Editorial Policy
How ChessAtlas verifies its content
Chess content on the internet has an accuracy problem. This page documents how we try to avoid adding to it: PGN validation on every opening, citations for every statistic, and an explicit correction process when we get something wrong.
Last updated: 2026-04-24.
Scope
This policy applies to every page under chessatlas.net that presents chess content: opening pages (/openings/*), feature pages, comparison pages, and the blog under /blog/*. It does not apply to user-generated content inside the web app (private repertoires, notes) which is the user’s own.
Who writes and reviews
Product copy, repertoire construction content, and the construction-of-repertoire articles are written by Antoine (see about). Drafts are assisted by AI tools for outlining and first-pass structure, but every move, variation, and factual claim is reviewed by a human before publication and cross-checked against the sources below.
PGN and move validation
Every move sequence that appears on an opening page or inside a published repertoire passes through chess.js validation at build time. If a move is illegal in the claimed position, the build fails and the page does not ship. This guardrail catches the largest class of AI-hallucinated chess content: plausible-looking move orders that do not actually work in the position.
Statistics and win rates
We do not publish a win rate, popularity percentage, or game count unless we can cite the source. When we use qualitative descriptors (“scores well at club level”, “popular among elite players”), that is intentional — it is honest about the fact that we are summarizing general consensus, not a database query. Our internal style guide forbids decimal win percentages and any “X% of players” claim without a link.
Grandmaster game references
When we cite a specific game between named players, the game is real, verifiable, and usually linked. Embedded Lichess games on blog articles use actual game IDs from our source PGN files (found in the repertoirebuilder/repertoires/generated/ directory). We do not invent matchups to illustrate a point. When we want to illustrate a pattern without a specific named source, we say “as seen in grandmaster practice” instead of attributing it to a player who never played it.
Rating-specific advice
Advice labelled “for beginners”, “for intermediate players”, or “for 1500–1800 ELO” reflects typical club-player reality, not GM-level theory. When a position is critical at the club level but irrelevant at master level (or vice versa), we note it. We do not repackage GM-level advice as universally applicable.
AI-assisted drafting
Long-form drafts are often structured with the help of AI writing tools. We disclose this once here rather than on every article because it applies to the whole blog. AI is used for outlining, first drafts, grammar, and translation assistance. AI is not used for: stating a move is good or bad, citing a player, citing a game, citing a statistic, or recommending a line at a rating level. Those judgments are made by a human reviewer.
Sources we actually use
- Lichess (game database, master database, opening explorer) — freely available at lichess.org.
- Chess.com (game database for verification of rating-band popularity).
- Chessable / ChessBase for contemporary-repertoire context where we cite them explicitly.
- Primary literature (Karpov’s My Best Games, Kasparov’s My Great Predecessors, etc.) for historical claims about specific games and players.
- Peer-reviewed sources for spaced-repetition claims: Karpicke & Roediger (2008, Science), the Ebbinghaus forgetting curve literature, and the open-source FSRS benchmark for algorithm comparisons.
Corrections
If you find an error — wrong move order, invented stat, dead link, misquoted source, anything — email contact@chessatlas.net or flag it in Discord. Verified corrections ship within 48 hours. Significant corrections are noted inline on the article (we do not silently edit content that someone has already cited).
Commercial disclosure
ChessAtlas is a commercial product. Comparison articles (/features/vs-*, blog posts comparing us to Chessable / ChessTempo / Lichess / ChessReps / Chessbook) include a visible disclosure that we are one of the products being compared. We try hard to describe competitors honestly; if a competitor is genuinely better for your use case, we prefer to say so rather than lose your trust.
Contact
Questions about this policy, a specific article, or a claim you want to verify: contact@chessatlas.net.