How many customers find brands through ChatGPT and AI assistants?
Published by Zigpoll. Updated
About 0.2-5% of purchases are directly attributed to an AI assistant on typical post-purchase attribution questions today, but a much larger share of shoppers say they use AI during research, the gap comes from measurement and buyer behaviour. Zigpoll’s post-purchase answers show AI named by 1.10% of shoppers in August 2026 (up from 0.23% in December 2025), while consumer surveys and discovery studies report far higher rates of AI usage for research (ranges vary by survey and question). (zigpoll.com)
How many customers now find brands through ChatGPT and other AI assistants?
Roughly 0.2-1.1% of orders name an AI assistant in post-purchase “how did you hear” answers, with wider survey estimates for AI use in discovery from about 14% up to 52% depending on the question and sample. Post-purchase attribution data from stores using Zigpoll shows shoppers naming an AI assistant rose from 0.23% (Dec 2025) to 1.10% (Aug 2026). That captures explicit, self-reported attribution at purchase. Broader consumer research asking whether people use AI tools to research brands finds higher shares, for example one industry report put AI discovery at 14.8% and a local discovery study found 52% of respondents used AI tools to research businesses. The difference reflects measurement (self-reported purchase attribution versus “I use AI to research”), plus shoppers checking brands elsewhere before buying. (zigpoll.com) What would change this answer: answers vary by vertical, country and how you ask (single-choice vs multi-select). Post-purchase surveys reflect eventual purchase attribution; behaviour surveys reflect research habits and will always be higher.
Is ChatGPT a major product discovery channel?
Not yet for most brands, it is an emerging discovery touchpoint rather than a dominant channel. Many consumers consult ChatGPT or similar assistants while researching, but explicit purchase attributions to AI remain small in transaction surveys, and many people verify AI recommendations through search, reviews or brand sites first. Surveys show heavy month-to-month use of ChatGPT among AI tool users, but discovery impact depends on which users, categories and prompts are measured. If your category is local services, travel or niche consumer goods, AI recommendations can play a larger role sooner; for high-trust purchases or regulated categories shoppers mostly verify recommendations before buying. (semrush.com)
How can I measure AI-assistant-driven discovery?
Use multiple methods: direct post-purchase surveys, instrumented landing pages/UTMs, server-side logging, and visibility monitoring for AI answer appearances (AEO). Each method finds different parts of the funnel and has tradeoffs.
| Option | What it catches | Strength | Weakness / when it fails |
|---|---|---|---|
| Post-purchase “how did you hear” surveys | Purchaser’s stated last known touch | Direct, simple, catches dark social and voice referrals | Depends on recall and question wording; small AI rates in purchase answers. (zigpoll.com) |
| UTMs + landing analytics | Clicks from links provided by assistants or downstream pages | Precise click-level attribution | Misses recommendations where users re-search or don’t click; lost to in-app browsers and privacy features. (business.adobe.com) |
| Server-side / first-party logging | Referrer, session paths, API logs (when assistants use site links) | Harder to block; more control | Requires engineering; still misses non-click, voice or text-only referrals. |
| AEO / visibility monitoring | Whether your content is cited by AI answers | Shows presence in assistant outputs | Visibility ≠ clicks or purchases; needs manual interpretation. (business.adobe.com) |
Zigpoll’s data also highlights a large tracking gap: across stores using its surveys, 51.8% of orders had no tracked source and many channels’ purchases arrived in analytics with no usable source (for AI assistant purchases that figure was 50.3%). That means measurement must accept and compensate for large “unknown” shares. (zigpoll.com)
How do I capture AI attribution in post-purchase surveys?
Ask one clear, short question plus two quick follow-ups. Example:
- “How did you first hear about us?” (single best answer), include explicit choices: “AI assistant (ChatGPT, Google Gemini, Microsoft Copilot, Siri, Claude, other)”, “Search”, “TikTok”, “Friend/family”, “Ad”, “Other (please specify)”.
- If “AI assistant”: “Which assistant?” (free text / pick list).
- “Did you click a link from the assistant, or did you search or visit us later?” (click / searched later / don’t remember).
These reduce false positives (people who say “ChatGPT” but actually discovered you via a link or search) and capture verification steps users take, most shoppers still verify AI suggestions before buying. Use a short single-question flow on post-purchase pages or email; tools that do post-purchase attribution surveys (including Zigpoll) can automate collection and push results to analytics and CRMs. Always pilot wording for your audience. (morningstar.com)
What would change this advice: if your product has very short purchase cycles (impulse buys) or you sell only in closed ecosystems, UTMs and server logs may work better than surveys; if you prioritize understanding long, multi-step journeys, combine surveys with session stitching and first-party data.
