com/pdf-tools/): 39 free PDF tools that run in your browser: merge, split, convert to and from Word, Excel and JPG, sign, redact and protect. No file is ever uploaded.
- [CartHacker Premium](https://carthacker.com/premium/): one subscription, monthly or yearly, that upgrades every tool at once:
  - 50 trust badges in 30 languages, switched automatically
  - 180 days of Core Web Vitals, per page
  - All 39 PDF tools, unlimited and never uploaded
  - 200 QR codes, tracked and editable after printing
  - No ads on any guide or tool
  - Every new guide 2 weeks before the public
  - Product Feed Manager and new tools included
- Every quota and price, tool by tool: https://carthacker.com/llms.txt