Two of advertising’s measurement players spent the first week of September making the same admission from opposite directions: nobody can reliably see what happens once a purchase decision starts running through an AI agent instead of a browser. NIQ and Similarweb answered by building a joint measurement product for it. Google answered by rethinking how it counts a conversion at all. Neither company is finished, and that gap between the problem and the fix is the actual story.
The Trail AI Shopping Doesn’t Leave
Traditional shopping measurement runs on digital exhaust: clicks, sessions, referral strings, add-to-cart events. A consumer who asks ChatGPT or Gemini to compare two products and then buys one, inside the chat itself or on a site the assistant recommended, generates almost none of that exhaust. The purchase happened. The trail measurement tools rely on to credit it did not.
That gap is becoming a business problem faster than most measurement stacks can adapt to it, which is why two unrelated fixes surfaced within days of each other, aimed at opposite ends of the same funnel. The commerce side of this problem is already being built out; the protocol connecting Microsoft, Stripe and OpenAI now routes a growing share of AI-agent checkouts without a single click for anyone to measure.
NIQ and Similarweb Bet on Instrumenting the Agent
NIQ and Similarweb announced a collaboration to measure what they are calling agentic commerce: how AI assistants influence discovery, intent, traffic and sales before a brand ever sees a session start. An initial version is due in the fourth quarter of 2026, covering a limited set of categories and markets, with coverage across ChatGPT, Gemini, Google AI Mode, Perplexity and Claude.
“AI is becoming a new commerce channel, and NIQ intends to make it measurable,” said Troy Treangen, Chief AI and Product Officer at NIQ. Similarweb Chief Revenue Officer Susan Dunn framed the same gap from the buyer side: “AI’s influence does not stop when a consumer leaves an AI platform.”
What’s Actually Being Measured
The companies describe five tracked areas: consumer intent (what shoppers ask AI assistants), agentic shelf visibility (where and how products surface in AI recommendations), product content readiness (whether a brand’s data is structured well enough for an AI assistant to use), AI-driven traffic, and AI-driven conversion. Notably absent from the announcement: pricing, named launch markets, a beta program, or a published measurement methodology. This is a bet on the category before the category has been proven out, not a finished instrument.
Google Is Rebuilding the Conversion Window Instead
Google’s answer, laid out on its Ads and Commerce blog the same week, does not try to see inside AI platforms at all. It rethinks how long a marketer is allowed to wait before crediting a conversion. The company introduced Qualified Future Conversions, a two-stage mechanism: a click first needs a qualifying signal within seven days, such as a branded search or an add-to-cart event, and once qualified, Google will count a resulting conversion for up to 180 days afterward, instead of stopping at a fixed lookback window.
John Chen, Senior Director of Product Management for Ads Measurement at Google, argued fixed windows never matched how people actually buy: “We think that’s probably a little bit too arbitrary.” Chen also pushed back on treating any single measurement approach as authoritative, describing attribution, incrementality testing and marketing mix modeling as three separate gauges on one aircraft rather than competing systems: no single instrument should serve as the north star. Alongside the change, Google is widening access to Meridian, its open-source marketing mix model, through a Google Cloud integration and a version built into Analytics 360.
Same Pressure, Two Different Fixes
NIQ, Similarweb and Google did not coordinate these announcements, and they are not solving the same technical problem. But both are responses to the same underlying shift: purchase journeys are stretching out and moving through interfaces that do not behave like a browser tab. Google’s fix stretches the calendar. NIQ and Similarweb’s fix tries to watch a new venue. Marketers who wait for either one to mature before adjusting their own tracking are, in effect, choosing to keep measuring the funnel that is disappearing rather than the one replacing it. It is the same instability that recently pushed Snap to hand attribution math back to advertisers and that a study cited in our own reporting used to show marketing mix models overstating ROAS by 2.5x: every layer of the measurement stack is being questioned at once, not just the AI-facing one.
What This Means for the Marketing Leader
Treat the NIQ-Similarweb product as directional, not authoritative, for at least a full quarter after its Q4 2026 launch. Without a published methodology or independent validation, its early numbers describe a trend, not a budget-shifting fact. Google’s Qualified Future Conversions is easier to act on immediately: it changes how existing Google Ads accounts count conversions, so anyone running long-consideration-cycle campaigns, in B2B software, financial services or big-ticket retail, should audit how much of their reported conversion volume shifts once the 180-day window applies, before drawing conclusions about channel performance from it.
The broader takeaway is Chen’s own framing, applied one level up: no single new metric, whether it is an AI-visibility score or a longer attribution window, should become a marketing organization’s sole gauge. The instruments are all being rebuilt at once precisely because none of them, individually, was built for how people are buying now.
Source: NIQ
