How to Read the FoeGlass Index Correctly

The FoeGlass index is a map of how AI models describe iGaming brands to real players, and reading it correctly means understanding what each cell is actually saying. FoeGlass tracks how ChatGPT, Gemini, Perplexity, Grok, and Google AI Overviews represent brands across geo, language, query intent, and player type. Every cell carries a state, not just a number, and the number only means something once you know the state behind it. Misread a gap as a failure and you draw the wrong conclusion. Here is the mental model.

The four cell states, and what each one means

Every cell in the FoeGlass grid sits in one of four states. The difference between them is the difference between a real problem and a non-event.

• Present: measured, and the brand was named. The cell has a score. This is a real reading of how AI positions the brand for that geo, language, intent, and persona.
• Absent: measured, a market exists there, and the brand was not named. This is an honest zero. The player asked, AI answered, and the deposit went to someone else. Absent is the state that should sting, because demand was there and the brand missed it.
• No market: measured, but there is no real market in that cell. It is excluded from the math entirely. No market is not a zero and not a penalty. It simply does not belong in the denominator.
• Not measured: never measured. This is not data. It is never painted as bad, never scored, never shown as a weakness. A blank here means we have not looked yet, nothing more.

The single most common misreading is treating not measured like absent. They look similar as empty cells, but one is an honest zero where the brand lost a real player, and the other is simply unexplored ground. FoeGlass keeps them visually and mathematically distinct on purpose.

The witness rule: why a red verdict needs proof

A low score only counts as a genuine miss when the market clearly existed at the moment of the query. FoeGlass enforces this with the witness rule. For a run to score a brand as absent, at least one tracked brand must be named confidently and non-negatively in the same AI answer. That named brand is the witness. It proves a player could have received a real recommendation there.

If no witness appears, the run is reclassified as no market and leaves the denominator. This stops the index from punishing a brand for an answer where AI recommended nobody at all. A zero is only fair when someone else got the deposit. The witness rule is what keeps absent cells honest.

The 0 to 100 score, and why money intents weigh more

When a cell is present, its score runs from 0 to 100 as a weighted average of four signals FoeGlass extracts from the AI answer:

• Role: was the brand recommended, merely mentioned, or flagged as a warning?
• Tone: was the framing positive, neutral, or negative?
• Position: where in the answer did the brand appear, first or buried?
• Source quality: what kind of sources did AI cite around the mention?

These four are not weighted equally, and neither are the queries. Money intents, the questions where a player is close to depositing, weigh more than informational ones. Being named first for "where to deposit and play slots for real money" counts for more than a neutral mention in a general "what is a crypto casino" explainer. The exact weights are locked and not disclosed, which keeps the index stable and non-gameable across cycles.

Reading a brand's row without misreading the gaps

Put it together and a brand's row tells a clear story once you separate the states. Green present cells show where AI already recommends the brand. Absent cells show honest zeros: real markets where the deposit went elsewhere, and the first place to focus. No market cells are noise to ignore, since they are excluded from the score. Not measured cells are simply the frontier, not a verdict.

A few practical rules keep your reading accurate:

• Judge the score only on present and absent cells. Those are the measured reality.
• Never read a blank as a failure until you confirm it is absent, not not measured.
• Remember that data is collected manually from the required geos and languages, by real people using prepared persona accounts, so each cell reflects a real player's experience, not a scraped guess.
• Treat published snapshots as fixed. The public view shows the current cycle and the one before it, so movement between them is the signal that matters.

Read this way, the index stops being a wall of numbers and becomes a clear priority list: defend the green, close the honest zeros, and decide what to measure next. If you want that turned into a concrete plan for your brands and geos, a FoeGlass report and ongoing analysis is where that work happens.