Editorial Policy

RTP, volatility, reel layout, paylines, bet limits and paytables are not authored. They are read out of the game — from its own configuration where that is readable, and from its in-game help screens otherwise — and written directly to the page.

Where our pipeline read a figure itself, it carries a marker naming the source it came from, and the page states when the game was last opened and read. The marker describes how the pipeline obtained the figure; it is not the only check a page receives, and its absence does not mean a figure is unreviewed — our editor reviews pages independently of it. See How We Verify.

The prose around those figures — how a game plays, what a bonus round does, the questions and answers — is drafted with AI assistance from the verified data and from observations recorded while the game was played. The model is given the measured numbers and is not permitted to introduce specifications of its own.

We think readers are entitled to know that, which is why it is stated here rather than left to be inferred.

We do not manually review every page before it is published. With thousands of games, each with specifications read directly from the game itself, page-by-page review would be slower than the data changes — and it would not make the numbers more accurate, because the numbers are measured rather than written.

What our team designs and supervises is the system that produces them: how each provider's data is extracted, the order of precedence that decides which source wins when they disagree, and the constraints the language model works under.

That oversight is specific, not nominal:

• Per-provider extraction is monitored. We track which specification fields each provider's extraction actually covers. A provider whose extraction stays incomplete is flagged for a person to work through by hand, and the games are re-processed once it is fixed.
• Consistency is checked automatically. Pages are audited for internal contradictions — a game claiming both a payline count and a ways-to-win count, a reel layout that disagrees with the reel count, an RTP outside a plausible range.
• Published pages are reviewed by hand, continuously. Rather than a fixed audit cycle, review runs alongside publication: pages are opened against the games themselves and checked for the errors an automated test cannot see. Particular attention goes to the captured images — that a paytable screenshot shows the symbols it is captioned with, and that a gameplay frame is the game it claims to be. This review is carried out by our editor.

Games change. Providers ship new versions with different RTP configurations, and a figure that was correct when we read it can stop being correct. We re-open games and re-read them, and the date on each page tells you how fresh that reading is.

If a figure on a page does not match the game, tell us at support@playguidepro.com. Reports are read by a person and checked against the game, and where we are wrong the page is corrected.

This is the part of the process where a reader is genuinely useful to us. An automated pipeline is good at being consistent and bad at noticing that it has been consistently wrong.