TL;DR: AI assistants send less referral traffic than traditional search, and some of the traffic they influence gets counted somewhere else. Report how often your brand is named on buying questions, and keep referral traffic as a secondary number.
The short answer
AI assistants (ChatGPT, Claude, Gemini, Google AI Overviews) send less referral traffic than traditional search. They sit at a different stage of the buying journey.
Say somebody asks ChatGPT which invoicing tool to use and gets back three names with a short explanation for each. They probably won't click anything. They're more likely to:
- Open a new tab and search one of the brands directly
- Ask follow-up questions
- Close the session and come back later
None of these count as ChatGPT-referred traffic. Someone clicking a blue link for "alternatives to [option A]" is already in buying mode, so Google tends to get the last click.
Why attribution undercounts
Some assistants pass a referrer and some don't. A brand search that started with an AI recommendation looks the same in your analytics as any other brand search. It also shows up as direct traffic when someone types your address after seeing your name.
So the referral number in your analytics is a floor. The real influence is bigger, and you can't measure it exactly.
What to report instead
- Presence rate. On your tracked buying questions, how often you're named. "Named in 7 of 22 questions, up from 4" is easy to repeat each month, and it maps to being considered.
- Share against named competitors. Who gets named when you don't. That tells you who you're losing to and what to look at next.
- Sentiment and description. How the answer describes you. Being named badly is its own problem.
- Branded search volume. The best proxy for AI-influenced demand that doesn't show up as referral traffic. If AI recommendations are working, branded search and direct traffic should rise over time.
- Referral traffic. Include it as a secondary number and label it as undercounted.
If your name is one of the three a buyer sees, you get considered. If it's missing, traffic from other channels doesn't put you back on that shortlist.
When waiting makes sense
Waiting is reasonable if your buyers don't use AI assistants. That happens in some markets and with some demographics. Ask your customers or run a short survey to find out.
Most of the work also overlaps with good SEO (crawlable pages, content that answers real questions, a reputation outside your own site), so there's very little downside to doing it anyway.
If you're deciding how long to wait, Peak Answer's research measured the entry bar across 14 markets. The entry bar is how small a cited site can be and still get quoted, and it ranged from 15 to 56 out of 100.
What to tell your boss
Here's a short version for someone who only asks about traffic:
AI assistants send fewer clicks than Google. They also shape which brands people look up afterward, and that shows up as branded search and direct visits. I'm tracking how often the company is named on buying questions, who gets named when it isn't, and branded search over time. Referral traffic is in the report too, with a note that it's undercounted.
FAQs
AI traffic is in the low single digits, as a share of total traffic, for most sites today. There's no standard industry benchmark yet though.
AI traffic won't replace search traffic any time soon. It's still important to employ strategies that work for both traditional search and AI search because both are still important.
Google Analytics shows some AI referral traffic by referrer. Search Console shows AI Overview impressions. Both can undercount though so be aware.
Whether it's worth investing in AI search while the traffic is tiny depends on your market or niche. One real benefit of investing in SEO and GEO is that it compounds over time if done properly, and the strategies used for traditional search overlap quite a bit with those used for AI visibility as well.