Shoppers naming an AI assistant grew 4.8x in eight months

What 3,303,402 shoppers said about how they found a brand, measured on completed orders across 3,237 stores, and compared with the source each store’s own analytics recorded for the same order.

Responses

3,303,402

Stores

3,237

Orders with no tracked source

51.8%

Of those, said “don’t know”

8.3%

AI-assisted discovery

The share of shoppers naming an AI assistant when asked how they found a brand rose from 0.23% in December 2025 to 1.10% in August 2026 — a 4.8x increase in eight months. It remains a small channel. No other channel in this dataset moved at a comparable rate.

Complete calendar months only. The data window runs 18 September 2025 to 18 September 2026, so September 2025 and September 2026 are partial months and are left out of every time series rather than compared as if they were whole ones. October 2025 and November 2025 are excluded because fewer than 50 stores contributed.

How often analytics loses each channel

For every channel a customer named, the share of those purchases that reached the store’s analytics with no usable source recorded. This is a rate within each channel, not a share of all orders, so it does not depend on how common the channel is.

By vertical

The same comparison run separately for each retail vertical, with its own channel table. A vertical appears only where at least 50 merchants contributed, so some exist in the data and are not published.

Method

Every figure is drawn from responses to “how did you hear about us” questions on stores using Zigpoll, collected between 18 September 2025 and 18 September 2026. Only responses attached to a completed order are included, so that what a customer said can be compared with the source recorded for that same order.

Answers are grouped into channels by an open, rule-based mapping. Answers that cannot be confidently classified are reported separately rather than assigned to a category; 87.3% of responses were classified. No figure is published for any channel or month with fewer than 50 contributing stores.

Stores using Zigpoll are not a random sample of ecommerce. These figures describe that population, not the market as a whole. Shares are of all classified stated answers. Order values are recorded in their original presentment currency and are not exchange-rate adjusted, which is why no revenue figures are published here.

Citing this

These figures are free to reproduce with attribution (CC BY 4.0).

Zigpoll Attribution Index, September 2026. 3,303,402 post-purchase survey responses across 3,237 ecommerce stores. https://www.zigpoll.com/attribution-index

For the underlying tables, a breakdown by a channel not listed here, or questions about method, contact [email protected].

Questions or Feedback?

We are always ready to hear from you.