Corrections
When a published fact or a wording claim was wrong, it goes here with a date. This list is not a commercial announcement and it is not a ranking change-log.
First-person redemption language
Related pages: Fit checker
Read the original correction, evidence and limitations
First-person redemption languageResult cards no longer describe unverified seed speed bands as timings we recorded. Observation wording is reserved for internal-observation evidence.
Funrize California availability
Related pages: Alternatives to Funrize
Read the original correction, evidence and limitations
Funrize California availabilityFunrize Terms of Use §3.4 limits California, Indiana, Maine and Maryland to tournament-points mode. The finder no longer treats Funrize as a full Sweeps fit in those states.
High 5 redemption minimums and timing
Related pages: Alternatives to High 5 Casino
Read the original correction, evidence and limitations
High 5 redemption minimums and timingHigh 5 Official Rules §8.2 set a $100 cash minimum and a $50 non-cash minimum. §8.7 allows processing of up to ten days and does not guarantee timing. The $50 figure is no longer shown as the unqualified published cash minimum.
Sweepstakes explainer legal wording
Related pages: What is a sweepstakes casino, really?
Read the original correction, evidence and limitations
Sweepstakes explainer legal wordingRemoved the statement that sweepstakes play isn't legally a bet. The page now describes the two-currency structure operators present and does not decide the legal characterisation.
Scenario pages described rules the finder no longer had
Related pages: If you're here for the slots · If you want to keep the spend small · If you want to be able to cash out sooner · If this would be your first one
Read the original correction, evidence and limitations
Scenario pages described rules the finder no longer hadFour saved answer sets explained their results using constraints that had been removed. 'If you're here for the slots' said purchase size was capped at $25; 'If you want to keep the spend small' said a $5 cap excluded any site whose smallest package was above it; 'If you want to be able to cash out sooner' said a $25 redemption-minimum ceiling excluded anything above it and that payout speed was weighted heavily; 'If this would be your first one' described a payout-speed weighting. The purchase-size, redemption-minimum and redemption-speed preferences and their rules were removed on 2026-08-22 and the pages were not updated with them, so each was explaining results with a reason the engine no longer applied. The rules have not been restored. Each page now describes only what the finder actually evaluates, and the two pages whose subject the finder cannot filter on say so and point at the recorded figures in the comparison instead.
'Every casino has a no-purchase route' overstated the record
Related pages: Can I actually play without buying anything?
Read the original correction, evidence and limitations
'Every casino has a no-purchase route' overstated the recordThe free-play answer said every casino in our data had a way in that costs nothing and always a postal route in the terms. Nine of the twelve record a no-purchase route, three of those confirmed against the operator's own published information; for Pulsz, Jackpota and Moozi we have not established it either way, and we hold terms language covering the postal route for seven of the twelve. The page now states those figures and names the three we cannot confirm.
The sweepstakes explainer still said a free route has to exist
Related pages: What is a sweepstakes casino, really?
Read the original correction, evidence and limitations
The sweepstakes explainer still said a free route has to existAfter correcting the free-play answer we left the explainer saying a free way in 'has to exist', that a postal request is 'almost always' in the terms, and that 'a handful of people a year' use it. The first asserts a requirement we are not in a position to state, the second overstates what we hold, and the third is a usage figure we do not have. The page now describes what operators present, states the nine-of-twelve record, and says we have no figures on postal use.
Legal claims on the purchase-or-deposit answer
Related pages: Why do they say purchase instead of deposit?
Read the original correction, evidence and limitations
Legal claims on the purchase-or-deposit answerThe page said the word choice is 'the legal structure in one word', that a purchase 'changes what you're entitled to' and cannot be reclaimed, that the model is why these sites operate where a licensed casino cannot, and that consumer protections do not apply the same way. Those are legal conclusions, published without a source, on a site that declines legal questions. They are removed. The page now describes what operators say about the structure, points at the operator's own refund policy, and sends the protection question to the page that explains why we do not answer it.
Payout-speed figures we do not have
Related pages: If you want nothing to do with crypto · What should I use if I don't want anything to do with crypto? · How does redeeming Sweeps Coins actually work?
Read the original correction, evidence and limitations
Payout-speed figures we do not haveThree pages described payout speed as something our data holds: ruling out crypto-first sites 'removes some of the fastest recorded payout times in our data' and 'the quickest unverified seed payout bands in this dataset', and the redemption walkthrough gave a range 'from under a day to over a week' with an operator's stated window 'usually optimistic'. Payout speed is recorded as null on all twelve profiles and no operator's published window is recorded either, so there were no times, no bands and no range. The comparison has always shown payout speed as not confirmed for every site; the prose now agrees with it.
Two scenario pages pointed at recorded figures that are empty
Related pages: Comparison
Read the original correction, evidence and limitations
Two scenario pages pointed at recorded figures that are emptyCorrecting the purchase-cap and redemption-ceiling claims left both pages telling readers to compare 'the smallest recorded purchase' and 'every recorded minimum' in the side-by-side comparison. Smallest purchase and redemption minimum are null on all twelve profiles. Both pages now say the comparison has a row for the figure and that it currently reads not recorded, which is the true state and more use than a pointer to an empty column.
More payout description the dataset does not support
Related pages: How does redeeming Sweeps Coins actually work?
Read the original correction, evidence and limitations
More payout description the dataset does not supportThe redemption walkthrough said later redemptions at the same site are 'usually much quicker' and that what we hold is 'a speed band per site with an evidence status and a date'. Payout speed is null on all twelve profiles, so there is no band and no date for it, and its evidence status is unverified. The page now says we hold no timing data at all and that the comparison shows it as not confirmed on every result. The free-play answer's claim that generous free routes tend to come with higher redemption minimums went with it: redemption minimums are null on all twelve too.
Operator practice on a state withdrawal, stated as if we knew it
Related pages: Why does a casino suddenly stop accepting players from my state?
Read the original correction, evidence and limitations
Operator practice on a state withdrawal, stated as if we knew itThe state-availability answer said most operators let you redeem an existing balance after they stop accepting new play, and that the window is 'often short and communicated by email only'. We do not record any operator's terms on this. The page now says what happens to a balance you already hold is set by that operator's own terms and points the reader at the site's own notice.
How it works claimed a recorded speed band
Related pages: How the finder works
Read the original correction, evidence and limitations
How it works claimed a recorded speed bandThe how-it-works page said 'what we have is a recorded speed band and a date'. Payout speed is null on all twelve profiles and no operator's published window is recorded, so there is no band and no date. The page now says payout speed is not confirmed for a single site and that every result says so. Two more went with it: identity-check timing, described as a step that does not repeat and the slowest part of a first redemption, is recorded as unknown for all twelve; and free play 'takes weeks', which we cannot know without the redemption minimums or the daily bonus amounts, neither of which we hold.
Responsible play described operator controls we do not record
Related pages: Responsible play
Read the original correction, evidence and limitations
Responsible play described operator controls we do not recordThe responsible-play page said most operators also run their own limits and self-exclusion tools, usually in account settings. We do not record which operators provide which controls, so neither the proportion nor the location was ours to state. The page now says that if the site you use offers those tools they will be on that site, that we do not record which sites offer what, and that the operator's own page is the thing to check. Helplines and the rest of the page are unchanged.
Two pages still described a capability that had been removed
Related pages: Questions we cannot answer
Read the original correction, evidence and limitations
Two pages still described a capability that had been removedThe declined-questions page said 'we record how long redemptions took, in bands, with dates' and that identity checks are treated 'with a timing preference attached'. Payout speed is null on all twelve profiles, identity-check timing is unknown on all twelve, and no timing preference or rule exists — it was removed on 2026-08-22 with the others. Both entries now say what we would record, what we actually hold, and that the finder has no question about identity-check timing.
Unknowns written up as conclusions
Related pages: Can I actually play without buying anything? · If you want to be able to cash out sooner
Read the original correction, evidence and limitations
Unknowns written up as conclusionsHaving said we cannot put a figure on free play, the same pages went on to conclude 'nobody gets there quickly', to promise 'how slow it really is', and to state that operators build free play slow on purpose. A scenario caveat also said the lowest minimums in our data belong to the newest sites, when every redemption minimum is null. Not knowing a number does not license a conclusion drawn from it; all four are removed.
Guesses at why operators do things
Related pages: Question library
Read the original correction, evidence and limitations
Guesses at why operators do thingsSeveral pages explained an operator's reasons as if we knew them: free play is slow 'by design', daily bonus amounts are 'designed to bring you back tomorrow', package pricing is 'built to make the big one look sensible', small packages are worse value 'by design', and the redemption page promised 'where it usually stalls'. Two of those sentences were written while correcting earlier claims in this same pass, which is how an intent claim slips in — it reads like explanation rather than assertion. We can say what a thing is; we cannot say what it is for. All are removed, and statements about this site's own intent are unaffected.
Describing the pace of things we do not time
Related pages: Question library
Read the original correction, evidence and limitations
Describing the pace of things we do not timeEvery timing field in the dataset is empty — payout speed, published windows, identity-check timing, redemption minimums, daily bonus amounts. The pages described pace anyway: free play was 'slow', the redeemable balance 'accumulates slowly', taking money out was a 'slower' process, and the live-dealer answer said table selections are 'usually narrower' and live tables 'generally' cannot use a free balance, when all we record is whether live dealer exists at all. Four rounds of correction removed variations of one sentence and each time left a synonym behind. The pages now describe the mechanism — the redeemable balance is never sold to you, it arrives attached to something else, there is a threshold — and say we cannot tell you how long any of it takes.
Two claims about what people do, with nothing behind them
Related pages: Question library
Read the original correction, evidence and limitations
Two claims about what people do, with nothing behind themAcross the answer library: most people who try free play give up first; the postal route is used by almost nobody; playing the wrong balance is the most common cause of disappointment; most people do not save their sweepstakes balance; debit and bank transfer are the two routes most people already have; first redemptions stall almost always because of a photo or a name; a state withdrawal happens quickly and after somebody official sent a letter. We have no data behind any of them. All are removed rather than softened.
The helpline number was out of date, and we called it free
Related pages: Responsible play
Read the original correction, evidence and limitations
The helpline number was out of date, and we called it freeThe responsible play page gave the National Problem Gambling Helpline as 1-800-522-4700 and told readers that number was another name for 1-800-GAMBLER, available 24/7, confidential and free. The first part stopped being true in September 2025. A New Jersey court ordered the National Council on Problem Gambling to cease using 1-800-GAMBLER as the National Problem Gambling Helpline number after 29 September 2025, and NCPG's helpline number is now 1-800-MY-RESET (1-800-697-3738), with call, text and chat 24/7/365. 1-800-522-4700 has not gone anywhere and still reaches the same helpline. 1-800-GAMBLER also still works, but it is a separate service run by the Council on Compulsive Gambling of New Jersey, so presenting the two as one number was wrong. Free was ours rather than theirs: NCPG says no payment or insurance information is required, and that a carrier's standard text rates can still apply. The page now gives the current number first, keeps 1-800-522-4700 as the alternative it is, lists 1-800-GAMBLER separately with who runs it, and states the cost position in NCPG's own terms. Sources, all read on 26 August 2026: NCPG's helpline page, its 1-800-MY-RESET FAQ, its statement of 22 September 2025 on the court ruling, and CCGNJ's own site. Affects /responsible-play/. Helpline numbers now live in one record with the source behind them, and a test reads both, because this sentence was written correctly and went stale on its own.
Three claims on the responsible play page that went further than the record
Related pages: Responsible play
Read the original correction, evidence and limitations
Three claims on the responsible play page that went further than the recordFound by an independent review of the helpline correction above, on the same page. First, the opening said the numbers listed are confidential and don't ask for payment or insurance details. NCPG says that about its own helpline; the Council on Compulsive Gambling of New Jersey does not say it about theirs, so the sentence claimed for both what one of them publishes. It now says only that the numbers are confidential, which both sources do support. Second, the age section said you have to be 18 or over to use any of the sites here and 21 or over for some of them. That understated it in the direction that matters: ten of the twelve record a minimum age, eight of those set it at 21 or the age of majority where that is higher, one sets it at 18, one says only the age of majority, and for WOW Vegas and Moozi we have recorded no age at all. The page now states those figures and tells you not to assume 18 is enough. Third, it said operators enforce this at the identity check. Every kycTiming field in the dataset reads unknown and nothing records what any operator does at sign-up, so that sentence is removed rather than softened. Affects /responsible-play/.
Five more guesses at what an operator was thinking
Related pages: What is a sweepstakes casino, really? · What's the difference between Gold Coins and Sweeps Coins? · Why do they say purchase instead of deposit? · Can I actually play without buying anything?
Read the original correction, evidence and limitations
Five more guesses at what an operator was thinkingThe correction above this one removed a batch of sentences that explained an operator's reasons as if we knew them, and the test added with it looked for the words those sentences had used — "by design", "designed to", "meant to". An independent review then found the same claim live in five places, in shapes that used none of them. The sweepstakes explainer said operators treat the free attachment as the reason the buy button says purchase rather than deposit; the purchase-or-deposit answer said operators present the purchase model as the reason these sites are structured differently from a licensed casino. Both name a reason and attribute it, and neither reason is anything we can see. The Gold Coins answer said the play-money balance is not generosity but the currency that costs the operator nothing: denying a motive is still claiming to know what it was, and we hold nothing about what any coin costs an operator. The same answer said the model is set up so you cannot buy the redeemable coins directly, which is the purposive construction the earlier test already banned, one synonym out. The free-play answer said the postal route's existence is what makes the rest of the structure work, which is a claim about what the structure is for. All five are replaced by what is actually visible: the button wording, the fact that the redeemable balance is never sold to you, and the fact that the postal route exists. What operators say and publish stays sayable and is still said. Affects /questions/what-is-a-sweepstakes-casino/, /questions/gold-coins-vs-sweeps-coins/, /questions/purchase-or-deposit/ and /questions/can-i-play-without-buying-anything/. The test looks for vocabulary naming a reason, a purpose or a motive in a clause that is not about us, and is proved against every example the class has produced so far, including these five. It is still a lexical detector and not a decision procedure: it catches the shapes it knows, which is why it keeps being attacked and keeps being widened.
Five further unsupported descriptions were corrected
Related pages: Alternatives to Chumba Casino · Why do they say purchase instead of deposit? · Which casino is easiest for a beginner to understand? · If this would be your first one
Read the original correction, evidence and limitations
Five more, found by attacking the test rather than reading the pagesThe correction above added a test for claims about an operator's reasons. It was then attacked twice — once here and once by an independent reviewer — with sentences a person might reasonably write, and it caught almost none of them. Four live claims came out of that. On every alternatives page for a casino with a focused library, five pages in all, the reason to look elsewhere read "its library is recorded as small and easy to browse — deliberate, but limiting if you play a lot": whether a small library is deliberate is the operator's business and we have no way to know it. The word is gone and the sentence now says only what the profile records. On the purchase-or-deposit answer, "people buy a package expecting the balance to behave like a deposit they can withdraw" is a claim about what people expect, and we hold nothing about that; it is now addressed to you, which claims nothing about anybody else. The same answer said these sites are put together differently from a licensed casino — a structural comparison against an undefined thing, left behind when the reason attached to it was removed — and now says only that the purchase wording is part of how operators describe what you are buying. On the beginners scenario, crypto was set to avoid because learning two unfamiliar things at once "is how people give up"; the answer set is unchanged and the reason for it is now stated as ours. And the beginner question said "some people never want the bigger lobby", which is the same kind of claim about what people want, and now says a smaller lobby can be the one you keep. Affects /alternatives/chumba-casino/ and the same page for Global Poker, LuckyLand Slots, Jackpota and Moozi, plus /questions/purchase-or-deposit/, /questions/which-casino-is-easiest-for-a-beginner/ and /for/beginners/. One of the five was invisible because of where it lived: the alternatives rationales, and the drawbacks on every result card, are authored copy that no guard in this repository had ever read, and the corpus now covers both. The other four sat in copy the test already read, in shapes it had no rule for. The last of them was found by the widened rule itself, after this record had been written but before any of it was published, which is why the count in this entry is five and not four.
The footer was still saying what the responsible play page had stopped saying
Related pages: Methodology · How does redeeming Sweeps Coins actually work? · Why does a casino suddenly stop accepting players from my state? · Which sweepstakes casinos have live dealer games? · If you want nothing to do with crypto
Read the original correction, evidence and limitations
The footer was still saying what the responsible play page had stopped sayingAn independent review of the batch above found the age claim we had just corrected still sitting in the site footer, where it renders on every page — including the responsible play page, a few centimetres below the corrected version of itself. It read "you have to be 18 or over to use any of the sites mentioned here, and older in some places", and the record has an age for ten of the twelve; WOW Vegas and Moozi have none at all. The footer said next, in its first repair, that each site sets its own minimum age and states it in its own terms — which is the same kind of claim about all twelve, and two of them record no age at all. It now says only what we hold: a minimum age for ten of the twelve, 21 or over at eight of those, and read the operator's own terms before you sign up. Shared components had never been read by any of the tests in this repository, which is why the same sentence could be corrected on a page and survive in the chrome around it; they are read now. The terms page carries the same claim and is deliberately not changed here: what that page says about who may use an operator is an eligibility statement rather than a description of our data, and it is held rather than corrected. It is recorded in the test as a named exception so it cannot be quietly forgotten. Five more claims went in the same round. The methodology page gave "how long a redemption took" as its example of something we recorded ourselves, and no attribute in the dataset carries that status — every observed redemption speed is empty — so it now says that nothing currently is. The redemption answer said "the waiting is the part people underestimate", which is a claim about what people expect; it now says we hold no figure for how long that step takes. The state-availability answer said nobody outside the company gets advance notice of a change, which we cannot know about anybody but ourselves, and now says only that we do not. The avoiding-crypto answer set said these sites take ordinary payment methods "and always have", that the crypto-first ones are "heavily promoted", and that they therefore "feel more common than they are" — a history, a marketing claim and a claim about readers' impressions, none of them recorded — and now states the count instead. The live-dealer answer explained that live dealer is expensive to run and that most sites therefore skip it, and said the same thing again a section below in terms of what a live table costs per hour and whose maths it does not suit. The first was removed and the second was missed, which the next review found; both are gone now, the counts stay, and the answer says plainly that why the other ten do not carry it is not something we can see from the outside. Affects every page for the footer, plus /methodology/, /questions/how-does-redeeming-sweeps-coins-work/, /questions/why-does-a-casino-stop-accepting-my-state/, /questions/which-casinos-have-live-dealer-games/ and /for/avoiding-crypto/.
Two safety claims, a privacy claim, and six more things we could not measure
Related pages: What's the difference between Gold Coins and Sweeps Coins? · Responsible play · How the finder works · If you're here for the slots · If you'd rather play tables than reels · If you want to keep the spend small · If you want to be able to cash out sooner
Read the original correction, evidence and limitations
Two safety claims, a privacy claim, and six more things we could not measureA fifth independent review read the pages this pass had not been looking at. Two of what it found matter more than the rest. The Gold Coins answer called playing on the play-money balance "risk-free" — an unqualified safety word in gambling copy, on a page that also explains you can buy that balance with money. It now says what is actually true: it teaches you the site without spending anything, if you already hold some. And the responsible play page, after listing five signs worth taking seriously, said "none of those on their own means anything is wrong". That tells somebody who recognises one of them to put it down, which is the opposite of what the section is for. It now says any one of them is worth paying attention to, that several together is worth a phone call, and that you do not have to be certain anything is wrong before making it. The privacy page says your finder answers never leave your browser and no server receives what you picked. The answers go into the address bar so a result can be shared or bookmarked, which means reloading or sharing that link sends them to whoever serves the page. That page is not changed here: a privacy representation is a legal notice, and it is held rather than corrected. The how it works page made a smaller version of the same claim, and the first attempt at fixing it introduced a new error, saying nothing here stores your answers when the finder keeps them in the browser's local storage; the sixth review caught that and both that page and the finder itself now say what actually happens. Six smaller ones, all of them figures we do not hold: the questions list promised to explain "why the first one drags" when identity-check timing is unknown for all twelve; four scenario pages described how long a balance or a budget lasts, when we hold neither the daily amounts nor a single redemption minimum; the low-minimums page advertised "what the recorded redemption thresholds actually are" in its own description, and there are none recorded; two table-game surfaces compared these sites with a conventional casino, which the live-dealer answer on the same site says we cannot do; and the same page said a balance below the threshold "can't do anything at all", when it can still be played and simply cannot be redeemed. Affects /questions/gold-coins-vs-sweeps-coins/, /responsible-play/, /how-it-works/, /questions/, /for/slot-players/, /for/table-game-players/, /for/small-budgets/ and /for/low-redemption-minimums/.
What the finder said about privacy, and five figures we do not have
Related pages: How the finder works · If you want nothing to do with crypto · If you want to keep the spend small · If you want to be able to cash out sooner · If you're here for the slots · Which sweepstakes casinos have live dealer games? · Why does a casino suddenly stop accepting players from my state?
Read the original correction, evidence and limitations
What the finder said about privacy, and five figures we do not haveThe sixth review, and a sweep of the page descriptions this pass had not read. The finder itself told you "your answers stay in this browser — in its local storage and in the address bar above. Nothing is sent to us." The first two clauses are true and the third is not: putting the answers in the address bar is what makes a result shareable, and reloading or sharing that link sends them to whoever serves the page. It now says so. The how it works page had the same problem in the other direction — the repair written for the previous correction said nothing here stores your answers, and the finder stores them in local storage so a refresh does not lose them. Both now describe both halves. The privacy page still carries the original claim, and it is still held rather than corrected. The rest are figures: the crypto answer set promised cards and bank transfers "in both directions" and "on the way in and the way out", when the profile records how a site takes payment and has no field at all for how it pays out — so it now says what it records and tells you to check the cash-out step on the site itself. The small-budget page described "the recorded package sizes" twice and the low-minimum page said it puts "every recorded minimum" in front of you; every smallest-purchase and every redemption-minimum field is empty, and both pages now say that is what the comparison prints. The live-dealer description said live dealer is rarer here than at a normal online casino, which is the comparison the answer beneath it says we cannot make, and now gives the count. The state-availability description said availability changes fast, which we do not time. A scenario caveat said game counts move weekly, which we do not track. Affects /find-a-casino/, /how-it-works/, /for/avoiding-crypto/, /for/small-budgets/, /for/low-redemption-minimums/, /for/slot-players/, /questions/which-casinos-have-live-dealer-games/ and /questions/why-does-a-casino-stop-accepting-my-state/.
Telling you how a site pays out, when the profile has no field for it
Related pages: What should I use if I don't want anything to do with crypto? · If you want to be able to cash out sooner · Methodology
Read the original correction, evidence and limitations
Telling you how a site pays out, when the profile has no field for itThe seventh review found the same defect on five surfaces at once, and it is the most consequential kind this pass has produced: copy about how you get money back out. A profile records how a site takes payment in, and the crypto rule reads only that. There is no payout field anywhere in the dataset. Yet the crypto answer said a site could be "crypto-only in both directions", said the finder catches "the second kind" without saying it catches it on the way in only, described a card-in, crypto-out mismatch and said the profile notes it — it does not — and its finder run was introduced as being for someone avoiding crypto "in either direction". The Stake.us card said "crypto is the only way in and out". All five now say what is recorded, which is the way in, and send you to the site itself for the cash-out step. A test reads every authored surface for a claim about a payout route and fails while no profile has one; it relaxes by itself if that field is ever added. Two more from the same review. The finder's restored-session notice said your answers are "stored in this browser only", which repeats the claim corrected elsewhere in the same component; it now says they are in the browser and in the address bar. And /questions-we-wont-answer/ said a site promising to pay without confirming who you are "is either wrong or isn't planning to pay" — a false choice, and a guess at whether an operator intends to pay. It now says identity checks are a normal part of paying money out and that such a site is worth reading carefully. The low-minimum page's browser title still promised sweepstakes casinos with lower redemption minimums when no minimum is recorded, and the methodology page still explained that game counts move weekly, which we do not track. Both corrected. Affects /find-a-casino/, /questions/what-should-i-use-if-i-dont-want-crypto/, /questions-we-wont-answer/, /for/low-redemption-minimums/, /methodology/, and every result card carrying the Stake.us drawback.
Naming the destinations a redemption can go to, and promising to keep a payment method
Related pages: How does redeeming Sweeps Coins actually work? · What should I use if I don't want anything to do with crypto?
Read the original correction, evidence and limitations
Naming the destinations a redemption can go to, and promising to keep a payment methodThe eighth review, and two of its four findings are the same shape as the one before: telling you something about money that the record does not hold. The redemption answer said a redemption request comes "sometimes with a choice of destination — bank transfer, card, occasionally a gift-card style option", and the diagram beside it told you to choose an available destination and listed destination among the things that vary by operator. No profile records a payout destination for any of the twelve. The answer now says which destinations a site offers is not something we record and sends you to the site's own redemption page; the diagram says to fill in that form and states plainly that we hold no minimum, destination, document list or processing time. And the alternatives pages promised that everything below keeps "the same game types and the same ways to pay". It does not: the starting answers drop crypto from any profile that has other routes, and a site qualifies by matching one of those routes rather than all of them. The page now describes what the code does — including the case the first repair missed, where a profile records crypto as its only route and the starting answers keep it — and a test reads both a mixed-method profile and a crypto-only one so the sentence cannot drift from the behaviour again. The Stake.us card said that without crypto there is "a whole separate thing to learn before you can do anything here", while the same profile records a free route in and daily free coins; it now says before you can buy anything, and mentions the free route. Separately, and not repaired: the review reports that Stake.us's current terms name cards, wallets and bank accounts, which would make the crypto-only value on its profile wrong. That value decides eligibility, and an operator fact needs a source with a date behind it, so it is recorded as an open question rather than changed on the strength of a reading nobody here has done. What did change is the copy that inherits it, which now attributes the claim to our record instead of asserting it about the operator. Affects /questions/how-does-redeeming-sweeps-coins-work/, every alternatives page, /questions/what-should-i-use-if-i-dont-want-crypto/ and every result card carrying the Stake.us drawback.
A date on every attribute, and a list of documents nobody gave us
Related pages: How the finder works · How does redeeming Sweeps Coins actually work?
Read the original correction, evidence and limitations
A date on every attribute, and a list of documents nobody gave usThe ninth review. Four pages said, in four different ways, that every attribute in the dataset carries the date it was last checked: the about page listed twelve profiles with each attribute recorded with a status and a date and said every attribute carries the date it was last checked; how it works said every card carries the date its details were last checked; the alternatives pages said everything there is drawn from profile attributes, with dates. It is not true and it is a claim about how good our evidence is. Of the 212 attributes across the twelve profiles, 70 were read from an operator's own page and carry the date they were read; the other 142 are marked unverified and carry no date at all, and a card falls back to the profile's own review date to show one at all. All four now distinguish the two. The terms page says the same thing and is still held rather than corrected; it is recorded as a held item and pinned in the test so the exception cannot be forgotten. A test now counts the undated attributes from the live data and fails on any promise of a date for all of them; it stops guarding anything the day every attribute really is dated, and says so. Second: the redemption answer told you what to bring to a verification check — photo ID, something showing your address, a utility bill or bank statement from the last three months — on a page that tells you its facts come from the finder's dataset. We record nothing about what any of the twelve asks for. That advice is now replaced by the only honest version of it: read that site's own verification page, because we hold none of this. Third, the alternatives sentence repaired in the correction above was still wrong for one profile: it said the starting point is a casino's payment routes other than crypto, and where crypto is the only route recorded the code keeps crypto. It now says both. Affects /about/, /how-it-works/, every alternatives page, and /questions/how-does-redeeming-sweeps-coins-work/.
Calling an amount small when there is no amount
Related pages: What is a sweepstakes casino, really? · What's the difference between Gold Coins and Sweeps Coins? · Can I actually play without buying anything? · How does redeeming Sweeps Coins actually work? · If this would be your first one · If you want to be able to cash out sooner
Read the original correction, evidence and limitations
Calling an amount small when there is no amountThe tenth review. Three pages said the redeemable balance arrives "in small amounts", and one scenario weighed a generous daily bonus against a modest one. A profile carries a yes-or-no field for whether a daily bonus exists and no amount at all, which the same answers say out loud two sentences later. Small is a claim about how far a free route gets you, which is the thing somebody reading that page is trying to work out. All four now say what is actually observable: the balance arrives attached to something else rather than being sold to you, and where the size of a bonus and the height of a minimum are what matter, we say we hold neither. A test reads the profiles for a bonus amount and fails on any size given to one while none exists. Second, the redemption answer said that where an attribute is recorded it carries an evidence status and a date, and that payout speed "carries neither". Every attribute here carries an evidence status, including the empty ones — unverified is a status — and it is the date and the value that are missing. The correction log repeated it and is corrected too. Getting our own evidence model wrong in public is the same defect as getting an operator wrong. Third, the beginners answer set said the two things that go wrong in week one are about confusion "rather than money". We hold no record of what goes wrong for anybody in a first week, and telling a beginner the problems are not financial ones is not a reassurance we can give. It now says what the answer set is built around and that we hold no such record. Finally, the correction above said an undated attribute means a card can print the evidence date as unknown; it cannot, because a card falls back to the profile's own review date. Affects /questions/what-is-a-sweepstakes-casino/, /questions/gold-coins-vs-sweeps-coins/, /questions/can-i-play-without-buying-anything/, /questions/how-does-redeeming-sweeps-coins-work/, /for/beginners/ and /for/low-redemption-minimums/.
Grading a free route the finder does not grade
Related pages: If you're here for the slots · If you want to keep the spend small · Can I actually play without buying anything?
Read the original correction, evidence and limitations
Grading a free route the finder does not gradeThree pages described the quality of a free route as though something here measured it. The slots answer set said a free route is weighted as important "so a site with a thin one loses ground"; the small-budget page paired a small budget with a "strong" free route; the free-play answer said a generous free route and a low redemption threshold are separate things and a site can be good at one and poor at the other. Two rules read free play and both read a yes-or-no field: whether there is a free way in at all, and whether there is a daily top-up. A thin free route and a lavish one score identically, because nothing in the engine can tell them apart, and no profile records a bonus amount, a frequency or a threshold that would let it. All three now say what the rules do. This is the same shape as the retired-constraint corrections above — copy describing a distinction the finder cannot make. The check that catches those looks for the name of a rule we no longer have, so a sentence that invents a capability without naming one goes straight past it. A new test reads the two free-play fields out of the live rule set, fails if either stops being a yes-or-no, and rejects a free route given a grade. Affects /for/slot-players/, /for/small-budgets/ and /questions/can-i-play-without-buying-anything/.
A count of games, on a site that will not stand behind one
Related pages: If you're here for the slots · Which casino is easiest for a beginner to understand?
Read the original correction, evidence and limitations
A count of games, on a site that will not stand behind oneA High 5 drawback rendered "Thousands of games and not much help narrowing them down" on every result card that returned it. We record library breadth as a band — focused, wide, very large — and no profile records a catalogue size for any of the twelve; /for/slot-players/ says so in its own caveat, that a number here would be one we could not stand behind, two paragraphs after using one. The drawback now says the library is recorded as very large. Two more counts went with it, both figures of speech rather than claims about a named site: a thousand games you will never open on the slots page, four thousand titles on the beginners answer. A page that will not stand behind a number does not get to print one as a turn of phrase either, and no check that reads for a phrase can tell the rhetorical thousand from the asserted one. A new test reads the profiles for a catalogue size, confirms library breadth is still a band, and rejects a game count anywhere in the published copy. One instance is not repaired: the finder's own help under the library-size question still reads "Thousands of games is a lot to wade through". It is recorded here rather than quietly left, and a test asserts it is still there, so it fails the moment somebody repairs it. Affects every result card carrying the High 5 drawback, /for/slot-players/ and /questions/which-casino-is-easiest-for-a-beginner/.
The rules were saying things the rules had not read
Related pages: Methodology
Read the original correction, evidence and limitations
The rules were saying things the rules had not readThe sentences on a result card are generated from the rule set, and until now nothing here checked them. The checks read the pages we write by hand; a rule's sentence is written by hand too, and it is printed on every card and in full on /methodology/. Four of them claimed more than the single attribute behind them established. "It has no daily free top-up, so free play runs out and stays out" reads a yes-or-no field about a daily bonus and tells you your free play is over — on cards where the rule immediately beside it, reading a different field, said you could play there without buying anything. That one now says only that it has no daily free top-up — an independent review caught the first repair generalising a daily field into any schedule at all, and it was cut back to the field. "You can keep playing there without buying anything" promised an indefinite supply from a field that records only that a free way in exists; it now says you can play there without buying anything. "It's weaker on that device than the alternatives" compared the site with the other results, which that rule never sees; it now says it falls short of what you wanted. And "Crypto is the only way to pay there" asserted an operator fact from a value carrying no source at all — the same claim the drawback and the no-crypto answer were corrected for in the ninth review, left standing in the one place that was frozen at the time. It now says its profile records crypto as the only payment route, which is what we can actually see. The recorded value itself is unchanged and remains an open question: no source in our record names Stake.us's payment media, and the schema has no way to say we do not know — the field cannot be empty or absent, and two must-have rules read it directly. It is not corrected here. The rule set's sentences are now checked like the rest of the copy, and a further test rejects one that claims more than the field behind it records. No rule id, weight, evaluator, preference or recorded value changed. Affects every result card and /methodology/.
Two must-have questions, and no records to answer either of them
Related pages: If you want nothing to do with crypto · What should I use if I don't want anything to do with crypto? · Methodology · How the finder works · If this would be your first one · If you're here for the slots · If you want to be able to cash out sooner
Read the original correction, evidence and limitations
Two must-have questions, and no records to answer either of themThe finder asked two questions it had no records to answer. A profile has a slot for the routes a site takes money in by, and two must-have rules read it: one for the payment methods you said you would use, one for whether a site can be used without crypto. Every one of the twelve profiles carried a value in that slot and not one of them carried a source. No row in the source table stated a payment route for any operator. So the answers to both questions were being decided by something somebody had typed, in both directions at once — Chumba was ruled out of a PayPal search on the strength of a list nobody had read off a page, and Stake.us was ruled out of every card-and-bank search, handed to anyone who picked crypto, and given a drawback saying crypto was its only route, on the strength of one word. A review named the Stake.us instance; a sweep of the other eleven found the same hole under all of them. Three things have changed. The slot can now record that we do not know, which is not the same as recording that a site takes nothing, and all twelve now say so; the validator refuses a value there without an operator-published source behind it. Both rules that read it are withdrawn, and the payment question is no longer asked — a must-have the records cannot answer is not caution, it is a question we had no business putting to you, and leaving it in would have emptied the list for everybody rather than telling anyone anything about a casino. The neighbouring slot for whether a site takes crypto at all was filled the same way, and an independent review caught it doing the opposite damage: a Stake.us card reading 'It takes crypto, which you said you would prefer' directly above a drawback saying we hold no record of how it takes payment in. That slot is empty now too, the rule that read it is withdrawn, and with both of its rules gone the crypto question is no longer asked either. And the copy that counted those routes is repaired: /for/avoiding-crypto/ and /questions/what-should-i-use-if-i-dont-want-crypto/ said eleven of the twelve took a card or a bank transfer and one did not, which we did not know; both now say what we hold, which is nothing, and point you at the site's own purchase screen. The withdrawal is not permanent and it cannot be lifted by another guess. All of it returns by itself the day every profile in the finder carries both payment records — the routes and the crypto flag — each taken from the operator's own published page, with the source and the date it was read, and each matching what that page is recorded as saying. They come back together or not at all, because they are one fact written twice. A test holds the line at every part of that: not an unverified seed, not a page somebody else wrote about the operator, not a source that says something other than what the record says, not one field without the other, and not one sourced profile among the rest, which would put the must-have back and then exclude every profile nobody had read yet. Two smaller things ride along. The finder's own help under the library-size question still read "Thousands of games is a lot to wade through", the one game count the previous pass recorded as held rather than repaired; it is repaired now, and the allowance that held it is gone, so that line is read by the same guard as everything else. And the rule set version moves to 2026.08.26, because an evaluator changed. Affects the finder itself, /methodology/, /how-it-works/, /for/avoiding-crypto/, /for/beginners/, /for/slot-players/, /for/low-redemption-minimums/, /questions/what-should-i-use-if-i-dont-want-crypto/, every alternatives page and every result card carrying the Stake.us drawback.
The finder put your answers in a link, and now it does not
Related pages: How the finder works · Privacy
Read the original correction, evidence and limitations
The finder put your answers in a link, and now it does notTwo passes ago the finder began reading a run back out of the address bar, so that a link could carry somebody’s answers into a fresh browser. It had been writing them there for longer than that. A query string is part of a request, so every reload and every shared link handed what somebody had asked for to whoever serves the page — which is why the copy under the results was changed to say so, and why /how-it-works/ said the same thing about your state. The wording was accurate and the behaviour was the problem. The finder no longer writes answers to the address bar and no longer reads them from it, and a run is now kept in this browser’s local storage. Links from the older build still exist, so arriving on one now has the answers taken out of the address bar without being read: what that link already sent cannot be recalled, but the reload after it is clean. The two buttons that used to open the finder preloaded with a set of answers — at the foot of every scenario page and every alternatives page — now open the plain finder, and say so. A copied recommendation still spells the answers out in words, which is what makes it readable without following anything; the link at the end of it used to reproduce the run and now points at the finder. The sentences describing the old behaviour are gone from the finder and from /how-it-works/. /privacy/ is unchanged: it has always said your answers never leave your browser, and that is now true of the software rather than a claim held against it.
A site we no longer offer as a fit, offered as somewhere to go instead
Related pages: Alternatives to Stake.us · Alternatives to Zula Casino · Alternatives to LuckyLand Slots · Alternatives to Moozi
Read the original correction, evidence and limitations
A site we no longer offer as a fit, offered as somewhere to go insteadEvery alternatives page ran its results against every profile on file rather than against the set the finder itself will return. Two of the twelve do not clear that gate: LuckyLand Slots, whose official play URL serves a not-available page and whose lifecycle is recorded as winding down, and Moozi, whose excluded-state list we have not read and which we therefore do not geo-fit. Both were already held out of the finder, and both could still come back as a recommended alternative on the one surface where somebody is explicitly asking where to go instead. The gate is a single function and the page was not calling it; it filtered the record set directly, which is a second gate with a looser rule, and a rule written twice is a rule that will disagree with itself. The pages now run against the same population the finder returns, and the improvement axes offered at the top of each page are judged against that same population, so an axis cannot offer a filter the results below it will not apply. Both profiles keep their pages, their recorded drawbacks and their lifecycle: staying documented is not the same as being offered as a fit, and a page about leaving a site has to be able to name it. Where the subject is one of the two, the page now says it is not a site we currently offer as a fit, rather than saying it was left out of its own results — which implied it would otherwise have been on the list. No rule, weight, evaluator, preference or recorded value changed. Results did move, and that is the point: two pages were showing one of these two in their top three, and each has been replaced by a site that was already eligible and was being kept out by a profile that was not. /alternatives/high-5-casino/ showed LuckyLand Slots third and now shows Sportzino; /alternatives/stake-us/ showed Moozi third and now shows WOW Vegas. The ruled-out table underneath moved on ten pages: nine of the other ten, and /alternatives/stake-us/ a second time, under a different question. On each of them a row naming LuckyLand Slots or Moozi and the must-have it failed has gone, along with the count in the heading above it. /alternatives/zula-casino/ is the only page where nothing changed but the size of the pool. Affects every alternatives page, /alternatives/high-5-casino/, /alternatives/stake-us/, /alternatives/luckyland-slots/, /alternatives/moozi/ and /.
The methodology page said nobody had read the pages it lists
Related pages: Methodology · Sources
Read the original correction, evidence and limitations
The methodology page said nobody had read the pages it lists/methodology/ told readers that attempts to read operator help and terms pages had been refused, and that rather than cite pages nobody read, the values were marked unverified. Two sentences above it, the same page printed the evidence statuses actually present in the records, one of which is operator-published. One section below it, the same page listed sixteen transcribed operator documents, with the section numbers they came from and the date each was read. Both of those lists are generated from the records. Only the sentence between them was written by hand, and it understated what we hold — a false claim about our own evidence, which is the same defect as a false claim about an operator and not a safer one for pointing downwards. The standing notice under every set of results made the same blanket statement about the casino details being unconfirmed. Neither is a blanket matter. Eleven recorded fields carry an operator document on at least one profile: the minimum age, the states an operator says it will not take, the free entry route, whether coins can be bought, what the operator calls the product, and its VPN and multiple-account rules among them. Eleven carry none on any profile, including every field describing what a site is like to use, and those are the fields most of the scoring reads. The page now prints both lists, generated from the records, counting each field against the profiles that record it rather than against all twelve, and noting which fields the published rules read. The notice draws the line where it actually falls, which is per card rather than per site: coverage is uneven. A card can carry six operator-fact lines; WOW Vegas carries none of the six, and Pulsz, Stake.us and Jackpota carry no free entry route. So the notice now makes its claim line by line rather than site by site: what a card prints was read from that operator's document, and where it prints none of those lines we hold none of them. The excluded-state list is stated on its own, because we hold one for every site the finder can return, it is what a state answer is checked against, and no card prints it — an earlier draft of this repair put it inside the card sentence, which turned the sentence into a denial of the one eligibility fact we are surest of. The page also corrects itself on how the rules use an evidence status: only the state check reads one, and every other rule takes the recorded value whatever its status, so an unverified seed can earn the same credit a sourced fact would. Nothing became confirmed: no evidence status changed, no source was added, no value moved, and the pages we could not read are still named as pages we could not read. Affects /methodology/, /sources/, / and every set of results.
Unverified fields were still earning fit credit
Related pages: Methodology
Read the original correction, evidence and limitations
Unverified fields were still earning fit creditThe evidence status on a weighted field reached the card as a caveat but did not reach the arithmetic behind the result. A provisional interface classification, library band, game range, free-play flag or device list could therefore earn the same full or partial credit as a supported value. Weighted fields now enter the fit calculation only when operator-published evidence carries both its source id and retrieval date, or an internal observation carries its observation date. Anything else is shown as unconfirmed and counts neither for nor against the result. No recorded value or evidence status was promoted as part of this correction. The separate payment and crypto evidence gate, the must-have rules and the eligible operator population are unchanged. Affects the finder, saved scenarios, alternatives and /methodology/.
A must-have could be cleared by a value nobody had checked
Related pages: Methodology · How the finder works · If this would be your first one · If you're here for the slots · If you'd rather play tables than reels · If you want something that isn't exhausting to use · Which casino is easiest for a beginner to understand? · Which sweepstakes casinos have live dealer games? · What is a sweepstakes casino, really?
Read the original correction, evidence and limitations
A must-have could be cleared by a value nobody had checkedThe 2026-09-07 pass stopped an unverified field earning fit credit. It did not touch the must-haves, and two of them read a field that is an unconfirmed seed on every profile the finder can return: the game list. So a card could say "It has live dealer tables, which you said were a must" and "It has slots, which you asked for" on the strength of a value nobody had read off an operator page, and eight of the ten sites could be ruled out of a live-dealer search the same way. The payment gate does not catch this, because it withdraws a rule only when no profile records the attribute at all, and every profile records a game list. The scoring boundary does not catch it either, because it applies to weighted rules. Three things have changed and none of them is the outcome. The sentence a rule in this position produces now names the kind of value it used, on both the clearing side and the ruling-out side, rather than asserting what the operator has. /methodology/ prints the rules in that position and counts the profiles, generated from the records, so sourcing the field empties the section rather than leaving a paragraph to be corrected; /how-it-works/ says the same in one line. And the card no longer appends "and clears every must-have you set" as a flat clause: it says how many of them were cleared against something we have confirmed, and that clause now disappears entirely when no must-have was set, which it did not before. No rule, weight, evaluator threshold, recorded value or evidence status changed, and the same casinos come back in the same order. Whether these two rules should keep running at all is a product decision and is not ours to take: withdrawing them would remove the game-type and live-dealer questions from the finder, which is a larger change than this correction. The side-by-side comparison went with it: five of its nine rows — games, library, interface, apps and product class — printed a seed in the same ink and the same words as a value read from a document, two rows below the redemption rows that say "unverified seed" in as many words. Every cell now carries its evidence from the field rather than from the look of the string, and an unconfirmed one is marked in the ink, in a word, and in a footnote. Product class is marked on every row, because it is our reading of what a site is and no operator publishes it. The ordering sentence is repaired too: a site for which nothing could be weighed has an empty denominator and therefore a full score, which ties it with a site that met everything, so a run where every preference came back unconfirmed was in alphabetical order under a sentence claiming it was ordered by weighted fit. The sentence now says which of the two it was. Affects the finder, every saved scenario, every alternatives page, /methodology/, /how-it-works/, /for/beginners/, /for/slot-players/, /for/table-game-players/, /for/simple-interfaces/, /questions/which-casino-is-easiest-for-a-beginner/, /questions/which-casinos-have-live-dealer-games/ and /questions/what-is-a-sweepstakes-casino/. The rule set version moves to 2026.09.21, because the sentence a rule produces is part of the rule.
Clarifying the record: an explanation update and a later finder change
Related pages: How the finder works · If you'd rather play tables than reels · Which sweepstakes casinos have live dealer games?
Read the original correction, evidence and limitations
A disclosure-only repair and a behaviour change, logged as oneThe 2026-09-21 entry above records a disclosure-only repair, and it is accurate about what that repair did. It also says, in as many words, that whether the two game-list must-haves should keep running at all was a product decision and not ours to take. Twenty-two minutes later, in a second commit under the same review, they stopped running: a population gate now withdraws a hard rule reading the game list until every profile the finder can return carries an evidence-backed one, and no profile does. That is a behaviour change, it is not what the first repair reviewed, and the log did not say so — which left the earlier entry reading as though the review had approved it. It did not, and it is not being rewritten to say otherwise. This entry is the record of the second change. What the withdrawal actually did: the game-type-available and live-dealer-required rules no longer run, the live-dealer question is no longer asked because no live rule reads it, and the game-type question is still asked because the weighted coverage rule still reads it — but that rule scores an unconfirmed list as unknown, so the answer is recorded, shown back, and changes nothing. No recorded value, evidence status, verification date, weight, evaluator threshold or casino population moved, and the same sites come back in the same order for every game-type answer. What was left behind was the copy. The finder's own help under the game-type question still said a casino has to offer at least one of the games you pick to show up, which is the exact promise the withdrawal exists to stop making; it now says what we record and what it does not do. The live-dealer alternatives axis still said the preference does rule sites out; /how-it-works/ still attributed every withdrawn rule to how an operator takes money in, and still said nobody is asked about live dealer who did not mention table games, when nobody is asked about live dealer at all; the live-dealer answer told you to set it in the finder and watch every site without it be excluded; and the table-games scenario said its answer excluded sites outright. The live-dealer answer also printed a finder shortlist beneath itself, captioned as a run for someone who will not sign up anywhere without live dealer tables, on a page that says the finder cannot act on that; nothing in that list was filtered for live dealer, so the shortlist is gone and the page is the answer. All of them now describe the engine that exists, and the two /how-it-works/ sentences are generated from the live rule set so that the day the evidence arrives they change with it rather than being corrected again. A test reads the published records, fails if a game list becomes evidence-backed while this copy still says it cannot count, and pins the results, exclusions and ordering against the behaviour that is deployed, so a wording repair cannot quietly move a result. Whether the game-type question should stay at all, given that it asks for something nothing can currently act on, is still the owner's decision and is still open. Affects the finder, /how-it-works/, /for/table-game-players/, /questions/which-casinos-have-live-dealer-games/ and every alternatives page offering a live-dealer axis.