Following the money to the checkout page.

1. 1

   Choose the doors

   Stripe Checkout and Payment Links, PayPal, Gumroad, Square, Mollie, Razorpay, Paystack, Coinbase Commerce and the newer merchant-of-record players (Lemon Squeezy, Paddle, Polar, Creem, Dodo Payments).

2. 2

   Pull each referral report

   For every checkout domain we read the six-month referral report: which sites send visitors, how many, how that changed. The scraper we used cost three to four cents per report, so the whole pull, retries included, came in well under a dollar. On Lovely Router the same report is $0.06.

3. 3

   Merge into sites

   Subdomains (app.example.com, www.example.com) were folded into one site, giving 9,542 distinct sites.

4. 4

   Remove the noise

   Search engines, social networks, AI assistants and the payment providers themselves are channels, not sellers. We kept them out of the rankings.

5. 5

   Sort the AI products by hand

   Similarweb’s own categories are noisy (one AI video lab shows up under “Romance”), so we labelled the senders ourselves: the biggest by hand, the rest from our own knowledge, web searches and a look at their homepages. 262 of the AI products we found send people to Stripe Checkout; we sorted them into niches.

6. 6

   Rebuild the monthly curve

   Each site’s monthly share of the checkout’s traffic, times that month’s total, gives an estimate of its visits per month, which is how we measured April → September.

7. 7

   Sanity-check and export

   We compared totals across providers, flagged the odd ones (see “what this can’t tell you”), and exported the rankings as spreadsheets.

1 · AI is about half of Stripe’s hosted checkout

2 · One breakout, then a market that is still moving

3 · The top is concentrated, the tail is wide

4 · The challengers are small and growing fast

5 · PayPal is a different market

• It’s modelled data. Traffic figures are estimates from a third-party panel. Trust ranks and ratios more than exact counts.
• Visits aren’t purchases. We can’t see completion, price, refunds or renewals. The calculator’s answers are only as good as your assumptions.
• Some checkouts are blunt. paypal.com includes logins and wallet traffic; gumroad.com includes its marketplace. Stripe Checkout (hosted) is the cleanest.
• Missing referrers. Some sites strip the referrer or embed checkout in a pop-up, so they’re undercounted. Payments through embedded Stripe Elements don’t pass through a hosted page at all.
• Noisy months. A few checkouts moved sharply in single months (Razorpay, Shopify, Coinbase Commerce); we treat those as likely data artifacts and don’t build conclusions on them.
• Our AI labels are judgement calls, made largely by language models reading search results and homepages, and not verified with the companies. Some will be wrong and small unrecognised sites are left out, so treat the AI total as an estimate.

# who sends visitors to a checkout?  ($0.06 per call)
curl https://lovelyrouter.com/v1/tools/site-referrals/run \
  -H "Authorization: Bearer $LOVELY_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input": {"website": "checkout.stripe.com"}}'

# → { "items": [ { "domain": "higgsfield.ai", "visits": 9671010,
#                  "share_percent": 4.93, "global_rank": 866, ... }, ... ] }