For most of the past year, conversations about AI visibility measurement ran into the same wall. Everyone was measuring something. Almost nobody agreed on what.
That changed on August 3, 2026, when the IAB published Measuring Visibility in the AI Era, the first serious attempt at a shared vocabulary for this market. It is better than many people expected. It is also the reason this is the right week to talk about measurement rather than tactics.
Because the standard answers a question CMOs have been asking for a year. It does not yet answer the one they ask about ten seconds later.
The problem the IAB named out loud
The guidance opens with an observation that will sound familiar to anyone who has evaluated an AI visibility tool.
More than 20 companies now sell AI visibility measurement. They use different methodologies. They produce materially different results for the same brand.
That is not a minor inconvenience. It means a marketing team could buy two platforms, scan the same company in the same week, and bring two versions of reality into the same board meeting. When the numbers conflict, the conversation stops being about strategy and starts being about whose tool to believe.
Roughly 16 percent of brands systematically track AI visibility today. Most of this market has not started, which means most of it will start with whatever the standard looks like now.
The four pillars
The IAB’s answer was structure rather than another score. It organizes measurement into four pillars.
Presence. Does the brand appear at all? This covers mention rate, citation rate, share of voice, and visibility momentum across systems like ChatGPT, Claude, Gemini, and Perplexity.
Prominence. Where and how prominently does it appear? Placement, ranking within the response, and how substantively the brand is covered rather than merely named.
Portrayal. In what context, and how accurately? Sentiment, framing, hallucination rate, and factual inaccuracy rate. A brand described confidently but wrongly is in a worse position than one described modestly and correctly.
Persuasion. Does the visibility drive action? Recommendation strength and post-citation click-through.
The pillars are sensible and, more importantly, they are shared. Two vendors can now describe the same thing using the same word.
The more useful contribution
The pillars will get the attention. The tiering is the part that will change how reports get read.
The IAB splits measurement into directional and decision-grade.
Directional measurement suits trend spotting and internal briefings. Fewer than 50 queries, at least two intent types, monthly or quarterly testing.
Decision-grade measurement is what belongs in a budget conversation. Large, diverse query sets covering all four intent categories, weekly or more frequent testing, and reproducibility with defined variation ranges and stated confidence levels.
And one line in the document does more work than the rest combined: single-response measurement is not measurement.
If a vendor hands over a screenshot, that is an anecdote. If a vendor hands over a number with no range around it, that is false precision. AI answers are probabilistic. They shift by prompt, by session, by platform, and by model update. A single value implies a stability that does not exist in these systems.
What the four pillars do not ask
Read the pillars again and notice what each one describes.
Presence: did you appear. Prominence: where. Portrayal: how were you described. Persuasion: did it move anyone.
Every one of those measures an event. A response happened, and something was true about that response.
None of them asks whether it still happens next week.
That is not a criticism of the framework so much as a boundary of it. And it is the question that separates an interesting report from a report worth acting on.
Change the prompt slightly. Add a constraint. Change the company size in the buying scenario. Ask a follow up. Run it on Perplexity instead of ChatGPT. Run it after the next model update. Are you still there?
Why that question changes the decision
The reason persistence matters is structural rather than philosophical.
Presence, prominence, portrayal, and persuasion can all be true on Monday and false on Friday. A brand can look excellent in one scan and disappear from the same query a week later because a model updated, or because the evidence supporting the recommendation was thin enough that one new comparison article displaced it.
One recommendation is an event. Repeated recommendation is a position.
You cannot build a budget on an event. You can build one on a position.
This is also where the IAB’s own logic points. If single-response measurement is not measurement, the remedy is not only more queries. It is the same queries repeated over time, so a team can tell whether what they measured was a fluke or a foundation.
A brand appearing in eight out of ten runs of the same buying question has something real. A brand that appeared once and got screenshotted has a screenshot.
What to ask for this week
Pull the last AI visibility report your team produced and run it against five questions.
Does it tell you whether you appear in the specific buying conversations that matter, not just the broad category term? Does it tell you where you appear relative to competitors? Does it tell you how you were described and whether anything was wrong? Does it distinguish confident recommendation language from hedged mention language? And does it tell you whether any of that held across prompts, platforms, sessions, and weeks?
Then ask one question about the report itself. Is this directional or decision-grade? How many queries, across how many intent types, run how often, with what variation range?
If the report answers one or two of the five, that is visibility measurement. If it answers four, it is approaching recommendation intelligence. If it answers all five and explains what changed and why, AI stops being a marketing metric and starts being acquisition intelligence.
The standard arriving is good news. Making it meaningful is the work that follows.
FAQ
What is the IAB AI visibility measurement standard?
It is guidance titled Measuring Visibility in the AI Era, published by the IAB on August 3, 2026. It provides a shared vocabulary, quality criteria, and disclosure requirements for measuring how brands appear inside AI-generated answers. It was created because more than 20 vendors were producing conflicting results for the same brand.
What are the four pillars of AI visibility measurement?
The four pillars are Presence, Prominence, Portrayal, and Persuasion. Presence asks whether a brand appears, Prominence asks where and how significantly, Portrayal asks about context and accuracy, and Persuasion asks whether the visibility drove action. Together they describe what happened inside a given AI response.
What is the difference between directional and decision-grade measurement?
Directional measurement uses smaller query sets and less frequent testing, and is appropriate for trend spotting and internal briefings. Decision-grade measurement requires large diverse query sets across all intent categories, weekly or more frequent testing, and reproducibility with stated variation ranges. Only decision-grade data should inform budget allocation.
Why is a single AI visibility score not enough for a CMO?
AI systems are probabilistic, so the same question can produce different answers by prompt, session, platform, and model version. A single number hides that variability and implies a stability that does not exist. The IAB guidance states plainly that single-response measurement is not measurement.
What should a CMO add to the four pillars?
Persistence, meaning whether a result holds across prompt variations, sessions, platforms, and model updates. All four pillars measure a single moment in time. Persistence is what separates a one-time appearance from a durable competitive position.
How do I know if my AI visibility report is trustworthy?
Ask your provider how many queries were run, across how many intent types, how frequently, and with what variation range. A provider who can answer those questions is offering measurement. A provider who cannot is offering an impression with a number attached.
Start here: axissuite.ai
Axis Suite by TrendAxis