Every entry in The Vetted Index carries a date and a source. That discipline exists because, over twenty sessions of verification, this desk has retracted its own claims often enough to learn the shapes those errors take. There are nineteen of them. Each one is set out below with a first-party example, most of them errors we published ourselves and then withdrew.
The question sounds simple. Does this brand accept cryptocurrency?
It is not simple, and the reason is worth stating plainly before any of the patterns below make sense: a keyword search for “bitcoin” on a company’s website is not a measurement of whether that company accepts bitcoin. It is a measurement of how often a string appears in a document. Those are different questions, and the gap between them is where every error in this article lives.
The nineteen patterns fall into five families. The first family is about what a crypto reference means. The second is about a page outliving the fact it describes. The third is about a real source that says less than we said it did. The fourth is about a publication contradicting itself – and every example in it is ours.
Family A – The reference is real, and it is not a payment method
These are the cheapest errors to make and the easiest to catch, once you know to look. In each case the brand’s own site genuinely contains the word. It simply is not doing the work the reader assumes.
1. Crypto as a JavaScript API
A raw-markup keyword sweep of Compass returned one hit for “crypto”; Prestige Imports returned four. Read in context, all five were window.crypto.getRandomValues() – the browser’s Web Crypto API, sitting inside minified click-stream code.
This matters more than it first appears. A raw-markup sweep scores non-zero on essentially any modern website for this reason alone. On the sweep that found it, Prestige Imports out-scored several brands with genuine, documented acceptance. Any methodology that ranks candidates by keyword frequency will surface this pattern first and confidently.
2. Crypto as authentication
BlockBar‘s only Ethereum reference anywhere on its site was the phrase “Sign in with Ethereum Wallet”, offered alongside “Sign in with Google” and “Sign in with Apple”. That is a login credential. The same site’s FAQ answers the actual payment question by naming wire transfer.
A wallet used to prove who you are is not a wallet used to pay. The two appear within a few hundred characters of each other on a great many luxury e-commerce sites.
3. Crypto as an investment product
This one scores highest of all on a keyword sweep and is completely false. Galaxy Digital describes itself as an institutional digital-assets business; every one of its digital-asset references describes what Galaxy sells. Sygnum Bank, FINMA-regulated, carried exactly three BTC references, all of them one product: a fund offering directional exposure to BTC.
We had badged both as accepting BTC and ETH. The badge asserted you could pay these institutions in bitcoin. What they sell is bitcoin. The ticker is the product, not the payment rail. Both were removed.
4. Crypto as a price-display unit
The most recent addition, and the most elegant. Sotheby’s International Realty‘s site returned two clean instances of “bitcoin” – too few to verify, too many to dismiss. Read in context across 530,585 rendered characters, every instance sat inside a currency-display picker: BTC listed beside the Bahraini Dinar and the Chilean Unit of Account, carrying an exchange rate of 0.0000158962.
The site will show you a house priced in bitcoin. It will not take bitcoin for the house. A unit for displaying prices is not a way to pay.
Family B – The page is real, and the company has moved on
Family A is about misreading a word. Family B is different in kind: the page says exactly what you think it says, and it is no longer true.
5. The marketing page versus the checkout
Global Charter published two pages that contradicted each other. Its crypto-payments marketing page carried 34 references and an explicit roster of fourteen currencies – BTC, ETH, USDT, USDC, BNB, SOL, XRP, ADA, DOGE, AVAX, LTC, BCH, SHIB and DASH. Its actual payment page read: “CRYPTO PAYMENTS TEMPORARILY UNAVAILABLE.”
The marketing page is what a search engine indexes and what a journalist quotes. The checkout is what settles the transaction. When they disagree, the checkout is correct.
6. The orphaned page on a retired brand
PrivateFly’s bitcoin-payments page returns HTTP 200. It is live, first-party and detailed: it names BTC, BCH, ETH and four USD stablecoins, explains the BitPay mechanics, and dates the programme to January 2014. By every test this desk owns, it is gold-standard evidence.
The first line of that same page reads: “PrivateFly is now FXAIR.” The operating successor, FXAIR, renders 463,805 bytes with zero references to cryptocurrency, bitcoin, BitPay or crypto.
The page is not lying. It is orphaned. And it carried its own dating evidence, which is the reusable part: it still offered BUSD and PAX, neither of which has existed under those names for years, and it rendered the literal string {{telephone}} – an unsubstituted template placeholder. A page nobody has loaded is a page nobody maintains. Internal-consistency tells can date a document when no date is published on it.
7. The repurposed domain
A domain that once belonged to a luxury brand and now serves something else entirely will still satisfy a link check, a status-code check and often a brand-name check. The name in the WHOIS record and the business behind it are separate facts.
8. The brand that moved domains, read as a brand that died
This is the mirror image of pattern 7, and it is the error we sustained longest. Our own entry for the yacht builder Wally recorded the business as gone: “wally.it serves nothing.” That claim stood for twenty-six days across nine sessions.
Fetched over plain HTTP rather than HTTPS, wally.it returns 200 with a 94-byte empty placeholder on a corporate domain-holding range. The live business is at wally.com – discovered from parent company Ferretti Group’s own brand links, then loaded and read before being substituted. Wally trades normally: full sail and motor fleet, brokerage, contacts, 163,285 characters of first-party estate.
Its crypto acceptance remains unverified, because that estate contains no crypto reference. But a dead domain is evidence about a domain, not about a company – and we had recorded a live business as defunct on the strength of one failed request.
Family C – The source is real, and it says less than we said
This is the largest family, the hardest to catch, and the one that does most damage – because in every case there is a citation, and it does look checked.
9. The dated-and-scoped source
A source can be genuine, first-party and still fail on two axes that nothing about its appearance reveals: when it was published, and what it covers. A 2021 press release describing a pilot in one market is routinely cited as evidence of a standing global policy. The correct test is not “does a source exist” but “does this source, on its own stated date and within its own stated scope, name this brand doing this thing.”
10. The citation that does not name the brand
Our Hublot coin badges rested on a third-party blog post. Read in full, that post never names Hublot at all. The citation existed, resolved, and was about something else. A link in a footnote is not a reading of the thing it links to.
11. The real event, misdescribed
Having found that the Hublot badges rested on an absence, we read the primary source – Hublot’s own November 2018 release – and found evidence that is present and says something materially different. Three errors sat in one sentence under our own “Last Verified” stamp:
- Hublot’s release describes a limited edition purchasable only online. We had asserted five physical boutiques.
- It names OSL Limited of Hong Kong as the partner, quoted from its chief executive. We had named Bitcoin Suisse – flatly wrong.
- It describes an edition limited to 210 pieces. We had described a standing policy.
Nothing here was fabricated. A real event was reported with the wrong partner, the wrong channel and the wrong permanence.
12. The plan reported as a completed rollout
Our Ferrari guide stated, in three separate places under a dated stamp, that Ferrari’s BitPay programme “extended to European dealers by July 2024” and to its broader international network by the end of that year.
BitPay‘s own announcement, dated 17 October 2023 and still live, is scoped verbatim to “Ferrari dealerships within the United States”, names ten participating US dealers, and contains no mention of Europe, of July 2024, or of any international rollout. Reuters reported only that Ferrari intended to extend to Europe.
Nobody invented a company, a product or a programme – all three are real. What was invented is that the plan completed, and the month it completed in. A stated intention rendered in the past tense is one of the hardest errors to see, because every proper noun in the sentence checks out.
13. The real product, with a collateral class we invented
VP Bank‘s Lombard loan is entirely real. Its own page is titled “Flexible liquidity without selling securities” and describes the collateral as “your existing assets”. It contains zero crypto references.
We had asserted BTC and ETH accepted as collateral, with a settlement window and a verification date. Not a fabricated institution, not a fabricated product – a real product with an invented collateral class, and a timestamp that made the invention look checked.
14. The settlement route taken as acceptance
Several of the most recognisable names on any “brands that accept bitcoin” list do not accept bitcoin. What exists is a settlement path: the buyer converts crypto to fiat through an OTC desk, a broker or a title-and-escrow agent, and the vendor is paid in currency, frequently without ever knowing the source of funds.
This is a materially different transaction from the one the reader expects, and it carries a taxable disposal the reader may not have planned for. Four entries in our Index sit in this category and are now marked as settlement routes rather than acceptors, in their own field: Burj Al Arab, Rolls-Royce Motor Cars, Bombardier and Propy.
The distinction is worth defending precisely. When we later found that Ferrari had a genuine processor integration at dealer level, there was pressure to file it here too – it would have been tidy. We refused: a processor integration where the customer actually pays in BTC is closer to direct acceptance than to an OTC desk, and stretching a category to make a number fall is how a taxonomy stops meaning anything.
15. The brand-versus-reseller conflation
An authorised reseller accepting crypto is not the manufacturer accepting crypto. Bugatti was listed on our own Index on the strength of iLusso, an independent dealer settling through BitPay. The dealer’s acceptance is real. It is not the marque’s, and a buyer approaching the marque directly will be told so.
Family D – The contradiction is inside the publication
Every example in this family is ours. They are included because a verification method that only audits other people’s claims is not a verification method.
16. The badge contradicting the prose
Our entry for AMINA Bank described it, correctly, as “an infrastructure counterparty, not a retail crypto-accepting merchant.” It carried BTC and ETH badges.
The bank’s own “available cryptocurrencies” page settles it: a table with the columns Asset, Custody, Tradable, Staking and Loan collateral. That is a list of what the bank holds for clients, not a list of currencies you may pay the bank in. A sweep of all entries for badge-versus-prose contradictions found exactly one. This one.
17. Prose contradicting prose, inside a single entry
Two entries opened with an endorsement and only then withdrew it: “European exotic car marketplace with full cryptocurrency support. Unverified – …”
Both halves were written by us and both were live. A reader skimming a card meets the endorsement first and may never reach the caveat. Where a caveat exists, it leads. Both entries were rewritten so that it does.
18. The enumerated payment list that excludes crypto
The last pattern is the one piece of good news here, because it is the strongest positive evidence available for a negative claim.
Chronext, rendered rather than raw-fetched, returned zero crypto references across its homepage, FAQ and 26,805-character terms and conditions. But the T&C does not merely omit crypto – it enumerates what is accepted: bank transfer, credit card, Apple Pay, Google Pay, financing and Klarna.
A closed enumeration is a negative statement in positive form, and it is far stronger evidence than absence. Absence can mean the page is thin, or badly rendered, or that the answer lives somewhere you did not look. A published list of every accepted method that does not contain crypto is the company telling you the answer.
A caution attached to the same finding: a raw fetch of that site reported “eth” five to eight times and would have supported the claim we were retracting. Those hits were whether, together and method. Word-boundary matching returned zero.
Family E – The entity is not the entity
The four families above all assume the page in front of you belongs to the company you think you are reading about. The nineteenth pattern is the one where that assumption fails, and it is the pattern this desk found in its own work.
19. The name collision
Two live, unrelated companies share a name. The article describes one of them. The link points at the other.
This publication described Bezel, the authenticated pre-owned watch marketplace, and linked it to bezel.com. As of a check made on 20 August 2026, bezel.com returns HTTP 200 and its own page title reads “Coin Bezels and other fine Coin Jewelry”. It is a coin-jewellery business. The watch marketplace trades at getbezel.com.
The error survived every check this desk had. A liveness sweep saw HTTP 200 and passed it. A keyword sweep found the word coin and would have read it as corroboration rather than contradiction. A stale-stamp audit had nothing to say, because the stamp was not the problem. The claim was caveated, dated and sourced – and pointed at the wrong company.
Every other pattern in this taxonomy requires a reader to go and check something. This one a reader falsifies in a single click, by following our own link and finding a jewellery shop under a paragraph about a watch marketplace. That is what makes it the most damaging of the nineteen, and it is why it is written up here rather than quietly corrected.
What it changes in practice: a URL is a claim about identity, and it has to be verified as one. Reading the destination’s own title against the entity we say we are describing is a two-second check that none of our instruments were doing. It is now part of the standard.
A caution on the general case. Name collisions cluster in exactly the places this publication works: short, common English nouns adopted as brand names, where the .com almost always belongs to an older business and the newer company trades on a prefixed or suffixed domain. Bezel, in a corpus about paying for watches with coins, is the case stated at its most embarrassing.
What falls out of this
Four rules govern how this desk verifies, and all four are consequences of the patterns above rather than principles adopted in advance.
- Read the source the claim cites – and check its date, its scope, and that it names the brand. Patterns 9, 10 and 12 are all failures of this one rule.
- An empty or thin fetch is not evidence. Render the page. One homepage in our records returned 543 characters to a plain fetch and several hundred thousand when rendered.
- Distinguish a failure to reach a server from a fact about a business. A 521 or a TLS failure is real signal; a 403, 429 or 308 tells you nothing. Try plain HTTP before concluding – it is what found Wally.
- Never substitute a near-miss domain without loading it, and discover candidates from the company’s own links rather than from a search engine’s guess.
And one rule about counting, which this article had to apply to itself. A keyword hit count is not evidence. Read the context and use word boundaries. Pattern 1 and pattern 18 are the same lesson from opposite directions.
A note on the number nineteen
These patterns were numbered as they were discovered, across twenty sessions. In preparing this article the numbering was audited and found to be wrong: eighteen distinct patterns were being carried under a running count of sixteen. One number had been assigned to two different patterns, and one pattern had entered the list without ever being given a number.
The published figure was eighteen at first publication because eighteen is what was there. It is nineteen as of 20 August 2026: Family E was added after this desk linked a watch marketplace to a coin-jewellery company of the same name and had to write up its own failure. The number will keep moving, and each move will be dated here.
The original note, left standing:
The published figure is eighteen because eighteen is what is there. We record the correction here rather than quietly renumbering, for the same reason every other correction on this site is dated and left visible: a taxonomy of counting errors that will not audit its own count is not worth publishing.
Why this is published at all
Most of the examples above are our own errors. Publishing them is a deliberate editorial choice, and the reasoning is straightforward.
A verified index is only worth what its method is worth. Any publication can assert that its listings are checked. What distinguishes a checked listing from an asserted one is whether the publication can describe how it checks, show what it caught, and demonstrate that it retracts. The retractions are the evidence.
Last Verified: August 2026. Every example in this article is drawn from this publication’s own dated verification record. Source URLs, character counts and publication dates were re-read at source on the dates recorded in each entry of The Vetted Index.
Further Reading
- The Vetted Index – every brand we track, with its status, its sources and the date each was last checked.
- The Acquisition Desk – a dated, sourced written answer on whether and how a specific brand can actually be paid in crypto.
- Hardware Wallets Compared: Ledger and Trezor – custody before settlement, and why the order matters.
- Collaborative Custody – holding keys without holding all of them.
- The Crypto Luxury Tax Guide – why a settlement route and an acceptance route are taxed differently.
Affiliate disclosure: some links on this site, including to hardware wallet manufacturers, are affiliate links. We may earn a commission on purchases made through them. No brand appears in The Vetted Index as a result of any commercial arrangement, and no brand has ever paid for placement, verification or a change of status. The buyer pays this desk; the vendor never does.
Frequently Asked Questions
Does a mention of bitcoin on a company’s website mean it accepts bitcoin?
No. A keyword match measures how often a string appears in a document, not whether a company accepts a payment method. The word routinely appears as a browser JavaScript API, as a sign-in credential, as an investment product the company sells, or as a currency-display unit in a price picker. Each of those is a documented pattern with a first-party example in this article.
What is the strongest evidence that a brand does not accept cryptocurrency?
A closed enumeration in the company’s own terms and conditions. Where a brand publishes a complete list of accepted payment methods and crypto is not on it, that is a negative statement in positive form and is materially stronger than the mere absence of any crypto reference, which can also be caused by a thin page or a failed render.
Is a settlement route the same as accepting cryptocurrency?
No. In a settlement route the buyer converts crypto to fiat through an OTC desk, broker or escrow agent and the vendor is paid in currency. It is a materially different transaction from direct acceptance, it usually creates a taxable disposal, and the vendor may never know the source of funds. Entries of this kind are marked as settlement routes in The Vetted Index rather than as acceptors.
Does a dead domain mean a brand has ceased trading?
No. A failed request is evidence about a domain, not about a company. One brand in our records was listed as defunct for twenty-six days because its former domain served an empty placeholder; the business was trading normally at a different address, discoverable from its parent group’s own brand links. Try plain HTTP as well as HTTPS, and look for the company before concluding.
Why does this publication list its own errors?
Because a verified index is worth what its method is worth. Any publication can assert that its listings are checked. What distinguishes a checked listing from an asserted one is whether the publisher can describe the method, show what it caught, and demonstrate that it retracts when the evidence changes.
