Editorial standards

Our methodology

How an article gets from idea to publication, what we check it against, and how we handle corrections. Written so you can audit our process.

Article topics come from two places: (a) what users actually ask the app to identify, and (b) gaps in publicly available consumer-friendly material. We don't write filler for keyword volume — if you can't use it to make a better decision about a piece of jewelry, it doesn't ship.

We use large language models to produce a first draft. We're open about this because hidden AI use is the actual quality problem — disclosed, edited AI use is just a faster typewriter.

Every AI draft is treated as untrusted input until it's been checked against the sources in the next step.

Technical claims are checked against published standards from the bodies below. Anything the AI asserted that we can't verify in one of these sources is either qualified ("commonly believed…") or cut.

• →Gemological Institute of America (GIA)
• →American Gem Society (AGS)
• →International Gem Society (IGS)
• →Smithsonian National Museum of Natural History (Smithsonian)
• →CIBJO — The World Jewellery Confederation (CIBJO)
• →Mindat.org (Mindat)

A human reads the article end-to-end before publication. We look for: AI hallucinations (invented facts presented confidently), tonal flatness, overconfidence on uncertain claims, missing caveats, and structural issues. About one in four drafts gets a substantial rewrite at this stage; the rest get light copy edits.

At publish time we set the article's published date, add structured data (Article, BreadcrumbList, and where applicable FAQPage or HowTo schema), generate a hero image, and ping IndexNow so Bing, ChatGPT-search, and other indexers re-crawl quickly. Google picks it up via the sitemap.

When a reader flags an error, or we find one ourselves, we fix the article and bump its dateModified timestamp. The sitemap's lastmod updates automatically, which prompts search engines to re-crawl.

We don't silently rewrite articles to chase rankings — updates address factual accuracy, clarity, or new developments only.

Known limitations

Things we're honest about and working on:

• !No named human reviewers (yet). Our content is written and reviewed by the internal team. We don't list paid “Editorial Advisors” on this page because we haven't engaged any — and inventing some would defeat the point of an honesty page.
• !AI identification can be wrong. Multi-angle capture mitigates this but doesn't eliminate it. For pieces where the cost of being wrong is high — insurance, estate, sale — always use a certified appraiser.
• !Value estimates are not appraisals. They're derived from public market data and should be treated as ballpark guidance only.
• !Coverage gaps. Our content depth is strongest on diamonds, mainstream coloured stones, gold/silver/platinum and major design periods. Less so on regional traditions, ethnic jewelry, and specialty markets — we're expanding coverage based on user requests.