There is more bad advice about testing gold than about any other subject in jewelry, and most of it circulates because it sounds plausible. Gold is not magnetic, so a magnet must tell you something. Gold does not tarnish, so bleach must prove it. Gold is dense, so it should feel heavy in the hand.
Each of these is either wrong or so weak as to be useless, and acting on them is how people buy plated brass at solid gold prices. This guide sets out what actually works, what each test genuinely proves, and what the failure modes cost.
The hallmark system
The starting point is not a test at all. It is the stamp, and understanding what a hallmark does and does not mean is the difference between a useful signal and false confidence.
A hallmark is a statement of fineness — the proportion of gold in the alloy by weight — expressed in parts per thousand.
| Mark |
Gold content |
Karat |
| 375 |
37.5% |
9K |
| 417 |
41.7% |
10K |
| 585 |
58.5% |
14K |
| 750 |
75.0% |
18K |
| 916 |
91.6% |
22K |
| 999 |
99.9% |
24K |
Older pieces may use karat stamps instead — "14K", "18K", "22K" — and British pieces carry a set of marks rather than a single one: a maker's mark, an assay office mark, a date letter, and the fineness mark. French pieces carry an eagle's head or another guarantee mark. These systems vary by country, and reading them is a skill in itself.
What a hallmark proves, and what it does not
A hallmark proves exactly one thing: that someone stamped a number on the piece.
It does not prove the metal matches the stamp. It does not prove the piece is solid rather than plated. It does not prove who made it, when, or where. It is a declaration, and the question is who enforces it.
In the United States, hallmarks are self-declared. A manufacturer stamps "14K" on a piece on their own authority, and the mark is only tested if someone complains. The National Gold and Silver Stamping Act makes it illegal to misrepresent the quality of gold, and the FTC's Jewelry Guides set out what may be claimed, but enforcement is complaint-driven and light. A counterfeit 14K stamp costs a fraudster nothing to apply and is prosecuted rarely.
In the United Kingdom, hallmarking is a legal requirement. Any gold item above one gram must be submitted to an assay office — London, Birmingham, Sheffield or Edinburgh — which independently tests it and applies the marks. Selling an unhallmarked piece above the weight threshold is an offence. This makes British pieces substantially more trustworthy than American ones, and it is why a vintage British piece with a full set of assay marks is worth more than a similar unmarked item.
The practical consequence: a stamp is a reason to look further, not a reason to stop looking. On a $40 chain from a marketplace, "14K" is marketing. On a piece with a full British assay set, it is a legal attestation.
A note on prices
All prices are street prices observed in 2026 across US and EU retailers.
The tests, ranked by what they actually establish
Here is the honest hierarchy. Only the top of this list is definitive, and everything below it is a screening tool.
| Test |
What it establishes |
Destructive? |
Cost |
Reliability |
| XRF analysis |
Elemental composition to a fraction of a percent |
No |
$0–$50 at a jeweller |
Definitive |
| Fire assay |
True gold content, the industry standard |
Yes |
$75–$200 |
Definitive |
| Acid test on a scratch |
Approximate karat range |
Slightly |
$10–$30 kit |
Good, operator-dependent |
| Electronic gold tester |
Karat estimate by conductivity |
No |
$20–$60 |
Moderate |
| Specific gravity |
Density, which narrows the alloy |
No |
Requires scale |
Moderate |
| Magnet |
Rules out magnetic base metals only |
No |
$5 |
One-way only |
| Visual inspection |
Nothing about content |
No |
— |
Worthless alone |
| Bite, vinegar, bleach |
Nothing |
Yes |
— |
Worthless |
XRF analysis
X-ray fluorescence is the correct answer to the question. An XRF analyser fires X-rays at the metal and reads the characteristic emissions of the elements present, giving a composition readout — for example, 58.4 percent gold, 30 percent copper, 8 percent silver, and trace amounts of others.
It is non-destructive, takes under a minute, and is accurate to a fraction of a percent on the surface. Many jewelers and nearly all pawnbrokers and scrap buyers have one, and a reading is usually free or costs $10 to $30. If you want a definitive answer without damaging the piece, this is the test.
Two caveats. XRF reads the surface, so a heavily plated piece will read as the plating alloy unless the analyser is set to penetrate deeper — better instruments have a thicker measurement mode, and a good operator will use it. And XRF does not tell you whether a piece is hollow; it tells you what the metal is, not how much of it there is.
Fire assay
The industry standard for settlement, and the only test that measures the true gold content of the whole piece. A sample is melted with lead and cupelled, and the resulting bead is weighed. It is destructive — the sample is consumed — and it is what a refinery uses when you sell scrap.
You will not commission a fire assay on a ring you intend to wear. It matters because it is the standard against which everything else is calibrated, and because a scrap buyer quoting you a price is working from one.
Acid testing
The acid test uses a touchstone — a piece of fine-grained black stone — and a set of nitric acid solutions of different strengths. You rub the metal across the stone, leaving a streak, then apply the acid matched to the karat you expect. If the streak survives, the metal is at least that karat. If it dissolves or changes colour, it is lower.
It is a real test and it works, in trained hands. Its limitations are significant:
- It is surface-level. A thickly plated piece can pass.
- It is operator-dependent. Reading a streak is a judgement call, and the difference between 14K and 12K can be a subtle colour shift.
- It damages the piece slightly. You are removing a small amount of metal onto the stone.
- The acids are genuinely hazardous. Nitric acid causes serious burns, and kits are sold to consumers with inadequate warnings.
A jeweller charging $10 to $30 for an acid test is doing something reasonable. Buying a kit and testing your grandmother's ring is a good way to damage it and misread the result.
Electronic gold testers
These devices measure electrical conductivity and translate it into a karat estimate. They are non-destructive and cheap, and they are better than nothing. They are also easily fooled: conductivity varies with alloy composition, temperature and surface condition, and a plated piece often reads high. Treat the reading as a hint, not an answer.
Specific gravity
Gold has a density of 19.3 g/cm³, which is far higher than brass at 8.5, copper at 8.9, or steel at 7.9. Measuring density is the one reasonably accurate test a careful person can do at home.
The method: weigh the piece dry, then weigh it suspended in water, and divide the dry weight by the difference. The result is specific gravity. A 14K yellow gold piece should land near 13 to 14.5; 18K near 15 to 16.5; 24K near 19.3.
Three conditions make it work. The piece must be solid — a hollow bangle will read low because the calculation assumes a solid volume. It must be stone-free, because a set diamond has a density of 3.5 and drags the average down. And the scale must be accurate to 0.01 grams, because the whole calculation rests on a small difference between two numbers.
A result close to the expected figure is meaningful. A result far off is not proof of fraud on its own — it may be a hollow piece or a different alloy — but it is a reason to get an XRF reading.
The magnet test
Gold is not magnetic. Neither are aluminium, copper, brass, zinc, or most of the metals used to fake it.
So a magnet that does not stick proves nothing at all. This is the single most misunderstood point in the category, and it is the basis of an enormous amount of confidently wrong advice.
A magnet that does stick proves the piece is not solid gold — because gold is diamagnetic and will not attract a magnet. So the test works in one direction only. It can rule a piece out. It cannot rule one in.
Visual inspection
Colour, lustre and heft are worth observing and worth nothing as proof. Good gold plating looks like gold. Plated brass has roughly the right colour. Tungsten has been used to fake gold bars precisely because its density is close to gold's and it takes a gold-coloured plating well.
The home tests that do not work
A short list of things that circulate widely and should be ignored.
The magnet alone. Covered above. Absence of attraction is not evidence of gold.
The bite test. Gold is soft, so teeth leave a mark — the logic goes. What actually happens is that you damage your teeth and transfer bacteria onto the jewelry, and soft base metals mark just as easily. Worse, biting a plated piece can crack the plating and expose the base metal, which is the only thing the test achieves.
Vinegar. Gold does not react with acetic acid, so no reaction is supposed to mean real gold. But neither do brass, copper, stainless steel, or most other base metals at household concentrations. The test has no discriminating power.
Bleach. The claim is that real gold survives bleach while fake gold discolours. Real gold does resist bleach. So does stainless steel, and so does titanium. The test separates gold from a few reactive metals and nothing more.
Rubbing the piece on skin or a tile. Produces a streak that means nothing.
Checking whether it tarnishes. Gold does not tarnish, which is true and useful as a general observation. But rhodium-plated and gold-plated pieces also do not tarnish until the plating wears, and stainless steel never does. Tarnish resistance is a clue, not a test.
Online "verify your gold" tools. There is no database you can type a hallmark into that confirms a piece is real. Report numbers exist for certified gemstones, not for gold. Any site claiming to verify gold by stamp number is selling something.
Plated, filled, vermeil, bonded, tone
The confusion in this category comes from five terms that sound similar and mean very different things. These are the definitions that matter.
| Term |
Construction |
Typical gold thickness |
Durability |
| Solid gold |
Gold alloy throughout |
Entire piece |
Decades to indefinitely |
| Gold-filled / rolled gold |
Gold layer bonded under heat and pressure |
At least 5% of total weight (US) |
Years to decades |
| Gold vermeil |
Gold plating over sterling silver |
2.5 microns minimum (US) |
1–3 years |
| Gold-plated / electroplate |
Electroplated gold over base metal |
0.5–2.5 microns |
Months to 2 years |
| Bonded gold |
No legal definition in most markets |
Unspecified |
Unpredictable |
| Gold tone / gold colour |
No gold content requirement |
None |
Unpredictable |
How much gold is actually in plating
This is the number that reframes the whole discussion. Gold plating is measured in microns — thousandths of a millimetre — and typical commercial plating is 0.5 to 2.5 microns.
For scale: a human hair is roughly 70 microns thick. Good plating is therefore between one-hundredth and one-thirtieth of a hair's width. At that thickness the gold is a cosmetic layer, and it wears through at contact points — the back of a pendant, the clasp of a chain, the inside of a ring band — within months of daily wear.
Even at the top of the range, a 2.5-micron layer on a ring will show wear inside a year. This is not a defect; it is the nature of plating. The failure is not that plating wears, it is that plating is priced and described as though it will not.
Vermeil is the regulated exception. In the United States, "vermeil" requires at least 2.5 microns of at least 10K gold over sterling silver. That is a real standard and vermeil is meaningfully better than generic plating, but it is still plating and it still wears through.
Gold-filled is the underrated one. Because the gold layer is bonded mechanically rather than electroplated, it is typically 50 to 100 times thicker than plating. In the US it must be at least 5 percent of the item's total weight, and pieces are marked "14K GF" or "1/20 14K" — the latter meaning one-twentieth of the weight is 14K gold. A gold-filled chain worn daily will last years, and at a distance it looks identical to solid gold. It is not solid gold and should not be priced as such, but it is a genuinely good product.
Bonded gold has no legal definition in most markets. It is a marketing term, and you should treat it as unspecified plating.
Gold tone means nothing. There is no content requirement, no thickness requirement and no durability expectation. A gold tone piece may contain no gold whatsoever.
Reading a plated mark
Plated pieces sometimes carry misleading stamps. A "14K" stamp on a plated item may mean the plating is 14K gold, not that the piece is 14K gold. Legitimate marks for plated goods include "14K GP" (gold plated), "14K RGP" (rolled gold plate), and "HGE" (heavy gold electroplate). A bare "14K" on a thin, light, inexpensive chain is a reason to ask questions, not a reason to relax.
Common scams
Knowing the shapes of the common frauds is more useful than any test, because the tells are in the transaction rather than the metal.
Hollow pieces sold as solid. A hollow chain or bangle looks and weighs far less than a solid one, but it is priced near solid. The tell is weight: ask for the gram weight and compare it with the dimensions. A hollow 18K bangle that weighs 8 grams is not a $1,400 piece of gold; it is a thin shell. Hollow pieces also dent irreversibly and often cannot be resized.
Plated pieces stamped 14K. The stamp is applied to deceive. This is the most common fraud in the category, and the most common defence is the price: if a new 14K chain is priced 60 percent below every comparable piece in the market, the explanation is usually the metal, not the seller's efficiency.
Gold-filled sold at solid prices. A gold-filled chain is a legitimate product with a legitimate price, which is a fraction of solid gold. Sold as solid, it is a straightforward fraud. Check for "GF" or a fraction mark like "1/20 14K" — but note that these marks are also sometimes removed or over-stamped.
"Scrap gold" schemes. Two versions circulate. In one, a buyer offers to purchase your gold at a price based on a weight reading taken on an uncalibrated or manipulated scale, then pays far below the metal value. In the other, a "refiner" takes your gold for testing and returns less metal than was submitted, or returns nothing. The defence is simple: weigh the piece yourself on a kitchen scale accurate to 0.1 grams before handing it over, photograph it, and get a written receipt with the weight recorded. Legitimate scrap buyers in 2026 pay 75 to 90 percent of the metal value for clean pieces; anything below 60 percent is a poor deal.
Tungsten and brass bars sold as gold. Common online, particularly for coins and small bars. Tungsten's density is 19.25 g/cm³ against gold's 19.3, which defeats a specific gravity test. XRF will identify it immediately, and a real assay is definitive.
"Authenticity certificates" with no issuer. A certificate from an unknown "laboratory" that is not GIA, IGI, HRD, AGS or an equivalent recognised body proves nothing. Anyone can print a certificate.
Pressure and urgency. Genuine sellers of genuine gold do not need to close the sale in ten minutes. Time pressure exists to prevent you from getting a second opinion, which is exactly the thing that would end the transaction.
What to do if you have been sold a fake
The remedies are better than most buyers expect, and the clock matters more than the argument.
Document everything immediately. Photograph the piece, the hallmark, the listing, the receipt, and any messages with the seller. Screenshot the product page before it changes, and record the seller's claimed description of the metal.
Get an independent written assessment. An XRF reading from a jeweller, with a short written note of the result on their letterhead, is worth more than any argument you can make yourself. Expect to pay $0 to $50.
Move quickly on chargebacks. If you paid by credit card, most issuers allow a dispute for goods not as described, and the window is typically 60 to 120 days from the transaction. Chargeback rights under card network rules are strong on misrepresentation, and the burden falls on the merchant to prove the item matched the description. This is the single most effective remedy for an online purchase, and it is why paying by credit card matters.
Know your consumer protections. In the United States, the FTC Act prohibits deceptive practices in jewelry marketing, and the Jewelry Guides set out what may be claimed about gold content. In the United Kingdom, the Consumer Rights Act 2015 requires goods to be as described, and the hallmarking regime gives you an independent regulator to complain to. In the EU, the Unfair Commercial Practices Directive covers misdescription. In practice, a clear written complaint citing the specific misdescription resolves most cases, because the seller knows the exposure.
Consider small claims. For amounts within your jurisdiction's limit — commonly $5,000 to $10,000 in US states, £10,000 in England and Wales — small claims court is cheap, informal and does not require a lawyer. A documented misdescription with an independent XRF reading is a strong case.
Report it. The FTC, your state attorney general, and in the UK Trading Standards all take reports. A single report rarely triggers action; a pattern of reports against one seller does.
The deadline that decides most cases
Chargeback windows are typically 60 to 120 days, and they are enforced strictly. A buyer who spends three months arguing with a seller and then tries to charge back is usually out of time. If the piece is not what was described, open the dispute early and keep arguing in parallel.
A short field guide
If you take one process away from this article, take this one:
- Read the hallmark, but do not trust it alone.
- Ask for the gram weight. Weight plus a stated karat is checkable against the metal price, and evasion is itself informative.
- Get an XRF reading for anything above a few hundred dollars. It is fast, non-destructive and usually free.
- Distrust the price. Deep discounts on gold are not bargains; gold has a published commodity price and nobody sells it far below that by accident.
- Pay by credit card so the chargeback remedy stays available.
- Do not damage the piece testing it yourself. The tests that might work are the ones a jeweller performs.
Frequently asked questions
Is a magnet a reliable way to test gold?
No, and this is the most common bad advice on the subject. Gold is not magnetic, but neither are many base metals, including aluminium, copper and brass. A magnet that does not stick proves almost nothing. A magnet that does stick does prove the item is not solid gold — so it is useful as a one-way test only.
What do the numbers 417, 585 and 750 mean?
They are fineness marks expressed in parts per thousand. 417 is 41.7 percent gold, or 10K. 585 is 58.5 percent, or 14K. 750 is 75 percent, or 18K. 916 is 22K and 999 is 24K. The mark states the gold content by weight, and nothing about who made the piece or whether the mark is truthful.
Does the acid test damage jewelry?
Yes, usually a little. The nitric acid scratch test requires cutting or rubbing the metal onto a touchstone, which removes a small amount of surface material. On a plated piece it may cut through the plating. On a hollow piece it can leave a visible mark. It is best used on scrap or on inconspicuous areas, not on a finished piece you care about.
What is the difference between gold-plated and gold-filled?
Gold-plated jewelry has a micro-thin layer of gold, often 0.5 to 2.5 microns, applied by electroplating over a base metal. Gold-filled jewelry has a gold layer bonded to the base metal under heat and pressure, and in the United States it is legally required to be at least 5 percent of the item's total weight. Gold-filled lasts years; plating lasts months.
Can I test gold at home without damaging it?
Not conclusively. A magnet can rule a piece out if it sticks, and a jeweller's loupe can help you read a hallmark, but no non-destructive home method confirms gold content. Specific gravity measurement is the only reasonably accurate home test, and it requires an accurate scale, a known metal density, and a solid, stone-free piece.
The bottom line
There is exactly one test that answers the question, and it is XRF. Everything else is either a screening tool or a myth. If you are buying anything with meaningful money attached to it, have the piece read on an XRF analyser before you pay — it takes a minute and it settles the question permanently.
For everything else, the hallmark and the gram weight will carry you a long way. Read the stamp, ask for the weight, and compare the price against the metal content. A 14K chain priced at a third of the market is not a find. It is either hollow, plated, filled, or stolen, and none of those are what you were told.
And ignore the magnet. It has never told anyone whether their gold was real.
