How we evaluate sweepstakes casinos

Why the method is different here

A licensed casino review can lean on a regulator: licence status, enforcement history, approved games and published complaint outcomes are all public in most states. A sweepstakes review has none of that. What it has is the operator's own documents, the state's laws, and whatever the operator's support team will put in writing.

That changes what a good review looks like. It cannot be a score out of five based on vibes. It has to be a record of what was read, when, and what it said, with the gaps marked as gaps. The purpose of this page is to make that method auditable, so a reader can decide whether our conclusions are worth anything.

The consequence is that some well-known brands have no page here. If we cannot read the terms, we do not publish a review and we do not replace the gap with a summary from another site. The absence is the method working as intended, not an oversight.

The five criteria

Every operator page uses the same framework. The weights are fixed before we look at an operator, so a brand cannot be rewarded by a criterion we invented for it.

State accuracy

25%

Does the operator publish a specific excluded-state list, is it current, and is it enforced at sign-up and at redemption? We compare the published list with the 2026 bans and with the operator's own sign-up behaviour. A vague list scores worse than a short one, because a short list is at least a decision.

Redemption transparency

25%

Does the operator state the minimum, the maximum per request, the processing window and the verification requirements before a player deposits? A brand that leaves these to support is a brand whose payout behaviour cannot be predicted, which is the single most important thing a player needs to know.

Bonus clarity

20%

Are play-through, game contribution, maximum redemption and expiry published in the same place as the offer? We value a small, clean grant above a large, conditional one, and we say so on the page rather than repeating the headline number.

Product and payments

15%

Which games and providers are named, whether the lobby is usable on a phone, which payment rails exist in which states, and whether the operator imposes a matching rule it does not explain. Named third-party providers are a mild positive; in-house originals without published RTP are a mild negative.

Support and responsiveness

15%

We ask a specific pre-sale question with a date attached and record the answer. Template replies to specific questions score poorly, because the same behaviour is what a redemption dispute will meet.

The source hierarchy

Every factual claim on this site is ranked by the strength of its source. When two sources disagree, the higher one wins, and the lower one is noted if the disagreement matters.

What an audit records

An audit records what a brand's published terms say on a date: the excluded states, the play-through base and contribution, the redemption minimum, the window, the verification requirement, and the terms version. Nothing in that list requires playing, and nothing in it is an opinion.

The record is deliberately narrow. It cannot tell you whether a brand pays quickly, because we have not observed enough redemptions to say. It can tell you whether the brand has described its process, which is the only thing a reader can check before spending.

The vocabulary in that record (play-through, verification, state list, terms version) is defined in the glossary, so a reader can check the definition without leaving the method.

What we refuse to score

Game libraries, support quality, "trust scores" and payout speed are all things a site can invent and a reader cannot verify. We leave them out rather than dress them up as data. The absence is visible: pages say "not summarised" instead of filling the gap with a number.

When a term has changed

If a term has changed, send the clause and the date to editorial@casinotrustcenter.com. We will check it against the source, update the page and move the date. Corrections are published as changes to the page, not as a separate note, because a reader checks the page rather than the changelog.

What we test, and what we do not claim

Where it is lawful for us to do so, we open an account, complete verification, make a small purchase and request a small redemption. We record the request and receipt dates. That is the only way to observe a payout rather than repeat a claim about one.

We do not test every state, because we cannot be physically present in all of them, and a location check is part of the product. We do not claim to have tested a market we have not tested. When a page says a redemption window is published but not observed, that is exactly what it means, the operator's number, not ours.

We also do not claim to know an operator's internal queue, its approval rate or its dispute outcomes. Those are not observable from outside, and any site that publishes them is either guessing or repeating the operator's marketing.

Evidence standards

Every claim that can change carries a date. Every terms quote carries a version. Where a page depends on a screenshot or a document we cannot publish, we describe what it showed and when we saw it, rather than asserting the conclusion without the trail.

This is deliberately pedantic. In a market where a brand can change its state list in a week, an undated claim is worse than no claim, because it looks like current knowledge and is not. A dated claim at least tells the reader how stale it might be.

When an operator is excluded

Some operators do not get a page at all. The reasons are consistent:

• No published excluded-state list, or a list that contradicts the sign-up flow.
• No redemption section in the terms, or a redemption section that states nothing.
• Bonus conditions that cannot be found before deposit.
• Claims that a redemption is guaranteed, or language that treats it as a regulated payout.
• Terms that cannot be retrieved and read in a form we can cite.

Exclusion is not an accusation of fraud. It is a statement that we cannot verify enough to write something useful. The operator may be perfectly reputable; we simply cannot show it.

Corrections

When we get something wrong we correct the page and note the change date. We do not delete a review because a commercial partner asked us to, and we do not keep a claim alive because it ranks well. If a correction changes a conclusion, the page's date changes with it.

If you find an error, tell us at editorial@casinotrustcenter.com. Corrections that change a factual claim are published with a date; corrections of typos are not announced, because nobody needs a changelog for a comma.

AI, automation and the limits of both

We use software to monitor pages, diff terms and flag changes. We do not use it to invent findings, authors or test results. Every page that makes a claim about an operator has been reviewed by a human against the source, and the date on the page reflects that review.

Automation is good at noticing that a page changed and bad at judging whether the change matters. The judgement is the editorial work, and it is the part we do not delegate.

Independence

We earn commission through affiliate links, and we disclose it wherever links appear. Commission does not determine placement, and operators cannot pay for a page, a conclusion or inclusion in a list. Our affiliate disclosure sets out the arrangement, and our editorial policy sets out the separation between commercial and editorial work.

What we deliberately do not do

• No pages generated for every city, ZIP code or keyword permutation.
• No recommendations for operators that exclude the reader's state.
• No invented authors, credentials or testing claims.
• No “last updated” stamps without a substantive change behind them.
• No payout rankings based on claims we have not observed.

How a state page is built

A state page is assembled from four layers, and knowing which layer a sentence comes from is how a reader can tell a fact from a judgement. The first layer is the jurisdiction record: the regulator, whether a casino or sports licence exists, the tier and the one-line summary. It is a data file, it is edited in one place, and it drives the badge, the table and the comparison pages.

The second layer is the longform block: five sections per state covering the regulator, the market history, three player profiles, the tax and records position, and a short version. That prose is written per state, dates its check and does not reuse a template beyond the section headings.

The third layer is the brand table, which is computed rather than written. It reads each operator's published state list and renders what the operator says, with one exception described below. The fourth layer is the FAQ, which is generated from the same records and is the only part of the page that is structurally identical across states.

The advantage of the split is that a change is local. When an operator removes a state, the brand table updates and the prose does not need rewriting; when a state changes its law, the jurisdiction record and the longform block change together. A reader who sees a contradiction on the page can usually work out which layer is stale, which is also how we find our own mistakes.

How the closed tier is decided

The closed tier is not a list of states where the law is settled against the product. It is a list of states where we would not describe the product as available, and three different things can put a state there: a statute that reaches the product, an enforcement action or regulator position that reaches it, or the operators themselves deciding not to serve it.

The third cause is the one readers question most. A state can be closed in our model because every audited operator excludes it, without any statute naming the product. That is not a claim about the law; it is a claim about the market, and the page says which of the three causes applies.

The exception works in the other direction. When an operator's published list includes a state that has closed the product by statute, the statute wins. The brand table shows the state as unavailable and explains why, because a list that still includes a closed state is out of date and a reader who relies on it will discover the problem at the cashier.

Both rules exist because a reader needs one answer rather than two. A page that said the operator lists the state but the statute closed it, without resolving the contradiction, would be accurate and useless, so the page resolves it and names the reason.

What a depth number means, and what it does not

We measure the words on a page, and the measurement is a floor rather than a goal. A page that clears the floor with repetition is worse than a shorter page that does not, which is why the number is published alongside the checks rather than used as an argument on its own.

The measurement exists because this market is full of thin pages. A state page with three hundred words and a table tells a reader nothing about the regulator, the history or the alternatives, and it exists to rank rather than to inform. A floor is a crude defence against that, and it is a defence we state openly.

The number does not measure whether the facts are right, whether the page is readable or whether the judgement is sound. Those are checked by reading, and the checks that matter are on the link, structure and citation side rather than the word count.

A reader who wants to test a page rather than trust it should pick a claim, find the source and see whether the page describes it correctly. That takes ten minutes and is worth more than any methodology statement, including this one.

How long a check takes, and why pages go stale

A state page takes between one and three hours to check: the regulator's site, the licensing lists, the statutory position and the operator state lists. A brand page takes longer, because it requires reading the terms line by line and checking what each clause says about redemptions.

Pages go stale for three reasons. A regulator changes a position without a press release, an operator changes a state list without a version number, or a statute changes and the summary keeps saying what it said before. The first two are the common ones, and they are why the date on each page is the most important element on it.

The correction route is deliberately simple. A reader who finds a stale claim can send the page, the claim and the source, and the correction is made and dated. We do not require an account or an explanation of why the reader was looking.

The honest expectation is that some pages are stale at any given moment. A site of ninety-three routes checked by hand is never entirely current, and a reader who treats a dated claim as a claim about that date is using the page correctly.

How operator claims are handled

An operator's own page is the first source for what the operator offers, and it is never the last. A brand can publish a state list, a bonus and a redemption window, and each of those can be out of date, so the page records what the operator says and says that it is the operator saying it.

Where the operator's claim and an external record disagree, the external record wins and the page says so. A brand that lists a state which has closed the product is the clearest example, and a brand whose published redemption window contradicts its own terms is the second.

We do not accept an operator's marketing as evidence of anything except its own intention. A claim that a product is licensed requires a licence number and a regulator; a claim that a payout is instant requires an observation; a claim that a bonus is the largest requires a comparison we cannot make with two audited brands.

That position costs us the easy page. It also means a reader who finds a claim here can see where it came from, which is the only reason a reference page is worth reading at all.