What zero-click and dark traffic mean
These are industry terms, and neither Google nor the AI companies define them. Zero-click describes a search that ends on the results page, with no visit to any site. Dark traffic describes a visit whose source your analytics cannot see, so it files the visit under direct.
AI answers add a third case. A buyer reads an answer that names you, closes the tab and phones you a week later. Nothing in your reports links that call to the answer. Google says people now ask longer questions and follow-ups inside AI Mode,[6] so more of the research happens before any site is opened.
What the data shows
Pew Research Center followed the browsing of 900 US adults in March 2025. On Google results pages with an AI summary, people clicked a traditional result link in 8% of visits, against 15% on pages with none. They clicked a link inside the summary in 1% of visits. They ended their session on 26% of pages with a summary and 16% without.[1]
Google gives a different view. It says clicks from AI Overviews are higher quality.[2] Treat that as Google’s own claim. Both can hold at once: fewer clicks, from buyers who are further along.
One study of one engine in one month is not your market. Your own trend is better evidence than any average.
What your analytics can and cannot see
- Search Console. Google says sites that appear in its AI features are included in the overall search traffic, and that it shows in the Performance report under the Web search type.[2] The page describes no separate AI report, so treat it as a combined figure.
- Google Analytics. Google suggests tracking conversions and time on site in tools such as Google Analytics.[2]
- ChatGPT referrals. OpenAI says ChatGPT adds utm_source=chatgpt.com to referral URLs, and that publishers who allow OAI-SearchBot can track that traffic in an analytics platform.[3] Filter your reports by that source.
- Fetcher visits. OpenAI names ChatGPT-User for requests that people make inside ChatGPT,[4] and Perplexity names Perplexity-User for page fetches that follow a person’s question.[5] A script-based analytics tool may not record these. Your server logs can show them by name, so ask your technical team to check.
None of these sees the buyer who reads your name in an answer and leaves. That buyer is the dark traffic, and the tags cannot reach them.
A five-measure scorecard for the board
This is our suggestion, not an industry standard. No single number tells the story, so read the five together.
- Share of voice. Take 10 to 15 buyer questions. Count the answers that name you, and the answers that name each competitor. Repeat each month.
- Branded search. Watch the searches that include your name in the Search Console Performance report. A rise can follow an answer that names you. It is a signal, not proof.
- Direct visits and calls. Track them beside share of voice. If both rise together, the answers may be working.
- AI referrals. Count visits tagged chatgpt.com and any other AI source in your reports. Treat the total as a floor.
- Stated source. Add a “how did you hear about us” line to enquiry forms and call scripts, with AI assistant as a choice.
Share of voice is the one that analytics cannot give you, because it measures who is named, and no visit is involved.
How Vyzpro helps
A Vyzpro audit sets the baseline for the first measure. It runs 10 to 15 buyer-style searches for your market, shows which sites appear for each and shows where your two closest competitors are named instead of you. AI search readiness carries 25% of the overall score. Search results stand in for AI answers unless we ran a live test, and every finding says which one it is.
Monitor repeats the review every month for 12 months and shows what changed. Vyzpro uses public information only and needs no logins, so it cannot see your analytics, count your visits or tie anything to revenue. It covers the part your analytics cannot see: whether the tools name you. See the AI search audit, look inside our report or check pricing. Related guides: the AI shortlist and what is SEO.