The customer puts a thirty-gram bracelet on the counter, its 750 hallmark clearly legible. The buyer has a few minutes to price it, and the price hangs almost entirely on a number nobody can see: how many parts of gold per thousand the metal really contains. Meanwhile, the customer waits. If the bracelet is a gold shell around a low-grade core, the hallmark misleads and the scale won't notice. That is why fineness testing is the technical step that sets the margin of a gold buyer, a bullion dealer or a refiner. X-ray fluorescence made it fast and non-destructive. It also has a precise physical limit, and it pays to know it before trusting the number on the screen.
Operators weighing up an instrument for their own counter can start with a practical test. PITECH arranges a comparison measurement on the items you handle every day and prepares a tailored quote within 24 hours: just write to us through the contact form or use the WhatsApp button.
Parts per thousand, karats and legal finenesses: the number you are checking
The counter thinks in parts per thousand. A 750 is 18 karat, 750 parts of gold per thousand; 585 is 14 karat, 375 is 9 karat and 916 is 22 karat, while investment gold is expressed as 999 or 999.9. In Italy, article 3 of Legislative Decree 251/1999 sets the legal gold finenesses at 750, 585 and 375 and allows any fineness above the highest one. It also requires the fineness to be guaranteed on melting for every part of the article. For silver the legal values are 925 and 800, for platinum 950, 900 and 850, for palladium 950 and 500.
The same article rules out negative tolerances on legal finenesses, apart from a few listed exceptions for platinum, palladium and lost-wax cast gold. The hallmark, in other words, promises a minimum. For a buyer the consequence is practical. A 750 that reads 740 at several clean points deserves a second look, provided the gap exceeds the uncertainty the instrument has demonstrated on that alloy type. Internationally, ISO 9202:2026 recommends the range of finenesses for jewellery alloys and defers to national rules on marking.
What happens under the measuring window
The instrument hits the surface with an X-ray beam. The excited atoms re-emit radiation at characteristic energies, and the lines of gold, silver, copper and zinc each carry their own signature. The detector counts them and the software turns counts into percentages through a calibration. Within seconds, or a few tens of seconds when more precision is needed, the fineness appears with its spectrum.
The two belong together. The fineness tells you how much gold sits in the measured spot; the spectrum tells you what it is alloyed with. Unexpected nickel in a yellow gold, or silver too high for the colour of the piece, are clues the number alone would hide. Precision improves with measuring time, because statistical uncertainty falls as counts accumulate, and high fineness needs dedicated calibrations and samples of known composition. Marco Actis Grande's paper at the 2016 Santa Fe Symposium makes exactly this point: the shape, roughness, size and composition of the piece shift the result, and only a sound operating method keeps false values out.
Touchstone or XRF: two readings of the same surface
The touchstone survives because it is cheap and works anywhere. You streak the piece on the stone, apply acids matched to different finenesses and watch whether the streak holds. The result is a band, along the lines of "at least 585". It depends on the eye of whoever runs the test. The stone shares XRF's main limit, since it too reads only the skin. XRF adds a numerical value, the full composition and a spectrum you can store, show the customer or attach to the transaction file.
A hydrostatic balance measures density by weighing the piece in air and in water. It's a good check on bars and compact shapes, and a much weaker one on jewellery with stones, cavities or solder joints. At the far end sits cupellation, the fire assay of ISO 11426:2021, which the standard applies to homogeneous material with a gold content preferably between 100 and 999.5 parts per thousand. Christopher Corti's review of gold assaying at the 2001 Santa Fe Symposium calls it the reference technique against which all others are compared. It destroys the sample, though, and needs a laboratory. Nobody cupels the wedding ring a customer brings to the counter.
Above 999 the reference method changes too. For high-purity gold, platinum and palladium, ISO 15093:2020 specifies a difference method using ICP-OES: the impurities are measured and the fineness is obtained by subtraction. That is refiners' and bullion territory, a long way from daily counter work.
XRF only reads the surface, and the skin is thin
The weak point of X-ray fluorescence is the depth the signal comes from. Gold's L lines, around 9.7 keV, are absorbed quickly by the metal itself. Using the attenuation coefficients tabulated by NIST, in pure gold at normal incidence their intensity falls to just over a third within about 4 micrometres, roughly a twentieth of the thickness of a human hair. Four micrometres is very little. Higher-energy lines, such as silver's K lines, come from deeper: in the order of ten micrometres in a gold alloy, more in a silver article.
On a homogeneous piece the limit doesn't matter, because the surface represents the core. Trouble starts when surface and core differ, through manufacturing or through fraud.
Gold-plated items: how to spot them before paying
Plating is the classic case. Thin electroplated gilding still lets the signal from the underlying metal through, and a spectrum showing silver, copper or zinc in proportions that don't fit the reading is already a warning. As the layer grows towards a few micrometres, the core signal fades and the instrument sees only the gold skin. Surface enrichment is subtler: pickling and finishing can deplete the surface of copper and silver and leave it richer in gold than the core, with no intent to deceive, so readings come out above the true fineness. On white gold, finally, rhodium plating adds a rhodium signal that has to be recognised for what it is and not mistaken for an alloying element.
Take a worked example built on the opening bracelet, assuming a homogeneous alloy under the gilding. Three ten-second readings on the links come out between 748 and 752, in line with the hallmark. A fourth reading inside the clasp, where wear has rubbed the surface away, gives 590 with copper and zinc clearly visible. Filing a hidden spot confirms the second value. This time the hallmark was lying. From that point the bracelet has to be treated as gilded 585. On thirty grams, the gap between the two finenesses is worth about five grams of fine gold (30 × 0.165), paid for and never received.
The method that follows is easy to describe. It does take discipline. Clean the measuring spot, measure at least two or three points with one of them in a worn or hidden area, then compare fineness, spectrum and hallmark. In doubtful cases, with the customer's consent, file a spot that won't show and measure again. Instruments with a coating-measurement mode, of the kind described in ISO 3497:2000, can also estimate gilding thickness, provided the layer structure is known and the calibration uses suitable samples.
On your typical items, from mixed scrap gold to wedding bands and small bars, the difference between two configurations only shows up when you measure. PITECH can arrange a comparison test on your own material: request a test on your items or message us on WhatsApp.
Small bars and tungsten cores: when you have to look inside
With bars the risk changes shape. Tungsten's density is almost identical to gold's, about 19.3 grams per cubic centimetre for both. A gold-clad tungsten core can therefore pass a routine hydrostatic weighing as well as XRF, which sees only the shell. It is the worst case for a buyer. The LBMA, looking back at the concerns raised in the early 2000s, reports that such inserts were found in small bars but never in the large bars of the Good Delivery system. It also notes that London vaults introduced ultrasonic testing as a precaution, because the speed of sound in tungsten is about twice that in gold and a probe exposes an insert immediately.
For a bullion dealer buying small bars from private sellers, the practical rule is to pair the surface reading with a check that goes through the piece, accurate density or ultrasound, before paying full value.
Who sets the acceptance threshold at the counter
The law sets the finenesses and the official methods of analysis. Article 3 of Decree 251/1999 refers to the implementing regulation for the methods and the maximum permissible error. Fineness inspections fall to the Chambers of Commerce, which take samples and have them assayed by authorised laboratories. No rule says when a gold buyer may trust an XRF reading. That threshold is set by the operator. It is worth writing down, spelling out which deviation from the declared fineness triggers a second reading, which one triggers filing, which one sends the piece to the assay lab.
The basis for setting it is the uncertainty the instrument demonstrates on samples of known fineness from the same alloy family, re-checked at regular intervals. Without that rule, a three-digit number is just an opinion.
What Italian law asks of gold buyers, and where measurement helps
Since 2017 the trade has been governed by Legislative Decree 92/2017. Articles 3 to 6 require registration with the OAM register, customer identification and traceable payments from 500 euros upwards. Every transaction then needs a numbered record with a description of the item, the metal price taken from a reliable and independent source, the valuation and two digital photographs from different angles, all kept for ten years. The decree doesn't require a measuring instrument. An XRF reading saved with its spectrum, however, makes the valuation in that record reasoned and traceable, and protects the dealer when a customer or an inspector challenges it. Since 17 January 2025, as the Bank of Italy notes, professional gold operators under Law 7/2000 also fall under the OAM, following Decree 211/2024.
On hallmarking, Italy joined the Vienna Convention on the control and marking of articles of precious metals through Law 55 of 15 May 2023, and has been listed among the Contracting States since 15 December 2023. The resulting Common Control Mark (CCM) is voluntary. In Italy it is applied by the assay offices of the Chambers of Commerce of Alessandria-Asti, Arezzo-Siena and Vicenza, as the Ministry of Enterprises and Made in Italy explains. One boundary every buyer should know: a bench XRF produces a technical and commercial assessment, while official assaying of fineness belongs to laboratories and enforcement bodies. Promising a customer a certification the instrument cannot issue is a professional mistake.
When a counter reading closes the deal, and when the lot goes to assay
The line runs through the value at stake. For used jewellery gold that is homogeneous, clearly hallmarked and consistent across several readings, a methodical XRF measurement is enough to close the deal. For items with conflicting readings, for small bars from private sellers and for lots heading to refining, where a few parts per thousand weigh on kilograms of metal, the surface reading becomes the filter that decides what goes to assay. A counter with few purchases, almost all standard hallmarked jewellery, can stay with a touchstone and an outside lab; an instrument starts paying for itself as volumes, doubtful items and disputes grow. An X-ray instrument also brings the radiation-protection duties of Italian Legislative Decree 101/2020. They differ between a shielded bench unit and a handheld one, and the offer should make clear who takes care of them.
One criterion holds the rest together: the higher the value of a single piece, the more the decision moves from the counter to the lab. Readers who want to go deeper into the instrument itself, from detector to precious-metal calibrations and configurations, will find the details in the guide to XRF analysis of gold and precious metals. If the decision concerns your own counter, the next step is a test: PITECH compares the metals, volumes and finenesses you handle, measures your typical items and proposes a configuration only after that.
Frequently asked questions on verifying gold fineness
How do you verify gold fineness without damaging the item?
With X-ray fluorescence: the instrument reads the surface composition in seconds, without acids and without removing metal, and returns the fineness in parts per thousand together with the spectrum. On doubtful items it should be backed by readings at several points and, where needed, a small file mark in a hidden spot or a density check.
Touchstone or XRF: which is more reliable at the counter?
XRF, because it gives a numerical value, the full composition and a spectrum you can archive, while the touchstone gives a fineness band that depends on the operator's eye. Both read only the surface, so neither one, on its own, exposes thick plating.
How do you spot gold-plated items with XRF?
By measuring at several points, including worn or hidden areas, and by reading the spectrum: thin gilding still shows the metal underneath, while thick plating is exposed by a small file mark and a new reading. Bars and high-value items also need a check that goes through the piece, such as accurate density or ultrasound.
How deep does XRF read into gold?
A few micrometres for the gold lines: in pure gold the signal at about 9.7 keV drops to just over a third within roughly 4 micrometres, according to NIST attenuation coefficients. Higher-energy lines, such as those of silver, come from deeper layers, but stay in the order of ten micrometres in gold alloys.
Does XRF also measure silver, platinum and palladium?
Yes. An instrument configured for precious metals reads gold, silver, platinum, palladium, rhodium and alloying elements such as copper, zinc and nickel, and tells yellow, white and rose gold apart. High fineness needs dedicated calibrations, reference samples and longer measuring times.
How much does an XRF analyzer for gold buying cost?
It depends on the configuration: detector type, precious-metal calibrations, optional coating-thickness measurement, bench or handheld format and management software. PITECH prepares a tailored quote within 24 hours and arranges a test on your own items, so the choice rests on real readings rather than a price list.