What we ran
- Sites: the 100 online stores of anti bot software market share. All established, none a risk. We wanted the false-flag rate on good merchants.
- Tool: the engine behind the merchant onboarding website check. 13 rows (reach, HTTPS, domain age, same page for bots and ads, bot protection, shop signs, policy pages, policy text, contact, company number, payment brands, payment host, title) and a verdict: normal, thin, red.
- Where: our test box, one home IP, plain fetch, 8.4 requests per site. The certificate log answered for none, so domain age is skipped everywhere.
What we found
| verdict | all 100 | home page answered (54) | refused or failed (46) |
|---|---|---|---|
| normal | 3 | 3 | 0 |
| thin | 80 | 35 | 45 |
| red | 17 | 16 | 1 |
46 stores gave the check no page: 35 returned 403, a few 429, 418, 451 or 498, one 511. The content rows are skipped for those, so the verdict is “thin”: unreadable.
Of the 54 that answered, the row results:
| row | pass | warn | fail |
|---|---|---|---|
| same page for bots and ads | 24 | 20 | 10 |
| policy pages linked | 19 | 10 | 25 |
| contact details in the text | 15 | 9 | 30 |
| company name and number | 4 | 50 | 0 |
Why the 16 red stores are red: Googlebot or an ad click got a different page (10), no contact details in the HTML (10), policy links missing (7). The reasons overlap. Big retailers prerender for crawlers and print footers with scripts, so the check sees what a plain client sees.
The 3 normal stores were John Lewis, Gymshark and Casper.
How you’d use it
- Acquirer or PayFac: run known-good merchants through the check first, see which rows fire on them, and set your threshold from that. The tool is for the long tail of new, small merchants.
- Onboarding team: treat red as a queue for a person, thin as “load the page in a browser”, normal as no extra step. See KYB website verification.
- Merchant monitoring provider: the rows that fire on legitimate sites are the ones to tune or to replace with a browser run. See for acquirers and PayFacs.