The phrase "attention economy" was coined in 1997. Herbert Simon used the underlying concept even earlier, in 1971, when he wrote that a wealth of information creates a poverty of attention. Fifty years later, the marketing industry has built an entire vocabulary around attention: attentive seconds, active attention, passive attention, attention quality scores, attention-weighted reach.
What it has not built is a standard way to measure any of it.
This is not for lack of trying. Eye-tracking panels, facial coding, EEG studies, biometric research, panel-based surveys: the industry has thrown significant money at the problem. What it has produced is a collection of methodologies that are expensive, slow, panel-dependent, and structurally incapable of operating at the scale modern brand management requires.
The measurement gap is not a gap in ambition. It is a gap in architecture.
What the existing tools actually measure
The dominant attention measurement approaches fall into three categories, each with a structural limitation that prevents it from becoming a standard.
Eye-tracking panels measure where people look. They are accurate within their sample and useless outside it. You cannot run an eye-tracking study on every piece of content your brand produces across every channel where it appears. You run studies on selected assets and extrapolate. The extrapolation is the problem.
Viewability metrics measure whether an ad was technically visible: was it on screen, for how long, at what percentage of the creative. The IAB and MRC have spent years refining these definitions. Viewability is a necessary condition for attention. It is not a sufficient one. An ad that is on screen for two seconds while the user is looking at their phone has perfect viewability and zero attention value.
Predicted attention models use machine learning to estimate the probability that a piece of content received attention, based on historical data about similar content in similar contexts. These models are improving rapidly. They are still predictions, not measurements. And they are trained on data from the past, which means they are structurally slow to adapt to new environments like AI chat, where the behavioral patterns are fundamentally different from anything in the training corpus.
None of these approaches can tell you what is actually happening with your brand in the environment that is now driving the majority of high-intent discovery: AI-generated responses.
The architecture problem
The measurement gap exists because the dominant tools were built for a media environment that no longer describes how people find information.
Display advertising assumes a page with discrete ad slots. Viewability makes sense in that context. Search advertising assumes a results page with ranked links. Click-through rate makes sense in that context. Social advertising assumes a feed with interruptive placements. Engagement rate makes sense in that context.
AI-generated responses assume none of these things. There is no ad slot. There is no ranked link to click. There is no feed to interrupt. There is a paragraph of text, generated in real time, read by a user who asked a specific question and expects a specific answer. The behavioral patterns are different. The attention mechanics are different. The measurement framework needs to be different.
Building AI attention measurement by adapting display metrics is like measuring a telephone conversation using the tools developed for silent film. The medium changed. The instrument needs to change with it.
What a real measurement architecture looks like
A measurement architecture capable of closing the gap needs four properties.
It needs to be behavioral, not panel-dependent. Behavioral signals, scroll patterns, pause duration, re-read loops, cursor hesitation, interaction depth, are observable at scale without recruiting participants. They reflect what people actually do, not what they report doing in a survey or what a camera records them doing in a lab.
It needs to be continuous, not episodic. Brand attention is not a quarterly study. It is a signal that changes with every model update, every competitor move, every shift in how users phrase their questions. A measurement architecture that produces results every three months is not a measurement architecture. It is a snapshot collection.
It needs to be multi-channel, not single-surface. Attention in AI chat behaves differently from attention on web editorial, which behaves differently from attention in social feeds, which behaves differently from attention in video. A brand that operates across these channels needs a framework that applies consistent methodology across all of them, with surface-specific calibration, not four separate tools with four incompatible outputs.
And it needs to be auditable. A score that cannot be explained to a CFO or a media director is a score that cannot be used to justify a budget. The weighting, the sources, the methodology: all of it needs to be documented, versioned, and verifiable. This is not optional. It is the condition for institutional adoption.
Why AI made this urgent
For the first decade of attention economy discourse, the measurement gap was an inconvenience. Brands managed with proxies. Viewability stood in for attention. Engagement rate stood in for interest. The proxies were imperfect but functional.
AI changed the economics of that compromise.
When the primary channel for high-intent brand discovery was search, an imperfect attention metric still left you with click data. The user clicked or did not click. You could optimize for clicks. The loop was messy but closed.
When the primary channel for high-intent brand discovery is an AI response, there is no click. The user reads, forms an impression, and acts or does not act. The entire value of your brand's presence in that response depends on whether it received attention, at what depth, in what context, weighted by what sources. If you cannot measure that, you cannot optimize for it. And if you cannot optimize for it, you are spending marketing budget in a channel where you have no feedback loop.
That is not a gap you can manage around. It is a structural vulnerability.
What BAX measures and why it is different
The Behavioral Biometrics Engine at the core of BAX is not a panel study and not a prediction model. It is a behavioral measurement system calibrated on more than five million data points across content formats, channels, and platforms.
The BAX Index aggregates four dimensions: visibility, position quality, sentiment weighted by cognitive load, and exposure quality. Each dimension is scored against decay curves specific to the surface being measured. AI chat has its own decay parameters. Web editorial has its own. Social has its own. The methodology is consistent. The calibration is surface-specific.
The output is a single number, 0 to 100, that reflects the real attention value of your brand's presence across the channels BAX monitors. Not predicted attention. Not viewability. Not share of voice. Attention, measured behaviorally, at scale, continuously.
This is what the attention economy has been missing. Not another panel study. Not another prediction model. An instrument.
The standard that is coming
The IAB is now developing normative standards for attention measurement that will apply across channels, including AI environments. When those standards are finalized, the market will divide between platforms that meet them and platforms that do not.
BAX is built to meet them. The methodology is documented. The weighting is auditable. The decay models are versioned and published. The source layer is classified and tracked.
The measurement gap is closing. The question for every brand in the attention economy is the same one it has always been: do you want to be on the right side of the standard when it arrives, or do you want to spend the following two years catching up?
The poverty of attention
Herbert Simon was right in 1971. A wealth of information creates a poverty of attention. What he could not have anticipated is that fifty years later, a machine would be making the first cut: deciding which information the user sees at all, in what order, from what sources, with what framing.
The brands that understand this, that the machine is now the first editor, the first filter, the first recommender, are the brands investing in understanding how attention works inside that machine.
The ones that do not are still optimizing for a media environment that no longer exists.