A QR code that does not scan is worse than no code at all, because people stand in front of it, try twice and leave. Almost every failure comes from the same short list of causes: too much data crammed into too small a print, poor contrast, a missing quiet border, a surface that bends or reflects, or a link that goes nowhere. None of them are difficult to avoid once you know what a scanner is actually looking for.
You can follow along with the QR Code Generator, which tests each code as it is made, and the QR Code Reader, which shows what a code contains before anything opens.
What a scanner is actually looking for
A QR code is a square grid of dark and light cells called modules. The three large squares in the corners are finder patterns. A camera locates them first, and from their positions it works out the orientation, the size and any tilt or perspective distortion. Larger codes add smaller alignment patterns to correct for curved surfaces, and rows of alternating modules act as timing marks. Everything else is data, protected by an error-correcting code.
Codes come in 40 versions. Version 1 is 21 by 21 modules, and each step adds four, so the grid size is 17 plus four times the version number, reaching 177 by 177 at version 40. The version depends on how much you store and how much error correction you ask for. The single most useful thing to remember is that a bigger version has smaller modules for the same printed width, and small modules are what a phone struggles with.
How much data is too much
The largest code can hold about 7,000 digits, 4,300 letters and digits, or 2,950 bytes of ordinary text at the lowest error correction. Nobody needs that on a poster. The practical limit is legibility. These are real results from the generator at error correction level M, which is a sensible default:
| Content | Length | Version | Grid size |
|---|---|---|---|
| Short link | 24 characters | 2 | 25 × 25 |
| Wi-Fi login | 46 characters | 4 | 33 × 33 |
| Typical page URL | 55 characters | 4 | 33 × 33 |
| URL with tracking parameters | 120 characters | 7 | 45 × 45 |
| Contact card with six fields | 141 characters | 8 | 49 × 49 |
| Paragraph of text | 300 characters | 13 | 69 × 69 |
Now do the arithmetic for a code printed 30 mm wide, counting a four-module border on each side. The short link has 33 modules across, so each is about 0.9 mm. The 300-character version has 77, so each is about 0.4 mm, less than half the size. Both are the same printed square, but the second is far harder to scan from a table or a wall. The fix is rarely a bigger print. It is usually shorter content: remove tracking parameters, use a short path on your own domain, or link to a page that holds the long text.
Choosing error correction
Error correction lets a code survive damage by storing redundant data. There are four levels, and they can restore roughly 7% of the data at L, 15% at M, 25% at Q and 30% at H. Higher levels are not free: they add modules, so the same content needs a bigger version. The table above shows the effect, since the 120-character link is version 6 at level L and version 11 at level H.
- L or M for screens and clean indoor prints, where nothing will scuff the code.
- Q or H for stickers, packaging, outdoor signs and anything that will be handled, scratched or dirtied.
- H when a logo covers the middle of the code, because the logo hides real modules and the redundancy makes up for them.
Colour, contrast and inversion
Scanners look for a difference in brightness, not in colour. A dark navy code on a pale yellow background reads perfectly, while a red code on a green background can have almost no brightness difference despite looking vivid. A useful target is a contrast ratio of at least 4 to 1, and a quick check is to view your design in greyscale: if the code disappears, it will not scan.
Avoid inverted codes, which are light modules on a dark background. Many modern phones read them, but some older apps and dedicated scanners do not, and the failure is silent. If you need a dark design, keep the code itself dark and place it on a light panel. The generator flags low contrast and inverted colours in its print checklist.
The quiet zone and placement
The specification requires an empty border of four modules around the code, called the quiet zone. Camera software uses it to tell where the code ends and the surrounding artwork begins. Text, borders or patterns pushing into that margin are a common reason a code scans on a screen and fails on paper. Modern readers tolerate less than four, but the generator warns below two and four is the safe choice for print.
Placement matters as much as design:
- Put codes on flat, matte surfaces. Glossy laminate reflects the phone's own light, and curved bottles distort the grid.
- Keep them away from folds, seams and staples, which break the pattern.
- Give people a reason to scan with a short line such as “Scan for the menu.” Codes without a label are ignored or feared.
- Size for distance. A common rule of thumb is that the code should be about one tenth of the scanning distance, so a code read from one metre away should be roughly 10 cm wide.
Content that phones understand
A QR code is just text. Phones recognise particular prefixes and offer an action. A web link begins with https://, a phone number with tel:, an email with mailto:, and a contact is a block of text in the vCard format. Wi-Fi uses its own format:
WIFI:T:WPA;S:CafeGuest;P:latte-and-cake-2026;;Here T is the security type, S the network name and P the password, and a trailing pair of semicolons closes the record. If the name or password contains a semicolon, comma, colon, quote or backslash, each of those characters must be preceded by a backslash. A network named Home;Net with the password pa:ss is written like this:
WIFI:T:WPA;S:Home\;Net;P:pa\:ss;;Getting this wrong is a common reason a Wi-Fi code scans fine but fails to join the network. The generator escapes the characters for you.
Static and dynamic codes
A static code contains the final data. It never expires, works offline and cannot be edited or switched off, but the printed pattern is permanent. A dynamic code contains a short web address that belongs to a service, which redirects to your real destination. That lets you change the destination and count scans, at the cost of depending on the service: if a subscription lapses or the company closes, every printed code stops working.
A sound approach is to use a static code that points to a page on your own domain, which you can redirect whenever you like, and to use analytics on that page. You get flexibility without handing your printed materials to a third party.
Adding a logo without breaking the code
A logo in the centre replaces real modules with artwork. It works only because error correction fills the gap, so raise the level to H, keep the logo to about a fifth of the code's width or less, and give it a plain background patch. Never cover the corner squares. Decorative styles such as round dots or gradients follow the same rule: they are fine as long as the scan test still passes. If it fails, go back to square modules.
Testing before you print
- Scan the screen version with the built-in camera on both an iPhone and an Android phone.
- Print a proof at the final size on the real material, not on office paper.
- Scan it from the real distance, in the real lighting, at a slight angle.
- Check the destination by scanning with the reader, which shows the full address. Confirm it uses https and lands on the right page on a phone.
- Keep a record of what each code points to, so you can find and fix it later.
Scanning codes safely
Because a code hides its destination, it has become a favourite tool for scams: stickers placed over legitimate codes on parking meters and menus, or codes in unexpected emails and letters. Reading a code is harmless. The risk begins when you open the link and then enter a password or payment details. Before you open one, look at the full address, check that the domain is the one you expect, and be wary of sites that ask you to sign in or pay after a scan. If a printed code looks like a sticker on top of another, do not use it.
Pre-print checklist
- Is the content as short as it can be, ideally under about 60 characters for a link?
- Is the error correction level suited to where the code will live?
- Is the code dark on a light background, with strong contrast?
- Is there a clear border of four modules and no artwork touching it?
- Is the printed width at least 2 cm, and about one tenth of the scanning distance?
- Did a printed proof scan on two phones from the real distance?
- Does the destination work on mobile and use https?



