There is a traffic source in your analytics growing at a rate nothing else on the list comes close to, and there is a very good chance you have never opened the report.
AI referral traffic to United States retail sites grew 393% year over year in the first quarter of 2026, according to Adobe Analytics data reported by TechCrunch. As of March it was up 269% on twelve months earlier. Adobe's figures cover more than a trillion visits to US retail sites.
The same report has a frustrating number in it. Roughly a quarter of the content on retailers' homepages has not been optimised for AI models, and around 34% of product pages cannot be properly accessed by them. The traffic turned up before the pages were ready.
Small business sites have a plainer version of that problem. A visitor from ChatGPT arrives looking to confirm one specific thing the assistant told them, and the page opens with something general. Put that fact in plain text at the top and these visitors become some of the easiest sales you will make.
Who these people are
Someone arriving from a Google ad is early. They clicked a promise and are still deciding whether the problem is worth solving.
Someone arriving from an AI assistant has already had a conversation. They described their situation, got options, asked follow ups and narrowed the field. Then they chose to come to you.
The numbers reflect it. In Adobe's retail data, AI referred visitors converted 42% better than other visitors in March 2026 and spent 48% longer on site. A year earlier they converted 38% worse. We cover that gap in more detail in why traffic from AI converts better and almost nobody is tracking it.
By the time they land, they know what you do, roughly what you charge and why they were sent. All that is left is to check it.
An AI referral is a warm introduction from a source your prospect already trusts. Most websites answer it with a hero banner and a slideshow.
Why they bounce anyway
The assistant told them something specific. Your page opens with something general.
Say the assistant recommended you because you do same week commercial callouts across the Central Coast. Your visitor arrives wanting confirmation of that claim, and lands on a homepage that says "Excellence in service since 2011" above a photo of a handshake.
Now they have to hunt for the fact that brought them. Plenty will not bother, because the assistant offered two other options in the same breath and one of those might confirm faster.
A second failure hides underneath. Sometimes the assistant's description of you is out of date, because what it found about you was thin or inconsistent. The visitor arrives expecting a price or a service area you do not offer, and leaves disappointed. Fixing your page helps, but so does fixing what the wider web says about you. Our guide to what AI is telling your customers about you walks through how to check.
Building the page that closes them
Three changes do most of the work.
Lead with the specific claim, in plain text, above the fold. Skip the headline about your values and state the fact. For the callout business above, that is "Same week commercial callouts across the Central Coast." Whatever an assistant would use to justify recommending you goes in the first thing a human reads.
Answer the follow up questions on the page. The prospect has already asked the assistant three or four clarifying questions, and they arrive with a fifth. Write out the questions your customers ask you and give each a full answer. An accordion of one liners written for search engines in 2018 will not do it.
Make the next step tiny and obvious. These visitors are close and do not need nurturing. Give them a phone number, a form with three fields and a clear statement of what happens after they submit it.
Plain text matters more than it sounds. A fact that only exists inside a slider image or a video is harder for an assistant to read and slower for a human to find. The sentence that confirms the claim for a visitor is also the sentence a model can quote next time, so one edit serves both audiences.
How to see this traffic
This used to be the step where people gave up. It is easier now.
Google Analytics 4 has a default channel called AI Assistant, for users who arrive "from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok". It only works when the click carries a referrer GA4 recognises. A visit with no referrer drops into Direct, the least informative bucket in your analytics, so treat the channel as a floor.
Three ways to fill in the picture, in order of effort.
- Check the AI Assistant channel, then your referral list. Open the traffic acquisition report and look for the channel. Then search your referrals for hostnames such as chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com in case any were missed.
- Watch how direct traffic behaves. Segment direct visitors who land on a deep page instead of the homepage. Nobody types a service page URL from memory, so some of that segment was sent by an assistant or a link in a private message.
- Ask. Add one optional field to your enquiry form: "How did you hear about us?" If people start typing some version of "ChatGPT said", you have your answer without buying an attribution tool.
Once you can see the channel, put it in your monthly report next to organic search and paid.
The compounding part
We think this is worth acting on now instead of next year, and the reason is reviews.
A business that gets named, then confirms the recommendation with a page that closes cleanly, wins the job. Satisfied buyers write reviews. Those reviews become more material for an assistant to read the next time somebody asks the same question.
A business that stays ambiguous gives the assistant nothing new to work with, and the gap widens each quarter.
What to watch for on site
Pull up a few session recordings from this segment and check them against what you would expect from someone who has already been briefed.
Our expectation is scanning, not reading. A visitor like this wants one or two specific facts, usually price, coverage or timeframe, and has little use for your navigation.
So design for scanning. Put facts near the top, in plain text, in the order an assistant would have mentioned them. Save the persuasion for people who arrived without a briefing.
A worked example: the physio clinic's hidden channel
This one is hypothetical. Say you run a physiotherapy clinic in Geelong. Your analytics show direct traffic creeping up and a handful of visits from chatgpt.com each month. Enquiries haven't moved, so you've never looked closer.
You build a segment of direct visitors who landed on deep pages and find most of them arriving on your "sports injury rehab" page. In the session recordings they scroll straight past the hero image, hunt for something and leave within seconds.
Next you ask a few assistants who they'd recommend for sports injury rehab in Geelong. You get named alongside two competitors, and the reason given is Saturday appointments and direct health fund claiming. Your rehab page mentions neither on the first screen. Saturday hours live in the footer and health fund claiming is on a separate FAQ page.
So you rewrite the top of the page: "Sports injury rehab in Geelong, with Saturday appointments and on-the-spot health fund claiming. Book online in under a minute." Below that, you answer the questions people ask at reception, and the booking button sits right there.
Your marketing spend is unchanged. You stopped making pre-sold visitors hunt for the reason they came. The details are made up, but most fixes in this channel have that shape.
Mistakes that waste AI referrals
- Sending every visitor through the homepage journey. Deep pages are where these visitors land, so deep pages need to stand on their own.
- Hiding prices entirely. If an assistant mentioned a price range, a page with no pricing signal at all looks evasive. A starting price or a range is often enough.
- Inconsistent facts across the web. Different hours on your site and your Google Business Profile confuse assistants and the people they send.
- Long forms. A ten-field form is a speed bump for someone who has already decided.
- Judging the channel on raw volume. This traffic is small and valuable, so count enquiries and sales before sessions. For the broader picture of pages that sell around the clock, see why your website should be your best salesperson.
What to do this week
Open your analytics. Find every session in the AI Assistant channel and every direct session that landed on a deep page. Look at what those visitors did.
Then take your top three landing pages and rewrite the first fifty words of each so they confirm, in plain language, the specific fact that would have caused an assistant to send someone there.
That is an afternoon of work on a channel that grew 393% in a year for US retailers and converts better than the rest of their traffic.
Few marketing fixes offer that ratio, so take this one. And if your site needs more than an afternoon of rewrites to get there, our web development team builds pages designed to confirm, answer and convert in the first screen.
Frequently asked questions
How do I track traffic from ChatGPT in Google Analytics?
Start with the AI Assistant channel in GA4's default channel group, which collects visits whose referrer matches Google's list of assistants. Then search your referral report for hostnames such as chatgpt.com, perplexity.ai and gemini.google.com. Visits with no referrer land in Direct, so also segment direct visitors who land on deep pages and add a "How did you hear about us?" field to your enquiry form.
Why is my AI referral traffic bouncing?
Usually because visitors cannot quickly confirm what the assistant told them. They arrive expecting a specific fact, like your service area or price range, and land on a general hero banner. Put that fact in plain text at the top of the landing page, answer common follow-up questions below it and make the next step obvious.
Is AI referral traffic worth optimising for if the volume is small?
Yes. The volume is small compared with search, but the visitors arrive further along in their decision, having already compared options with an assistant. In Adobe's US retail data they converted 42% better than other visitors in March 2026. The fixes are cheap too, mostly rewriting the top of key pages, so a modest number of extra enquiries covers the effort.
How do I get ChatGPT to recommend my business?
Make clear, consistent, specific information about your business easy to find. State your services, your area and your pricing in plain text on your site, and keep those details matching on your Google Business Profile and directory listings. Collect real reviews and earn mentions on independent sites. Assistants recommend businesses they can describe confidently.