Dev.to WebDev 🛠 Dev 👁 0 📖 11 min read

How Many Websites Use Visa? 1,022,781 Sites and How to Check a Card Mark Yourself

As of November 16, 2026, webatla counts 1,022,781 websites that show Visa, 0.31% of the 325.8 million websites it tracks. The live figure sits on webatla's Visa technology page, which is rebuilt from a fresh scan every d

How Many Websites Use Visa? 1,022,781 Sites and How to Check a Card Mark Yourself

As of November 16, 2026, webatla counts 1,022,781 websites that show Visa, 0.31% of the 325.8 million websites it tracks. The live figure sits on webatla's Visa technology page, which is rebuilt from a fresh scan every day.

For a developer the more useful question is what each of those rows is made of, and how far it holds up when you crawl pages, build features from detections or keep a store front end in shape.

Waffle chart of 400 squares standing for all 325.8 million scanned websites, with about one and a quarter squares filled green for the 0.31% that show the Visa mark

Visa's share of all websites in webatla's November 16, 2026 scan. Each square is 0.25% of the 325.8 million sites.

What the scanner matches when it reports Visa

Visa is a card network. Nobody installs it on a server, so "use" needs a definition. webatla's Visa page names four kinds of evidence, and each has to name the network. They are a logo asset, a script from a payment gateway, a payment button and a call into a payment SDK. In page terms that is usually an image or inline SVG icon, a script tag, a button element, or JavaScript that hands the network's name to a payment library.

The last form is easy to overlook. Wallet integrations list the card networks they accept as plain strings in the page's JavaScript, for example supportedNetworks in an Apple Pay payment request or allowedCardNetworks in a Google Pay configuration, where Visa appears as visa or VISA. A page can name Visa there without drawing a single card icon.

Every one of these forms is something the page says. The card payment itself runs between the gateway and the banks, outside anything a crawler downloads, so a row can only ever mean that a page names Visa as a way to pay. An icon left in a template keeps saying it after the merchant stops taking the card.

As a feature the mark is sparse. About 3 websites in every 1,000 carry it, so in a crawl of the general web the column is nearly all zeros, and the few ones are informative. The Visa page itself calls the mark a strong sign that a site takes money from visitors, which makes it worth keeping as an input whenever a pipeline has to separate storefronts from everything else.

Reproducing a detection on one domain

You can get a rough version of the signal with curl. Treat it as one quick check you can run yourself. It is not how webatla detects anything.

curl -sL -o /dev/null -w '%{http_code}\n' --max-time 20 https://example.com/
curl -sL --max-time 20 https://example.com/ \
  | grep -oiE '[a-z0-9_./-]{0,40}visa[a-z0-9_./-]{0,40}' \
  | sort | uniq -c | sort -rn | head -n 20

The first command prints the final HTTP status after redirects. The second counts the strings around every match.

  • If the status is anything other than 200, ignore the grep. Many stores answer scripts with a 403, a 429 or a challenge page, and an empty result then only means the page was not served. Because -L follows redirects, a store that sends you to a regional version is checked on that version.
  • Asset paths such as /icons/visa.svg, ids and class names that contain the word, SVG titles and wallet configuration strings are the matches worth trusting.
  • The bare word in body text is weak evidence. Travel, immigration and consular sites use "visa" in an unrelated sense, and the pattern also catches words that contain the four letters, such as advisable or improvisation.
  • Empty output with a 200 does not mean no Visa. curl returns the HTML the server sends and runs no JavaScript. Icons a script inserts after load exist only in the rendered DOM. Open the page in a browser, open DevTools and search the Elements panel with Ctrl+F (Cmd+F on a Mac), which finds text, CSS selectors or XPath in the live DOM.
  • Some stores show card icons in the footer of every page. Others show them only in the cart or at the final payment step, which a crawl of public pages may never reach. Your curl misses those too.

Shopify's Dawn theme shows how a store can produce the icons. When its footer setting for payment icons is on, the footer loops over shop.enabled_payment_types in Liquid and renders each entry with the payment_type_svg_tag filter. On such a store the icons are inline SVG built from the payment methods the merchant has switched on, so they arrive in the served HTML where the grep above can see them. Keep that loop in mind for the next section.

Visa and Mastercard as one feature, American Express apart

If you turn technology detections into features for a classifier, the Visa and Mastercard marks are close to duplicates. Mastercard shows on 1,022,520 sites, Visa on 1,022,781, a difference of 261. The gap stays that small in every slice the two technology pages publish. Country here is the server's, by IP, and most payment-mark sites resolve to Canada, most likely because of where one large store platform hosts them (more below).

Diverging lollipop chart of Visa sites minus Mastercard sites: +261 overall, +193 on Canadian servers, +40 on US servers, +150 on .com, +77 on .shop, +53 on .store, and Mastercard ahead by 22 on .de, 28 on .uk and 92 on .nl

Visa count minus Mastercard count, overall and by server country and domain extension, November 16, 2026. No slice shown differs by more than 261 sites.

Matching totals do not prove that the same domains carry both marks, since no published table crosses the two. They are what you would expect if both marks come from one source on the page, such as the icon loop above or one gateway script that names both networks. In a feature matrix the Visa and Mastercard columns are therefore likely to be almost identical. Check the row-level overlap in your own data, for example the share of domains with both marks or the correlation between the two columns, and if it holds, merge them. Two copies of one column make an additive score count the same evidence twice and split feature importance between two names for one thing.

American Express shows on 944,310 sites, so 78,471 or more Visa sites (7.7%) lack it. That figure is a floor from subtraction alone, because every site showing American Express without Visa would push the true number higher. The totals allow Mastercard to be a near copy of Visa, but they force a gap of at least 78,471 sites for American Express. Of the three card columns, it is the one that can carry a signal of its own.

The same logic shapes any Visa market share figure. Inside a peer group of the three card marks, Visa holds 34.2% of 2,989,611 detections, Mastercard the same 34.2% and American Express 31.6%. On presence the networks are level. Which one moves more money cannot be read from page markup.

Correlated jumps usually mean one change, not many

History tables invite trend stories. Before writing one, check whether the neighbors moved in the same month. webatla's history model has a clean example.

Slope chart of existing sites that added each technology, January to February 2025: Google Pay 9.06K to 37.3K, Visa 10.7K to 33.6K, American Express 10.2K to 31.5K, Apple Pay 11.9K to 26.1K, while Shopify fell from 18.8K to 15.6K

Existing sites that added each technology in January and February 2025, from webatla's monthly history model.

In February 2025 the model has 33,600 existing sites adding the Visa mark, against 10,700 in January. So the jump came from sites that already existed, and it hit five card and wallet marks in the same month, while the number of sites moving onto Shopify did not rise (it fell from 18,800 to 15,600). A scanner cannot tell you which change that was. It can tell you that a step shared by five payment marks in one month is most likely one event, and your analysis should treat it as one.

In a pipeline, a simple guard is to compare each series' month-over-month change with the median change of its peer group, and to treat a step as news only when one series moves and its peers do not.

The same rule covers November 2025. Recorded departures from Visa fell from 17,800 in October to 3,300 that month, and the other two card marks, Apple Pay and Shopify stepped down with it, so it may be a measurement effect and recent flows are best compared across technologies over the same window.

Server country follows the hosting platform

Start from one domain and look at who owns the address it resolves to.

ip=$(dig +short example.com A | tail -n 1)
echo "$ip"
whois "$ip" | grep -iE '^(orgname|org-name|netname|country):'

On a hosted store the address belongs to the platform, so any lookup by IP, whether a whois record or a geolocation database, describes the platform rather than the merchant. Behind a CDN or a reverse proxy you get the proxy operator instead. Either way, IP country measures the edge that answers the request.

webatla's country column is that kind of measure, the server's country taken from its IP address. That is why 951,086 Visa sites, 93.9% of those with a known country, resolve to Canada. The probable reason is that most of them live on one hosted store platform, and the Shopify technology page puts 94.5% of Shopify stores on Canadian servers. These figures describe infrastructure. They say nothing about where merchants or their buyers are.

For market questions the domain extension, which the merchant picks, is the better hint. .ca holds 16,270 Visa sites, .de 26,761, .au 33,491 and .uk 41,289.

Bounding overlaps when all you have is totals

Visa usage statistics, like most public technology counts, come as one total per technology plus country and domain tables, with no co-occurrence. You can still get hard limits. For two technologies A and B, the sites running both inside one country cannot outnumber the smaller of the two counts there. Add that up over countries and you have a ceiling on the overlap.

For Visa the arithmetic is short because the mark is so concentrated. Any Visa site that also runs some stack X either sits on a Canadian server, where X has a published count, or is one of the 71,695 Visa sites hosted elsewhere or with no known country. So the overlap is at most X's Canadian count plus 71,695.

Stack Websites Hosted in Canada Visa overlap, at most
React 11,053,868 31,974 103,669
WooCommerce 3,344,511 20,994 92,689
Next.js 1,973,502 7,191 78,886
Stripe 386,361 5,669 77,364
Joomla 497,752 4,196 75,891

Lollipop chart of the largest possible overlap with Visa as a share of Visa sites: React 10.1%, WooCommerce 9.1%, Next.js 7.7%, Stripe 7.6%, Joomla 7.4%, with 7.0% of Visa sites hosted outside Canada or of unknown country

Ceilings on the overlap between Visa and five stacks as a share of Visa sites, from the per-country counts on each technology page.

The same reasoning covers all five at once. At most 141,719 Visa sites run any of them, so at least 881,062 Visa sites, 86.1%, run none. If you write WooCommerce plugins, Stripe integrations or React storefronts, the million sites with a Visa mark are mostly not where your code runs. Turned around, at most about 20% of Stripe sites and at most 2.8% of WooCommerce stores can show the Visa mark at all.

A ceiling is not an estimate. The true overlap may be far smaller. The bound holds because each site has a single server country in every table, and it tells you nothing about where inside the range the real value sits.

How fast a snapshot of payment marks decays

webatla's history model rebuilds monthly counts of active sites from archived copies of page code. Two change figures on the Visa page come from it, +22% across the last 12 months and +52% across the three years from October 2023 to October 2026. In that last year 245,000 sites came in and 61,900 left. Early years are approximate, and model months are counted differently from the daily file, so quote the change figures and do not compare the last model month with today's count.

Net growth hides a lot of turnover. The calendar year shows it better than the last 12 months, since most of 2025 comes before the November step in recorded departures.

Split bar of Visa flows in 2025: 361K arrived (210K new sites launched with it, 151K existing sites that added it) and 254K left (148K shut down, 107K dropped the mark), net change +107K

Sites that arrived and left the Visa count in calendar year 2025, by kind, from webatla's history model.

In 2025 the model recorded 361,000 arrivals and 254,000 departures, about seven out for every ten in. Of the departures, 148,000 sites shut down and 107,000 removed the mark while staying online. In 2021 the same table shows 235,000 in and 93,700 out, about four out for every ten. Shut-downs grew 4.3 times between those two years while the year-end count grew 1.8 times, from 497,000 to 874,000. The 2025 departures equal about a third of the 767,000 sites counted at the start of that year.

  • Cached detections go stale quickly. Re-check any stored snapshot before you rely on it, and expect a good share of shut-down sites in an old one.
  • Compare snapshots row by row. Two totals can match while thousands of domains have swapped in and out between them. If your own crawler keeps per-domain results, a join between two runs shows which sites added or dropped the mark, and a total hides that.

For the full tables by month, country and domain extension, see webatla's Visa statistics article.

Notes for anyone who maintains a payment page

The Visa page makes a point maintainers should take literally. A lot of the card art on the web was copied into themes years ago and is maintained by nobody, and every crawler reads that art as a current claim.

  • Render the icon row from configuration. The loop over enabled payment types is a sound pattern. Turning a network off in settings then removes its mark, and nobody has to remember a hard-coded image.
  • Check the rendered page after a gateway change. If an old icon survives, every scanner keeps counting your site for that network.

Giving each icon a text alternative, an alt attribute or an SVG title element, helps screen reader users and makes the mark readable by machines as well.

FAQ: the Visa count in short

How many websites use Visa?

1,022,781 as of November 16, 2026, or 0.31% of all 325.8 million scanned websites. The daily figure is on the Visa technology page.

Does a Visa icon prove that a site accepts Visa cards?

No. The icon is a claim the page makes. Authorization happens between the gateway, the card network and the issuing bank, and no crawler sees it.

Should I keep Visa and Mastercard as separate features?

Probably not. Their totals agree within 261 sites in every published slice, so check the row-level overlap in your data and merge the two if it is as high as the totals suggest. Keep American Express apart.

Why does curl find no Visa icon on a store that shows one?

Either the page was not served to curl (check the status code first), or a script adds the icons after load, or they appear only in the cart or at the payment step. Search the live DOM in DevTools instead.

Sources

📰 Read the original article on Dev.to WebDev

Originally published by Dev.to WebDev. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.