You can’t buy a slot in a ChatGPT answer. There’s no admin console, no feed spec, no bid. Vinod Kumar, co-founder and CEO of Syntheum AI, put a name on that on the Unscripted SEO Podcast: AI search is a probabilistic marketplace, and nobody has solved how a brand controls what shows up in it. Here’s how he frames the problem, what his team does about it, and what it means for the way you report to clients. You can watch the full episode here or listen on Castos.
Deterministic marketplaces vs probabilistic ones
Vinod spent years at Salesforce building technology for brands and retailers. His frame starts with marketplaces brands already know. On Amazon or Walmart, a brand logs into a console, chooses which products to feed in, sets business rules, and the marketplace shows exactly what it was told to show.
“The brand can say, I want product A, B, and C to show. And the marketplace will show products A, B, and C. But a GenAI-driven marketplace is more of a probabilistic marketplace.”
In a chat, there’s no definite way to say which products appear. Vinod’s point was that sometimes even the engineers who build these models don’t know how they’ll behave. So the question for a brand isn’t how to configure the listing. It’s how to influence a system that doesn’t take instructions.
Two problems: be in the answer, then own the answer
Vinod broke it into two parts, and they map neatly onto work you’re probably already selling.
- Participation. Can these engines crawl your site on whatever schedule they keep, and when they do, do they get the right, current information? If blue is your newest mug color, how do you tell the crawler blue should come first?
- Control. When your product does come up, is your own site the source, or a third party? Search for a Dell laptop and the answer might lean on a review site instead of dell.com.
“So how do brands control that narrative? How do brands control that story is another piece of the puzzle that needs to be solved.”
And he didn’t pretend anyone has solved it.
“Even people who say they’ve solved it actually don’t have an answer. It’s more theoretical hand waving at this point.”
That’s a refreshing thing to hear from someone whose company sells AI. If a vendor pitches you guaranteed AI visibility, keep that line handy.
Fewer clicks, better buyers, and a reporting problem
I told Vinod the SEO side of this has felt like someone eating our sandwich. We built the top-of-funnel content, Google sent the traffic, and that deal got rewritten one step at a time, first by featured snippets and then by AI Overviews. I’d seen a chart showing HubSpot losing a quarter or more of its organic clicks while revenue kept climbing.
That leaves agencies in an awkward spot. We’re asking clients to accept fewer clicks for the same amount of money on the promise that the visitors who do show up convert better, and inside a black box we can’t fully map that. What I’ve seen myself, researching with Claude, is that you end up arriving at the bottom of the funnel with the comparisons already done, pretty much just there to buy the thing.
Vinod agreed, and added that shoppers can now go a long way down the consideration path inside ChatGPT, Perplexity and Copilot. He called it total greenfield. If you’re explaining this to clients, show them the conversion side of the report, not just sessions.
Start with an audit of what AI says about you
Vinod’s practical answer is a free tool Syntheum launched the week before we recorded: AVA, the AI Visibility Auditor. You enter a URL and AVA simulates shopping-style conversations in ChatGPT and Gemini, then compares what those engines say with what your homepage says.
- Where your brand shows up and where it’s doing well
- Which competitors appear in the conversations relevant to you
- Themes and campaigns on your homepage that never make it into the answers
- Things the engines say that your homepage no longer does
It’s built for e-commerce. Vinod was clear that it isn’t exhaustive or comprehensive, but it gives a brand a list of strengths, weaknesses, opportunities and risks to start from. For an agency, that’s a decent opener for an AI visibility conversation with a retail client.
A content team of one, with a human on the publish button
The other half of the episode was about running a lean startup with agents. Vinod said his content strategy team is “a big team of one.” They use Perplexity Pro for topic research, and they’ve been building workflows in n8n and Lindy, which he described as WYSIWYG agent builders where you daisy-chain steps.
“The workflow can then go to the Google Sheet, take the last row, which is the latest content idea, expand on it using OpenAI, let’s say, and then build a complete blog post and say three LinkedIn posts”
Everything lands in Notion or a Google Doc for review. Vinod said that job usually takes two to three people. Now the team drops ideas in a sheet, reviews what comes out, and declines anything they don’t like. Notice what stays human: the ideas going in and the approval going out.
The full conversation, including the khaki vs khakis example that explains Syntheum’s core product, is in the episode recap on Unscripted SEO.