Tom Storey has been getting leads from AI. Here is what he found out about them.

Tom Storey has been in real estate for nearly two decades and has led his Toronto-based team for close to ten years. He is not someone who chases trends. So when he spent several months deliberately optimizing for AI search, it was worth paying attention to what happened next.

He signed up for Rank My Agent. He added an AI disclosure to his website. He opened his YouTube video library to LLM indexing, a setting in YouTube’s backend that most agents do not know exists and that allows large language models to source video content directly. He rebuilt his approach to text-based content on the assumption that AI search rewards specificity and credibility over traffic optimization.

It worked. Leads started coming in from people who had never watched his videos, never met him at an open house, never been referred by anyone he knew. They had typed a question into Google or ChatGPT, received an AI-generated answer, and his name was in it.

“Some have turned into sales,” Storey said during the episode. “I think we’ve done five so far this year from this source.”

But the leads were not what he expected. And the gap between what he anticipated and what he got is the most useful part of his experience for other agents to understand.

What AI leads actually look like

The first thing Storey noticed was that the leads did not behave like referrals or YouTube leads. They arrived with no prior relationship, no context and no trust. Many came through his website contact form or called directly. Some booked through his calendar link.

What they had in common was a script. Every one of them asked roughly the same ten questions – questions that AI had apparently suggested they ask when interviewing a real estate agent. It felt, Storey said, like an interrogation.

“They didn’t know who I was. They didn’t care. They just got my number,” he said. “And buyers, who have never really done this in my career, were like, just want to let you know, we’re interviewing five people.”

The volume was also higher than he anticipated, to the point where the incoming leads became a workflow problem. His calendar link, which he had previously used with clients who already knew him, was being booked by strangers who had no relationship with him before the call.

Taylor Hack named the issue directly on the episode: Storey had been using content as a filter for years, building trust and familiarity before anyone reached out. With AI leads, that filter was gone. He was, as Hack put it, “raw-dogging the internet.”

Where AI leads rank as a source of business

Storey was straightforward about how he now categorises these leads relative to his other sources.

Repeat and referral clients sit at the top. They arrive with trust already established, the conversion rate is high and the relationship exists before the first conversation.

YouTube leads rank second. These are people who have watched his content, formed a view of him over time and reached out when they were ready. The quality of those calls is high because the work of building credibility happened before the phone rang.

Open house, Realtor.ca and other traditional sources sit in the middle.

AI leads currently rank closer to Facebook leads in terms of conversion difficulty. Not worthless – five closed sales this year is a real number – but requiring a different approach and a different expectation going in.

“They’re not shit,” Storey said. “But they’re complicated.”

Why Tom is ranking in AI search

Hack’s explanation for why Storey’s name is appearing in AI search results is worth understanding, because it is not primarily about the Rank My Agent profile or the AI disclosure on the website.

It is about the video library.

AI search systems filter for credibility by looking for volume and consistency of evidence. Storey has been producing real estate video content for years. His library covers news headline breakdowns, neighbourhood guides, first-time buyer explainers and long-form step-by-step guides. When he opened that library to LLM indexing, he gave AI systems access to an enormous body of text – every transcript from every video – that demonstrates deep, consistent, local expertise.

“Look at how much evidence you have of being an expert,” Hack said. “Look at how little evidence there is that you are not credible. You win. You made the largest, deepest knowledge base for someone competing for this probability spot.”

Reviews, by contrast, are weighted differently by AI than by traditional Google search. Hack noted that AI systems are increasingly sceptical of review patterns that look solicited or abnormal, like too many reviews mentioning the same name in a short window, unusual spikes in volume. The credibility signals that matter most to AI are harder to manufacture: consistent publication, media mentions, verifiable expertise demonstrated over time.

The four video types that serve different purposes

Storey outlined the four categories of video he produces and what each one actually does for his business, a framework that applies regardless of market.

News headline breakdown videos generate views and engagement, but do not directly produce appointments. Their value is reach and relevance, not conversion.

List videos – the seven best neighbourhoods, the ten things to know before buying – are evergreen. They do not spike in views, but they accumulate over two or three years and build credibility over time. These are the videos people find when they are doing research rather than when they are entertainment-scrolling.

Market update videos produce appointments. Clients and prospects watch these to understand what is happening and often reach out after seeing one.

Long-form masterclass videos – thirty minutes or more, covering everything someone needs to know about a specific process or decision – get the fewest views but the highest quality of engagement. These are for the small percentage of people who are close to making a decision and want to go deep. Storey’s first-home buyer guide in Canada is at 60,000 views, has produced approximately ten sales and has generated outbound referral fees from leads in markets he does not serve.

“The long form videos are the best for converting actual leads,” he said. “Less views, less engagement, but the quality of view is much higher.”

The twenty-dollar CRM

Storey also demonstrated live on the episode a team CRM he built entirely using Claude and Lovable, a no-code development tool. The system cost twenty dollars a month to run and replaced paid tools, including Trello and a previous CRM platform.

It includes a deal pipeline with drag-and-drop stages from active to closed, a client database with contact history and home anniversaries, a goal tracker that calculates required weekly appointment volume based on income targets and conversion rates, a property evaluation and CMA tool, and a team dashboard showing each member’s active clients and upcoming tasks.

His process was straightforward: take screenshots of the tools he was already using, give them to Claude with instructions to replicate the functionality, then move the output into Lovable to build and iterate the interface. Every member of his team has a login.

The security caveat Hack raised is worth noting: systems built this way require the same attention to data protection as any other client-facing tool. AI-built login systems are not inherently secure, and live client data warrants proper security review rather than assuming the default setup is sufficient.

What to do when the interrogation call comes in

The practical question Storey raised toward the end of the episode was how to handle the AI interrogation call, the lead who arrives with ten prepared questions and no prior relationship with you.

Hack’s answer was not to answer the questions as presented. The agents who just work through the list are indistinguishable from each other. The move is to change the conversation to the things the lead does not know to ask.

His framework: acknowledge the questions, then pivot to what actually determines whether an agent is the right fit – house knowledge, neighbourhood knowledge, market knowledge and deal knowledge. Most buyers can give a rough sense of what they want. Very few can tell you what they do not know they are missing.

“The question they’re asking is ‘how long have you been doing this,'” Hack said. “The question behind that question is ‘Can I trust you with the most expensive thing I will ever buy?’ Answer the real question.”

Storey’s working solution is to pre-answer the ten most common questions in short video format and send that automatically when a new AI lead books a call. The goal is to use the time on the call for the conversation that actually determines fit, rather than working through a checklist the AI generated.

The full episode covers the real estate health check and how Storey has adapted it for renters, the specialist versus generalist debate and how to hold a niche without losing general business, and how to use sold data rather than active listings when starting a new buyer relationship.

Join the conversation live

Each week, The Leads Are Sh*t brings working agents together to dissect what’s actually working in real estate right now.

Be part of the live recording, get your questions answered and join the after-show conversation.

👉 Register for the next live session.

Related Posts