Two unlike questions, one instrument
Accuracy is not a single measurement here, and separating it is the first useful thing to do. An identifier is asked to perform two unlike jobs at once: to say what a thing is, and to say what it is worth. It is markedly better at the first than the second, and the honest error bars on the two are different sizes.
Naming a round brilliant in a four-claw platinum mount is a recognition problem, and recognition from photographs has become very good indeed. Saying what that stone would fetch is a valuation problem, and valuation from a photograph is an estimate resting on an estimate. An instrument that reports both in the same confident tone is flattening a real distinction.
What a photograph reads well
A camera captures geometry, colour and surface. That is a good deal more than it sounds, because most of what a jeweller notices across a counter is also geometry, colour and surface.
- Shape and cutting style — round brilliant, oval, emerald, cushion, pear. The facet pattern is unmistakable in a clean image.
- Setting and metal — claw, bezel, pavé or halo, usually with the metal's colour and finish.
- Hallmarks — 585, 750, PT950 and a maker's mark, provided the inside of the band is photographed and in focus.
- Apparent size — the millimetre spread read against the finger or the mount, which is how carat weight is inferred rather than measured.
- Condition — a chipped girdle, worn claws, a stone sitting low or slightly crooked in its setting.
What a photograph cannot read at all
The tests that actually separate a diamond from its imitators are not photographic, and an image carries none of them. Thermal and electrical conductivity, refractive index, specific gravity, hardness — every one of them needs the stone in the hand and an instrument against it.
Moissanite settles the argument. It is close enough in appearance that a picture cannot rule it out, and common enough that it has to be ruled out. The doubled facet edges that give it away are a loupe observation, made at an angle and a magnification the camera was not holding. This is why the honest answer to “is my diamond real” is never a photograph, however good the photograph is.
Origin is the second thing no image reaches. A laboratory-grown diamond and a mined one are the same crystal, with the same hardness and the same optical behaviour; the equipment that separates them reads growth structure and trace elements, not appearance. An instrument that claims to tell you from a photograph which of the two you are holding is reporting something it cannot know.
Why the four Cs come back as a range
A grading laboratory works under conditions a telephone does not reproduce, and does not try to. The stone is unmounted and cleaned, and viewed loose under controlled, colour-corrected light. Colour is compared against a set of masterstones rather than judged in isolation. Clarity is assessed at ten times magnification. More than one grader sees it, and a report is issued only where they agree.
A photograph of a set ring has none of that. The metal around the stone throws colour into it, so a yellow-gold basket will warm a stone that graded near-colourless while it was loose. The inclusions that decide a clarity grade sit well below what any telephone camera resolves at that distance. And carat weight is not weighed at all; it is inferred from spread, which a setting can flatter by a tenth of a carat without anyone intending it to.
So the four Cs an instrument returns are a considered estimate of a band, and are worth reading as one. Where a reading and a laboratory report disagree, the report is right.
Where the value figure comes from
The price is not observed in the photograph at any point. It is modelled. The instrument settles on a likely shape, a likely spread and a likely grade band, looks up what stones in that band have been trading at, and returns a range.
Every uncertainty upstream widens that range. A clean read on a plain solitaire yields a tighter band than a cluster whose centre stone has its edges hidden by the mount. And the model knows nothing at all about the two factors that move a real sale most: the specific stone in front of a specific buyer, and whether it arrives with paperwork.
How to make the reading as good as it can be
Most disappointing results are photographs rather than models. The instrument can only work with what it is handed, and a few minutes of care changes the answer more than anything else available to you.
- Clean the stone first. Skin oil and hand cream flatten brilliance and read as a poor cut.
- Use indirect daylight. A window with the sun off it beats every lamp in the house; direct sun and a phone torch both blow out the facets and lose the pattern that identifies the cut.
- Fill the frame, then stop. Tap to focus on the table of the stone, and move back if the focus will not hold — a sharp small stone reads better than a blurred large one.
- Photograph it more than once: face-up, then at a slight angle, then the profile. The profile is what carries depth and girdle thickness.
- Photograph the hallmark separately. A close, sharp image of the inside of the band is often the most informative frame in the set.
- If a grading report exists, photograph that as well. A report answers in seconds what an image cannot answer at all.
What the reading is actually for
An indicative figure has real uses, so long as it is not asked to be a different kind of thing. It tells you whether you are holding something worth tens or something worth tens of thousands, which decides everything you do next. It gives you a number to carry into a jeweller's, so that the first figure you are told is not the only figure you hold. It is enough to establish that a piece ought to be scheduled on a policy rather than left under a contents cap.
What it is not: a certificate, an appraisal, or a basis on which to sell. Nobody at the house sees the stone — the house says so on every screen it publishes, and it bears repeating here. Before money moves, a gemological laboratory grades the stone and a qualified appraiser values it. Those are two different services, and an insurance claim generally wants both.
The short answer
On identity of kind — shape, cutting style, setting, metal — an instrument is dependable enough to rely on in ordinary use. On the four Cs it returns a band rather than a grade, and the band is honest. On whether the stone is a diamond at all it cannot be conclusive, and any instrument implying otherwise is overreaching. On value it offers a starting figure, whose whole job is to stop you being surprised by someone else's.
That is a narrower claim than this category usually makes on its own behalf. It is also the claim that survives contact with a laboratory report, which is the only test of an instrument that finally counts.