Colophon

About this reference

This site answers one question for each state: what does the law actually say about gold and silver, and which document says it. It exists because that question is usually answered with a number nobody can take apart.

What this is

A state by state record of four things: whether a bullion purchase carries state sales tax, whether gold and silver are recognized as legal tender, how the state treats a gain on a sale, and whether a state bullion depository exists. Every finding names the document it rests on. Every published state carries the date a person last read those documents, printed at the top of its page.

The reason to prefer a page here over a summary elsewhere is that a page here can be disagreed with productively. The definitions are stated, the arithmetic is shown, the denominator is printed next to every ratio, and the states that would move a figure under a different definition are named. If a reader concludes the count should be different, this site has given them everything they need to say why.

The counterpart to that is restraint. Where nobody has read the statute, the record says so and stops rather than filling the gap from another summary. An empty cell is a fact about the research and it is published as one.

The sourcing standard

These are commitments rather than aspirations, and they are written so that a reader can catch the site breaking one.

Fetched, not guessed: No URL is cited here that has not been retrieved and read. A citation is a claim that a specific document at a specific address says a specific thing, and a plausible looking address for a document that ought to exist is not evidence that it does.
A reader proxy is not a fetch: Content served through a caching reader or summarizing proxy does not satisfy the rule above. Such a service can return a stale copy, a partial render, or its own paraphrase, and it can return something for an address that no longer serves anything. Where the only access to a document was through one, the record says so and the citation is downgraded rather than presented as an official record.
Amendatory bills are read as images: A bill that amends existing law is printed with the removed words struck through and the added words underlined. Extracting text from that PDF silently discards the strikethrough, so an automated read of an amendatory bill returns the old law and the new law run together as though both were enacted. Bills of that kind are read as rendered pages instead, and where a quotation depends on which words were struck, that is stated.
Unreachable hosts are recorded, not routed around: Several official state hosts refused connections during research. Where that happened, the state record says which host was unreachable and what was cited in its place, rather than quietly substituting a secondary source and presenting the result as though the official one had been read. A named gap is usable. A hidden one is not.
Disagreement is published, not resolved: Where two sources this site relies on say different things, the state page prints the divergence under its own heading and leaves it standing. Picking a winner and deleting the loser would remove exactly the information a reader needs to judge the finding.

Sources are also ranked, and the ranking is visible on every citation. The text of a statute, a legislature's own record of a bill, and a state tax authority speaking about its own tax are official records. Everything else is marked secondary, with a hollow marker and the word printed beside it, at the same ink and the same size as every other kind. Secondary means not an official record. It is not a verdict on the text, and it is not a warning: some quotations here carry that marker for no reason other than that the state's own host would not answer when it was asked. The methodology page defines each kind and the rules for using it.

What is not sold here

Nothing. This site sells no product and no service, runs no advertising, accepts no sponsorship, takes no affiliate commission, and carries no commercial outbound link. It does not link to a dealer, a depository operator, a broker, or a marketplace, and it does not rank them.

That is not a statement of virtue. It is the only thing this site has. Every page that competes for these questions belongs to somebody who profits when the answer is yes, and a reference that also profits when the answer is yes is worth nothing to a reader trying to check a claim. The absence of a commercial interest here is the reason a figure on this site can be cited, so it is enforced by the build rather than by good intentions: the verification script fails the build outright if a commercial outbound link appears anywhere in the output.

Nothing here is legal or tax advice. It is a record of documents, and a statute can change the day after a page is checked.

Corrections

A claim on this site is worth exactly as much as the document under it. If the document says something different from what the page says, the page is wrong, and no amount of careful phrasing here changes that. The citation is not decoration; it is the appeal route.

What happens when a claim is found to be wrong, in order:

The cited document is fetched again and read, under the same rules set out above.
The state's record is corrected at the source, in the data the pages are generated from, so every page and every figure that depends on it changes together. Figures on this site are computed at build time, not written into pages, which is what makes that possible.
The record's verification date is reset to the day of the recheck, so the correction is visible on the page itself rather than absorbed silently.
Where the error was a real disagreement between sources rather than a mistake, it is recorded as a divergence on the state page and left standing instead of being smoothed over.

Every researched state carries that verification date, and the dates on this site currently run from July 22, 2026 to July 27, 2026. A reader can therefore judge staleness directly instead of taking freshness on trust. A page checked two years ago says so.

One gap should be stated plainly rather than papered over: this site publishes no contact address, so there is currently no channel for reporting an error to it. That is a real limitation of the site as it stands. Until one exists, the check available to a reader is the one the whole apparatus is built for, which is that every claim names its document and the document overrides the page.

How much of the country this covers

26 of the fifty states have not been researched. No statute has been read for those records, they are counted in no figure anywhere on this site, and they appear in the comparison table marked as unresearched rather than filled in from a secondary summary. They are: California, Colorado, Connecticut, Georgia, Hawaii, Illinois, Indiana, Iowa, Kentucky, Maine, Massachusetts, Michigan, Minnesota, Mississippi, Nevada, New Jersey, New Mexico, New York, North Dakota, Ohio, Pennsylvania, Rhode Island, South Carolina, Virginia, West Virginia, Wisconsin.

The researched states are also not a random sample. Most were selected because something happened there legislatively, which tilts the set toward states that have exemptions, so a ratio drawn from them should not be scaled up to fifty. That caveat is repeated wherever a ratio appears, and the methodology page works through what a different definition of the key terms would do to the number.

Research continues, and because the figures are derived from the records rather than written into the pages, they move on their own when it lands. The fifty state table is the fastest way to see the shape of the gap.

Who stands behind it

No named author, no organization, and no institutional affiliation. Rather than invent one, the honest thing is to say what actually governs the work, which is the standard set out in section 2 and the definitions on the methodology page.

That has a consequence a reader should weigh. Nothing here should be believed because of who wrote it, since nobody is offering a reputation as security. What is offered instead is that every assertion is traceable to a document, that the documents are named at full size rather than buried, that the classification rules are published, and that the arithmetic can be reproduced. A reference built that way can be audited by a stranger, which is the only kind of trust available to an anonymous one.