# What AI traffic is worth > What purpose AI traffic serves for the organisation, and how much the organisation is willing to pay for it. If the organisation does not decide what the channel is for, the tool supplies an implicit answer. > What purpose does the traffic arriving from AI serve for us, and how much are we willing to pay for it? > Does it still make sense to rely on the traffic metric? Upstream of this question is another — how much do we want AI to bring people towards us? — which is decided in [Staying in AI answers](https://www.shaped-by-ai.com/en/impacts/the-decisions-ai-forces/staying-in-ai-answers/). ## The question What purpose does the traffic arriving from AI serve for us, and how much are we willing to pay for it? The "what purpose" is not decoration: it is the part that makes this a decision and not a measurement. ## Why it is a decision Because the metric the tools return by default is volume, and volume in this channel always goes down. If the organisation does not decide what the channel is for, the tool supplies an implicit answer: *it is going badly.* ## What the literature says * **Volume.** CTR from −15% to −89%; 93% zero-click in AI Mode; Reuters at −33% and −38%; Search Console giving impressions only since 3 June 2026. * **Aggregate benchmark.** The pattern is not sector-specific. Across all Google searches, not just commercial or publishing queries, [SparkToro](https://sparktoro.com/blog/in-2026-less-than-one-third-of-google-searches-still-send-a-click/) measures 68.01% of US searches ending without a click in the first four months of 2026 (Similarweb panel), against 60.45% in 2024 (Datos panel) — down from 374 to 276 clicks to the open web per 1,000 searches. A [follow-up country breakdown](https://sparktoro.com/blog/zero-click-searches-highest-in-the-uk-lowest-in-germany-and-france-has-the-most-efficient-searchers/) puts the UK at 69.5%, the US at 68.0%, France at 65.3%, Canada at 63.8%, Italy at 63.4% and Germany at 62.1% — a four-point spread across markets on the same metric. The measure on real behaviour is in [The conversations you cannot observe](https://www.shaped-by-ai.com/en/impacts/what-is-happening/search-becomes-conversation/the-conversations-you-cannot-observe/), which reports the Pew figures on clicking in the presence of an AI summary. Sector-specific measures remain the operating figure: these numbers place them inside a broader trend. * **Quality.** 0.5% of sessions produce 12.1% of signups: roughly 23 times the average rate. * **Composition.** AI-influenced traffic arrives far less as direct (19.9% against 38.8%) and far more as branded search (55.9% against 40.4%): the user gets the recommendation in chat and then searches for the brand on Google. That means the AI channel does not just bring visits — it shifts the moment when purchase intent forms. * **Volatility.** The presence of AI Overviews on e-commerce queries has dropped from 29% to 3.2%: the channel changes shape without notice. The organisation must assess whether the channel will still be the same channel tomorrow. * **Non-observability.** The *informed non-opener*: an informed user who opens nothing. It is not an absence of interest, it is an absence of trace. * **Downstream behaviour.** 58% of shoppers use AI to inform themselves, 37% start a purchase journey with it, only 17% are comfortable completing it; and in chat the measured conversion rate is a third of that of click-outs, while generating twice as many new customers. The pattern is clear: people discover products in AI and buy on the site. If the channel is there to acquire, the number to watch is not in-chat conversion. * **Sources to cross-reference**, because none is enough on its own: long-tail from Search Console, keywords from Search Ads, AI queries from Bing, internal site-search queries. ## What each choice entails **Choosing to rewrite the metric.** The organisation stops measuring the part of the phenomenon destined to shrink for external reasons, independent of channel performance. The organisation loses the time series and comparability with the other channels, exactly where management asks for them. **Choosing to keep the traffic metric.** The organisation preserves a comparable historical series that is legible to whoever approves the budget. Budget decisions continue to rely on a number that measures the part destined to shrink. **Not deciding.** Volume stays the implicit metric, and the channel gets judged on the easiest number to obtain — even when downstream behaviour shows people discover in AI and buy on the site. ## Criterion for deciding Measure conversion, qualification and traffic composition together with cross-referenced sources — Search Console, Search Ads, Bing and internal search — accepting a declared discontinuity. The metric should be chosen based on the channel's purpose, not the continuity of the time series. Declare coverage before interpreting the result: impressions contain no question, and simulated prompts contain no users. Agentic traffic must be separated from human traffic; task success, completion of delegated operations and failure points describe the system, while mixed sessions and funnels can describe software rather than people. ## Possible evolution Moving to conversion, qualification and traffic composition means accepting **a declared discontinuity**, not finding an equivalent metric. Anyone promising equivalence is reconstructing the old number with worse data. ## Where it falls on roles The choice of metric is not a technical matter: it rewrites the work of whoever produces it and whoever is accountable for it. * **[Operations & Data](https://www.shaped-by-ai.com/en/impacts/roles/operations-data/)** — Owns the decision. Rewriting the metric means redefining channel groups and attribution windows, separating AI referrers, branded search and direct, isolating agentic traffic before measuring anything, and documenting what attribution cannot see. * **[Marketing & Growth](https://www.shaped-by-ai.com/en/impacts/roles/marketing-growth/)** — The object of the work changes: if the channel acquires customers rather than visits, in-chat conversion is no longer the number to watch. * **[SEO, GEO and AI Visibility](https://www.shaped-by-ai.com/en/impacts/roles/seo-ai-visibility/)** — Evidence has to be built by cross-referencing Search Console, Search Ads, Bing AI queries and internal search. * **[Leadership & Strategy](https://www.shaped-by-ai.com/en/impacts/roles/leadership-strategy/)** — Accepting a declared discontinuity is a decision about objectives and budget allocation. ## Who decides Digital / Web Analyst together with whoever owns the channel budget. The decision is about allocation, not measurement — and it has to be argued to the second, not the first. ## Related phenomena * **[The customer journey shortens](https://www.shaped-by-ai.com/en/impacts/what-is-happening/search-becomes-conversation/the-customer-journey-shortens/)** — If awareness and consideration move into chat, fewer visits measure a shift in stage, not a loss of interest. * **[The conversations you cannot observe](https://www.shaped-by-ai.com/en/impacts/what-is-happening/search-becomes-conversation/the-conversations-you-cannot-observe/)** — The informed non-opener is not a tracking gap but a part of the journey that leaves no trace by construction. * **[Intent-based interaction](https://www.shaped-by-ai.com/en/impacts/what-is-happening/search-becomes-conversation/intent-based-interaction/)** — AI traffic is already filtered for intent, which is why volume averages say little. * **[The user expects the system to act](https://www.shaped-by-ai.com/en/impacts/what-is-happening/agentic-ai-from-showing-to-doing/user-expects-system-to-act/)** — Agentic traffic requires segmentation and outcome metrics, not undifferentiated funnels.