Same Brand, Different Scores: How to Read AI Visibility Measurement With the IAB's Four Ps
Marketing teams have started putting a new number in the monthly report: how often the brand shows up in answers from ChatGPT, Gemini and other AI tools. The trouble is that the number changes when the tool changes. On August 3, 2026, the Interactive Advertising Bureau (IAB) published a 36-page guide, Measuring Visibility in the AI Era, which sets out a shared vocabulary and quality criteria for AI visibility measurement for the first time. It does not tell you which tool to buy. It teaches you how to read the number in front of you and whether that number is fit to base a decision on.
Why AI visibility measurement is so messy: 20-plus tools, no agreement
According to PPC Land's coverage of the framework, more than 20 vendors now measure AI visibility with different methods, and the same brand can get materially different results. At the same time, only 16% of brands track their AI visibility systematically.
There are structural reasons the numbers do not line up, and the research cited in the framework spells them out:
- Platforms rarely cite the same sources: a Similarweb analysis from November 2025 found only about 11% overlap in citations between major AI platforms, and roughly half of the cited sources change each month.
- Few brands appear consistently everywhere: of more than 1,200 brands Semrush tracked in June 2026, only 36 had consistent visibility across every platform.
- Different questions give different results: the same brand can perform very differently on recommendation questions and on comparison questions.
In other words, different numbers do not necessarily mean someone is wrong. They often mean people are not measuring the same thing.
The IAB's four Ps: presence, prominence, portrayal, persuasion
The IAB arranges the metrics in four levels, from shallow to deep:
- Presence: are you in the answer at all? This covers mention rate, the rate at which you are cited as a source, share of voice within your category, and how these change over time.
- Prominence: where do you appear? First mention, or last in a long list.
- Portrayal: how does the AI describe you? Positive or negative, market leader or budget alternative, and whether the description is accurate.
- Persuasion: does the AI actively recommend you or mention you in passing? Do people then click through to your site?
The third level deserves the most attention from brands. The framework separates hallucination rate (the AI invents something about the brand) from factual inaccuracy rate (the brand is genuinely referenced, but the specs, price or features are wrong). Both must be reported per platform and must not be filtered out by the tool. To measure factual inaccuracy, the brand has to supply a source of truth to compare against. The framework cites an NP Digital survey from February 2026 in which 47.1% of marketers said they run into AI inaccuracies every week.
This echoes the point we made in Why Being Cited Is the New Ranking in AI Search: being mentioned is only the start, and how you are described is what counts.
Directional or decision-grade: can this number move a budget?
The most practical part of the framework is that it splits data into two tiers:
Directional data:
- Suitable for: trends, internal briefings, competitive awareness.
- Question types: at least two of the four intent types.
- Testing cadence: monthly or quarterly.
- Platform coverage: a single platform is acceptable if disclosed.
- Methodology: summary level.
Decision-grade data:
- Suitable for: budget allocation, vendor selection, executive reporting.
- Question types: all four intent types, with results available by intent.
- Testing cadence: weekly or more often.
- Platform coverage: platforms covering most AI usage in the target market, reported separately.
- Methodology: full disclosure.
The four intent types are informational, comparison, recommendation and transactional questions. The framework also sets a floor: a query set of fewer than 50 questions is exploratory only and does not even qualify as directional.
For most small and mid-sized brands, directional data is enough for now. The key is not to use it as if it were decision-grade, for example by rushing to shift budget because share of voice slipped two points.
Five questions to ask before buying an AI visibility tool
The framework lists what vendors should disclose. Condensed into five questions, you can ask them at the next demo:
- Which platforms and model versions do you measure? Are results reported per platform, or only as one combined score?
- How many questions are in the query set, and where do they come from? Written in-house, based on real search volume, or drawn from observed user behavior? What is the mix of intent types?
- How is the data collected? Simulated queries, an observed user panel, or data from the platforms? Is live web retrieval on? Results from different methods cannot simply be added together.
- How are wrong mentions handled? Are hallucinations, factual errors and accurate mentions reported separately?
- What happens to historical data after a model update? Are re-baselining dates and reasons recorded? Is the data before and after shown as two series, or stitched into one line?
A vendor that cannot answer is not necessarily selling a bad tool. It does mean the number is directional at best for now.
Conclusion: AI does the measuring, people decide which questions matter
The IAB framework covers organic, non-paid visibility only. Paid placements and attribution are left for later documents. The standard is still taking shape, so there is no need to chase a precise score yet.
Two things are worth doing first, and both need people. One: list the 30 to 50 questions your customers actually ask before they decide. That query set matters more than any tool. Two: write down your brand's source of truth, covering product specs, prices, service scope and the points people often get wrong. A tool can run every question for you each week, but only you know what to ask and what counts as right.
TitApeX helps brands build these two foundation documents and carry them through to website content and brand strategy. If you want to see how your brand looks inside AI answers, talk to us.
Comments