What should happen after an AI assistant sends a buyer to your client’s site? Rich Thomas, co-founder and VP of Sales at CHAOS Supplies and the sales lead for its wheel-service brand CHAOS Moto, gave me two answers on Unscripted Small Business. First, the page should already look like the thing they asked about. Second, a person should call them back within the hour. He says that second habit alone won CHAOS three new customers in a month.
Most of the AI search conversation stops at getting cited. Rich’s story starts where that one ends, with a visitor who already arrived, and it’s a useful reminder for anyone who reports on AI-referred traffic: the referral is the start of the sale, not the end of it.
Watch the full episode on YouTube or listen on Castos.
An AI-referred visitor should land on a page that matches what they asked
Rich’s marketing is run by Alex Storch, a friend he hired who has a master’s in digital marketing from USC. Rich says “Alex has forty AI agents working for him,” and one of the things they do is shape what a visitor sees on arrival:
“If somebody searches for a product that’s related to CHAOS Moto through ChatGPT or one of the AI sites, when they go to our site, whatever they’re searching for, that homepage, that image comes up.”
He says the site “immediately becomes customizable and follows their previous searches for similar products.” The payoff he described was a utility company with 10,000 people that needs prescription safety glasses. They found CHAOS through AI, landed on the site, and saw utility workers. Rich joked that “if they were looking for cowboys, they probably would have found cowboys wearing safety glasses or clowns.”
I haven’t audited how the CHAOS site does this, so treat it as Rich’s description, not a teardown. The lesson for you doesn’t depend on the tooling, though. If you can see which prompts or topics send AI traffic to a client, check what the landing page shows those visitors. A generic homepage makes a buyer who asked a specific question start over. For the retrieval side of the same problem, see how to get recommended by an LLM before you’re in the training data.
The lead is only worth what happens in the first hour
The utility company booked a Zoom call for the next morning. Rich was clear that AI’s part of the job was over:
“AI will get you there to one point, but now you have to do the rest of the work.”
Then he told me about three customers who switched to CHAOS in the last month, and why. “We called back within an hour of the lead coming in off of AI.” His theory is that competitors assume the software handles it: “They think that the computer can do it all right down to closing the deal.” He called that “the glitch in the system.”
The data says he isn’t imagining the gap. Jake Melendy of Ignitvio mystery-shopped service businesses, and in his study over half never responded at all. Of the ones that did, only 2% followed up more than once. Alex Zweydoff makes the same case for property managers in owner intent over tenant traffic. If you sell SEO, response time is part of your results whether it’s in your scope or not. We laid out how to measure it in speed-to-lead: the SEO fix that is not SEO.
Nobody can say which channel gets the credit, so stop fighting over it
I brought up something Keith Perhac of SegMetrics calls the messy middle. A buyer finds your site through Claude on a phone, emails it to their spouse, the spouse runs a Google search, and someone finally types the URL in. Was that AI, Google or direct? Yes. All three.
The danger isn’t the reporting gap. It’s what companies do with it. Too many set sales and marketing up in a death match over who gets credit, and in my words on the show, “you’re cannibalizing when you should be collaborating.” Keith’s practical fix is to stop chasing a perfect model, and his two-question method for first touch and last touch attribution is the version I’d start with.
Give sales the article before the objection comes up
The collaborating version looks like this. Marketing learns the friction points that come up on sales calls and answers them in content. Then, when a prospect says they don’t know if the product can do something, the rep can say “we wrote an article about that” and send the link. That same page is now there for the people who never got on a call, because they find it through AI and Google.
Rich’s reaction: “it couldn’t have been explained better what you just said.” Then he added a limit. For cheap, simple purchases, buyers are happy with no human at all. For anything where you’re spending real money, have questions and expect to come back for service, his line was simple: “I want to talk to somebody.”
AI skims the surface, so plan for the buyers below it
Rich’s analogy was fishing for bass. AI skims the waters and picks up whatever is near the surface. Pair it with a salesperson and “you can get the fish on top and you can get the fish down below that are deep that aren’t on the surface.” His conclusion: “you can’t eliminate salespeople and you can’t eliminate employees on a lot of levels just by throwing AI in. It’s not the answer.”
For your reporting, that means AI-referred leads and human-closed revenue belong on the same page. If you show a client AI citations and sessions but not what happened after the visit, you’re reporting the surface catch.
Questions to ask before you send a client more AI traffic
- Which prompts and topics send AI-referred visitors to the site, and what does each landing page show them?
- When a lead comes in from AI, how many minutes until a human calls back, and who owns that step?
- Which sales objections keep coming up, and is there a page that answers each one?
- Do sales and marketing agree on how a lead gets credited, or are they arguing about it?
- Where does the time AI saves the team go: more output, or fewer people?
The full episode recap is Rich Thomas on why AI brings the lead and a human closes it on Unscripted Small Business, and my own take is on hiring your opposite and handing off what bores you. You can see the products Rich talked about at CHAOS Moto and CHAOSRXOptics.