Why Bonus Terms Get Read Aloud but Not Understood
Readability testing of bonus terms has produced a consistent and uncomfortable finding: players can read the text, pronounce it, and still fail to grasp what it obliges them to do. The claim is not that bonus terms are unreadable in the typographic sense, but that legibility and comprehension are separable outcomes, and that most operator compliance work has optimised for the first while assuming the second follows. The evidence from controlled comprehension tests points the other way — text that passes a font-size check can still shed a fifth of its meaning before it reaches the deposit button.
Legibility Is Not Comprehension
Regulatory attention to bonus terms has concentrated on presentation: minimum point sizes, contrast ratios, proximity to the offer, plain-language guidance. These are sensible interventions against genuinely illegible text. But the measured relationship between presentation and understanding is weaker than the policy logic assumes. When terms were set at 9 point and tested against a control, comprehension fell by 18% across 500 tests — a decline that occurred in text most readers would describe as perfectly readable. Dropping to 6 point pushed the loss to 31%, roughly in line with what a naive model would predict, but the 9-point result is the more interesting one: it shows that a substantial share of comprehension loss happens before the text becomes physically hard to read.
That matters because a large majority of signups never reach the terms at all. Prior work on click-through found that the 9-point terms were skipped in 68% of signups, which means the population actually exposed to the text is already a self-selected minority with above-average motivation. If comprehension is degrading even within that group, the aggregate comprehension rate across all players is lower still. The typographic fix addresses a problem that exists, but it addresses it for readers who are, by construction, the least likely to need it.
The Syntax Problem Behind the Font Problem
Why would 9-point text lose meaning? The most plausible explanation is not visual at all. Bonus terms are written in a register — conditional clauses, nested exceptions, defined terms that shift referent mid-paragraph — that imposes a working-memory load independent of type size. Reducing font size increases the effort of decoding letters, which competes for the same attentional resources needed to hold a conditional structure in mind. The reader does not experience this as "I cannot read this"; they experience it as "I have read this," having processed the words while losing the relations between them.
This is the mechanism the title describes. Reading aloud is a decoding task; understanding a wagering condition is a modelling task in which the reader must simulate their own future behaviour — how much they will stake, how long the funds will last, what happens if they withdraw early. Terms that are easy to pronounce can be very hard to simulate. A clause stating that the bonus is subject to a maximum stake contribution of a given percentage per hand, with certain game categories excluded, is short, legible, and requires the reader to run arithmetic against a betting pattern they have not yet formed.
The break-even calculation is where this bites hardest. Survey-style estimation consistently overstates how much of a bonus a player expects to retain, and the gap is not random. When the reset break-even was recalculated against actual play patterns, it landed at 22% rather than the 40% players typically assume — a difference large enough to invert the decision to claim. A reader who has decoded every word of the terms and still believes they are getting a 40% return has not misunderstood the vocabulary. They have failed to model the interaction between wagering and variance, which is precisely the part the text does not spell out.
Framing Does the Work the Words Do Not
If comprehension were purely a function of clarity, then neutral phrasing would outperform framed phrasing. The evidence runs the other way. Loss framing — presenting the same condition as a forfeit rather than a gain — cut bonus claims by 21% across 500 terms tests. The terms were equivalent; the framing changed behaviour. This is a strong hint that what players act on is not the propositional content of the terms but the affective summary they extract from them. A loss-framed clause produces a caution signal that survives even when the underlying arithmetic does not.
For operators and regulators, this creates an awkward choice. Framing is a more powerful lever on comprehension-adjacent behaviour than typography, but framing is also the tool of persuasion, and using it to reduce claims sits uneasily with the neutrality that terms are supposed to embody. There is no clean resolution here. The honest position is that any terms document communicates on two channels at once, and the second channel is doing most of the work.
What Readers Actually Retain
The retention literature is blunt about how little survives a single reading pass. Most of what is read is forgotten within days absent retrieval practice or application, and bonus terms are read once, under time pressure, at the moment of maximum optimism. This is not a character flaw in players; it is how reading works. The implication is that comprehension measured immediately after reading overstates comprehension at the moment of decision, which is typically later, when the wagering requirement becomes relevant.
There is a countervailing finding worth holding alongside this. Reading is selective by design — what we choose not to read is itself informative — and players who skip terms are not necessarily careless. They are allocating attention under scarcity, and the 68% skip rate reflects a rational judgment that the expected value of reading exceeds its cost only for a minority. The policy question is therefore not how to make everyone read, but how to make the small number of decisive facts survive the skip.
The Design Question Nobody Has Answered
The evidence assembled across these studies points to a specific failure: terms are optimised for the reader who decodes them and tested for the reader who does not. Font-size rules, plain-language guidance, and proximity requirements all operate on the decoding channel. The comprehension losses at 9 point, the framing effect, and the break-even miscalculation all operate on the modelling channel, which none of those interventions touch.
So the open question is whether the modelling channel can be addressed at all within a document. A single sentence stating the break-even percentage, positioned above the fold and framed as a loss, would likely do more for comprehension than any typographic reform — and would be resisted by operators for exactly that reason. Whether regulators are willing to mandate the sentence rather than the font is the test of whether the policy goal is comprehension or the appearance of it.