Academy · Foundation
Testing gold yourself — and what each test cannot see
There are five tests you will meet, and they form a ladder: magnet, density, touchstone and acid, XRF, and fire assay. Each one is quicker and cheaper than the one after it, and each one leaves a specific hole that only the next one closes.
You can do the first two yourself with a kitchen scale and a glass of water. The third you can watch being done. The fourth happens at the counter in thirty seconds. The fifth destroys the piece, which is why almost nobody runs it on a chain.
The useful thing to learn here is not how to test. It is what each test still cannot rule out — because that is the gap a determined seller works in, and it is the same gap a professional assayer spends a career learning to close.
1 · The magnet, and what it really proves
Gold, silver, platinum, palladium and copper are all non-magnetic. A strong magnet held against a piece of real gold does nothing. If the piece jumps to the magnet, there is steel or nickel in it, and you are finished testing.
That is the entire value of the test, and it is genuinely worth thirty seconds. But note what it does not prove. Tungsten, lead, brass and bronze are all non-magnetic too. A magnet catches the crudest fakes and passes every serious one.
There is a better magnet test for silver, which is worth knowing because it uses different physics. Silver is the most electrically conductive metal there is. Slide a strong neodymium magnet down the face of a genuine silver bar and it visibly slows, dragged by the eddy currents it induces. On a plated or base-metal bar it slides straight off. That test works on bars with a flat face; it tells you nothing about a thin chain.
2 · Density — the best test you can run at home
Every metal has a fixed density. Gold's is unusually high, and that is what makes the test work: very few things are as heavy for their size.
Weigh the piece dry. Then weigh it again fully submerged in water, hanging from a thread so it touches nothing. The difference between the two readings is the weight of the water it displaced, and because a millilitre of water weighs a gram, that difference is the volume in cubic centimetres.
Illustrative figures. Use a scale that reads to 0.01 g, and take each reading three times.
The arithmetic, in full
A bangle sold as 22 karat.
- Weight in air: 20.00 g
- Weight submerged: 18.87 g
- Water displaced: 20.00 − 18.87 = 1.13 g, so volume = 1.13 cm³
- Density: 20.00 ÷ 1.13 = 17.70 g/cm³
A 22-karat yellow gold alloy sits around 17.7 to 17.8. This piece is consistent with what it claims to be.
Run the same piece at 18 karat and the submerged weight would have been about 18.71 g — a displacement of 1.29 g and a density of 15.5. The whole difference between 22k and 18k, on a 20-gram piece, is 0.16 g of water. That is why the scale matters more than the maths.
What density still cannot rule out
Three things, and the third is the serious one.
- Hollow pieces and air. A trapped bubble under a hollow bangle displaces water it should not. Submerge slowly, tilt the piece, and watch for anything escaping.
- Solder and stones. Heavy solder at the joints and any stone set into the piece both drag the reading away from the metal you are trying to measure. Density is only clean on a plain, unset, single-alloy item.
- Tungsten. Tungsten has a density of 19.25 g/cm³. Pure gold is 19.32. Those two numbers are 0.4% apart, and no arrangement of a kitchen scale and a glass of water will ever separate them. A tungsten core under a real gold shell reads as gold on every density test ever devised — this is exactly why the counterfeits that matter are made of tungsten and not lead.
Density is not a weak test. On jewellery it is a genuinely good one, because tungsten is hard, brittle and nearly impossible to work into a chain. On a large plain bar, which is easy to core, it is close to worthless on its own.
3 · Touchstone and acid
The oldest test still in daily use, and you will see it done in front of you. The piece is rubbed on a dark, fine-grained stone to leave a streak of metal. Graded acids are dropped on the streak: each acid dissolves everything below a certain purity and leaves everything above it. Whichever acid the streak survives sets the karat.
A skilled hand does this quickly and gets close. Two things limit it.
It reads only the streak — which means it reads only the surface. A plated piece rubbed lightly leaves a streak of the plating and passes. The defence is a deep rub, or a small notch filed into an inconspicuous edge so the streak comes from inside the metal. If nobody files, nothing has been tested but the coating.
And it is comparative, not quantitative. It tells you the streak behaves like 22 karat. It does not give you 91.6 as a number, and it cannot tell 91.6 from 90.5 — a gap that is invisible to acid and quite visible on a large bill.
4 · XRF at the counter
The handheld gun most hallmarking centres and larger shops now use. It fires X-rays at the piece; the metals fluoresce back at energies unique to each element, and the machine reads off the composition. It is fast, non-destructive, gives numbers, and identifies the alloying metals as well as the gold.
Its limit is not accuracy. It is depth. The signal the machine reads comes from the outermost layer of the piece — a few tens of micrometres at most, and less for the lighter elements. Below that, the fluorescence cannot escape the metal to be counted.
So XRF sees the skin. Ordinary decorative plating is thin enough that the machine sees straight through it and reports the substrate honestly. A deliberately heavy gold layer, laid on thick enough to exceed the reading depth, is a different matter: to the gun, that surface is simply what the object is made of. This is the same blind spot as the touchstone, arrived at by better physics, and it is why a serious assay office does not stop here on anything valuable.
Two smaller points worth carrying. XRF reads what it is pointed at, so a reading taken on a clasp is a reading of the clasp. And surface enrichment — the thin gold-rich skin that some finishing processes leave behind — biases the result upward on a piece that is genuinely solid.
5 · Fire assay, the referee
The method every other test is ultimately calibrated against, and the one the world's refiners settle disputes with. A sample is cut from the piece, wrapped in lead and silver, and heated in a porous cup. The base metals oxidise and are absorbed into the cup; what remains is a bead of gold and silver, and the silver is then dissolved away in acid. Weigh what is left. That is the gold.
It resolves purity to a fraction of a part per thousand, it is indifferent to plating, cores, solder and surface enrichment, and it has been doing this since long before anyone had X-rays.
The catch is in the first step. You have to cut a piece off. Fire assay is destructive, it takes hours rather than seconds, and it costs money — so it is used on bars, on scrap lots, on refinery intake and in disputes, and essentially never on a necklace someone is thinking of buying.
Which test answers which question
- Is there steel in it? Magnet. Thirty seconds, free, conclusive when it fails.
- Is the alloy roughly what they say? Density, on a plain unset piece, with a 0.01 g scale.
- What karat is the surface? Touchstone and acid, or XRF for a number.
- What is inside it? Nothing on this list, short of fire assay — or cutting it open.
That last line is the honest summary of the whole lesson. Every fast test reads the outside. The only tests that read the inside destroy the thing they read.
What this lesson cannot tell you
It cannot make you an assayer. Reading about a touchstone is not the same as having done ten thousand streaks and knowing by the colour of the smear before the acid lands. The tests are simple; the judgement is not, and the judgement is what people are actually paying for when they pay an assay office.
It also cannot tell you whether the price is fair. A piece can be exactly 22 karat, correctly hallmarked, honestly weighed, and still be an expensive way to buy gold — because purity and price are two entirely separate questions. The second one is Lesson 04.
Kept in this browser only — no account, and nothing is sent anywhere. Finish all eight and you can claim a certificate.
Educational content only. Not financial, investment or tax advice. Aurix does not buy, sell, store or broker metal, and does not quote executable prices.