How to Identify Gemstones in Jewelry

Last updated: August 2026

A mounted stone takes some tests off the table — not all of them. Here's what a setting still allows, what it forbids, and how estate buyers, jewelers, and counter trade work a piece without pulling a single stone.

What mounting changes

Identification works by measuring physical and optical properties and matching them against known values. A setting doesn't change the stone's properties — it changes which ones you can reach. The skill of working mounted goods is knowing which observations are still trustworthy and letting go of the ones that aren't.

One rule before anything else: never scratch-test jewelry. Hardness testing has a place on rough material you own — not on a finished piece, and never on goods that belong to someone else.

What still works — and what doesn't

Loupe & microscope

Inclusions, facet wear, doubling, gas bubbles, mold lines. The single most productive test on mounted goods — glass and many synthetics give themselves away here.

Refractometer (table facet)

A flat, polished, accessible table takes a normal RI reading mounted. Bezels and very low settings can prevent optical contact — skip rather than guess.

UV fluorescence

Non-contact, works in any setting. LW/SW response patterns narrow candidates fast — and some synthetics fluoresce differently from their natural counterparts.

Chelsea filter

Non-contact. For red and green stones, the reaction separates several classic confusion pairs in seconds.

Specific gravity

Hydrostatic weighing needs a loose stone — the metal's weight makes the measurement meaningless. This is the big loss when working mounted.

Scratch / hardness testing

Never on jewelry. Destructive, and unnecessary — the non-contact tests above carry the identification.

Closed-back and foil-backed settings additionally block transmitted-light observation — and in antique pieces, a foil back is itself a finding worth noting.

The mounted-stone workflow

  1. Clean and inspect the setting. Open back or closed? Is the table reachable for a refractometer?
  2. Loupe at 10x. Inclusions, wear patterns, doubling of back facets, bubbles. Record what you see, not what you hope.
  3. RI on the table if accessible. One good reading typically cuts the candidate list dramatically.
  4. UV, then Chelsea for reds and greens. Both non-contact, both fast.
  5. Enter observations in GemID with "set stone" enabled. The engine excludes tests the mounting makes impossible and recommends the best available next test instead of one you can't run.
  6. Read candidates; refer when it matters. Results are candidate identities consistent with observations. Natural-versus-synthetic and treatment questions on consequential stones belong with an accredited laboratory.

Who works this way

Estate & auction intake

Mixed lots of mounted goods, worked stone by stone with lot numbers and per-piece notes — then a consolidated report for the consignor. The dealer workflow covers this end to end.

Jewelry stores & counter trade

Buying over the counter means deciding with the tests the setting allows, documenting what was observed, and flagging what needs a lab — the same discipline pawnbrokers apply to walk-in goods.

Inherited pieces

Before selling a family piece, the honest first step is the property record: what the observations are consistent with, documented. GemID gives you that record — a laboratory gives you certainty. Both have their place.

Common questions

Can a gemstone be identified without removing it from the setting?

Usually, yes — to the level of candidate identities. Loupe observation, a table-facet RI reading, UV response, and the Chelsea filter together narrow most stones to one or two candidates. What mounting genuinely prevents is specific gravity and, in closed settings, transmitted-light work. Certainty on natural versus synthetic often needs a laboratory regardless of mounting.

How can I tell if the ruby in my ring is real?

"Real" hides three questions: is it corundum (versus garnet, spinel, or glass — instruments answer this), is it natural (versus synthetic ruby, which is chemically the same mineral — microscopy of growth features answers this, often at a lab), and is it treated. A property-based examination settles the first question and flags the second two for referral when the piece warrants it.

Can a photo app identify the stones in my jewelry?

No — color and shape don't identify gemstones, mounted or loose. Red garnet, red spinel, ruby, and red glass can photograph identically. Identification comes from measured properties; here's the full comparison of the two approaches.

Working mounted stones? GemID's set-stone mode ranks candidates on the observations a setting actually allows — and turns the session into a client-ready report.

Open GemID →

Studying for the FGA or GIA practical? GemID drills instrument-based identification — refractometer, SG, and UV readings — and runs timed mock exams built around the real exam formats.

Practice exams →

Related