NIQ and Similarweb Partner on AI Purchase Measurement, Agentic Commerce Measurement Due Q4 2026
NIQ and Similarweb are building a solution that links AI-driven product discovery to verified sales. A look at the five measurement areas, why conventional attribution breaks down, and what merchants can start on today.
Key Takeaways
- On September 2, 2026, NIQ (NYSE: NIQ) and Similarweb (NYSE: SMWB) announced a collaboration on Agentic Commerce Measurement, a solution that connects AI-driven product discovery to actual sales. An initial version is planned for Q4 2026, starting with a focused set of categories and markets
- The solution covers five areas: Consumer Intent, Agentic Shelf Visibility, Product Content Readiness, AI-Driven Traffic and AI-Driven Conversion, with planned coverage across ChatGPT, Gemini, Google AI Mode, Perplexity and Claude
- The point is to close an attribution gap. Similarweb's own research found that only 8.8% of visits to AI-recommended brands arrived through direct AI links, while 55.9% arrived later as branded search
A declaration: AI-driven sales will be measured
NIQ and Similarweb help brands measure AI-driven discovery, traffic, conversion and sales in the agentic commerce era.
nielseniq.comIn a press release datelined Chicago on September 2, 2026, consumer intelligence company NIQ disclosed a collaboration with digital data company Similarweb. The theme is singular: make it possible to measure how much revenue an AI recommendation actually produced.
Troy Treangen, Chief AI and Product Officer at NIQ, put it plainly.
AI is becoming a new commerce channel, and NIQ intends to make it measurable. Our clients want to know where AI is already influencing their business, how quickly that influence is growing and what they should do about it.
The release rests on a premise: AI agents have moved past discovery and into the purchase itself. With new infrastructure such as Google's Universal Commerce Protocol (UCP) and OpenAI's Agentic Commerce Protocol (ACP), the journey from discovery through evaluation and recommendation to purchase can now complete inside a single AI experience. As that path retreats into the AI, it disappears from the brand's view.
On timing, NIQ says an initial version will arrive in Q4 2026 with a deliberately narrow set of categories and markets. Pricing, the specific categories, the launch markets and availability in Japan are all undisclosed.
Five measurement areas that define the problem
The five announced areas double as a list of what merchants currently cannot see.
| Measurement area | What it looks at |
|---|---|
| Consumer Intent | What consumers ask AI assistants, and which questions shape their purchase decisions |
| Agentic Shelf Visibility | Where products appear when AI recommends options, and how that compares with competitors |
| Product Content Readiness | Whether product content is complete and structured enough for AI to understand and recommend accurately |
| AI-Driven Traffic | How much product-page traffic originates from AI platforms, via direct clicks and subsequent visits |
| AI-Driven Conversion | How AI-influenced engagement connects to verified omnichannel purchasing |
The third one is the interesting outlier. Product Content Readiness measures whether product information is structured and legible to AI, which makes it a diagnosis of causes while the other four measure outcomes. Without separating a lack of exposure from a defect in the product data, there is no way to decide what to fix.
NIQ lists five target platforms: ChatGPT, Gemini, Google AI Mode, Perplexity and Claude. The cross-platform design says a lot about the nature of this collaboration.
What conventional analytics is missing
Here is the crux. Why is existing web analytics insufficient?
Research Similarweb published in 2026, covering US desktop web, answers with numbers. Brands recommended by AI saw visit rates 2.5x higher than non-recommended competitors within seven days. But break that traffic down and only 8.8% came through direct links from AI platforms, while 55.9% arrived as branded search, with users typing the brand name themselves (reported by PPC Land).
Consider what that structure implies. A shopper remembers the product the AI suggested, then searches for it later on their own. To an analytics tool, that is organic search or direct traffic. The fact that AI started the journey is recorded nowhere. Implementations that withhold referrer data, such as Google AI Mode, are spreading, so even a click often arrives with an empty referrer.
For most merchants, then, AI's contribution is not showing up as zero; it is being credited to a different channel. In a budget discussion that is worse than undercounting. It pushes teams to spend more on branded search ads while the real cause, unprepared product data, goes untouched.
The same study reported that AI-influenced visitors averaged 12.0 pages and 11.8 minutes on site, roughly double the 6.5 pages and 5.6 minutes of everyone else. The highest-quality traffic is arriving in the form that is hardest to measure.
Why POS data and clickstream data are being joined
The substance of this collaboration is the overlay of two very different datasets.
NIQ brings retail sales measurement and product intelligence spanning more than 90 countries, roughly 82% of the world's population and $7.4 trillion in consumer spend. Similarweb brings digital behavioral data from panels and clickstream: signals from inside generative AI platforms and from the journey that continues after the user leaves them. One knows what sold; the other knows what happened just before.
The pairing sits on the trajectory of NIQ Commerce Lab, which the company launched in April 2026. Commerce Lab set out a measurement layer built from six domains: preference, product, availability, purchase verification, channel measurement and optimization. This collaboration reads as sourcing the thinnest of those pieces, behavior inside AI platforms, from outside the company.
Susan Dunn, Chief Revenue Officer at Similarweb, notes that AI's influence does not stop when a consumer leaves an AI platform. Given that 55.9% figure, that is less a sales line than the technical premise the whole partnership rests on.
Where to discount the announcement
It would be premature to treat this measurement as definitive. At least three reservations apply.
First, the precision of estimated data. Similarweb's traffic figures are extrapolations from a panel, and the company's own data accuracy documentation is explicit about the use of sampling and estimation models. Accuracy tends to be close to reality for large sites, while the company's own documentation acknowledges that error margins widen on low-traffic properties. Whether the granularity will be trustworthy for a small or midsize merchant looking at its own numbers is an open question.
Second, the constraints of the underlying study. The Similarweb research cited above covered US desktop only, compared six brands across three verticals, and used a seven-day attribution window. Mobile behavior was excluded, and ChatGPT's share moved from 76.4% to 52.7% during the study period. These are not numbers to transplant directly onto the Japanese market or onto today's platform landscape.
Third, and most structurally, the AI platforms themselves disclose nothing. This measurement observes from the outside and infers. Which products actually appeared, and in what order, inside ChatGPT or Gemini cannot be confirmed unless the platform publishes it. There is still no first-party reporting window for AI search comparable to Search Console.
The size of the demand is also uncertain. Separate NIQ research found that as of May 2026, 42% of US consumers had used at least one AI tool to shop in the previous month, but only 5% had a fully autonomous AI agent place an order (a monthly survey of about 500 people). Use for discovery has reached 74%, while delegated purchasing is only starting. Merchants should be careful not to conflate which layer is being measured.
What merchants can start on now
There is no need to wait for a vendor solution. Two of the five areas can be addressed with no external data at all.
Structuring product information tops the list. That Product Content Readiness was elevated to a measured metric means, in effect, that defects in product data are starting to be priced as lost AI-driven revenue. Missing attributes, inconsistent naming, stale inventory and pricing, and the state of structured data markup can all be audited today.
The second is a provisional read on AI-driven traffic. A complete picture is out of reach, but simply monitoring branded search queries alongside referrals from AI platforms reveals a trend. If branded search is climbing without any increase in ad spend, suspect AI recommendation as the cause. That level of hypothesis testing runs fine on GA4 and Search Console.
The common failure mode in this field is waiting for the measurement stack to mature while competitors finish cleaning up their product data.
Conclusion
The NIQ and Similarweb collaboration marks AI commerce moving from a topic of conversation to a line item in budget allocation. Budget follows measurement, reliably.
Two things are worth watching: which categories and markets are chosen for the Q4 2026 initial version, and how far AI-Driven Conversion can be backed by real data rather than modeling. If that holds up, AI-driven revenue starts being treated as a channel alongside television and search. If it stays within estimation, it will remain a reference figure for a while longer.
Either way, the day when being findable by AI is judged in numbers is getting closer.
