Back to all
At Rithum LIVE, one of the most engaging discussions came from the audience Q&A with Accenture Song’s Caitlyn Ford and Brittny Cantor. As brands and retailers wrap their heads around Generative Engine Optimization (GEO), this unscripted session got into the harder questions about AI bias, agent monetization, liability, and what commerce looks like when the product page isn’t the final destination.
Here are some of the highlights.
Q: If AI agents start recommending products, how do we know they won’t be biased by paid placement? Can brands “bribe” the agent?
Brittny: That’s the million-dollar question. OpenAI is already standing up monetization capabilities. We expect some form of sponsored placements to roll out this holiday season, likely in beta. The big unknown is how they’ll balance monetization with trust. If sponsorship tilts results too far, it could undermine the whole experience.
Caitlyn: It’s like the early days of retail media. Brands need to prepare for both tracks: getting their product information right for organic visibility, and being ready to test sponsored placements as those capabilities come online.
Q: Who’s responsible when an AI gets it wrong, like recommending the wrong product or misrepresenting information?
Brittny: It’s still a gray area. Right now, no one’s eager to raise their hand and claim full responsibility. It’s a bit of the Spider-Man meme: brands pointing to marketplaces, marketplaces pointing to LLM providers. What we need is a set of shared guardrails that define ownership: who owns the inputs, who governs the outputs, and who steps in when something breaks.
Caitlyn: Until that’s clear, the best risk management is having a strong human-in-the-loop process. That means validating your content before it goes live, especially as more content is generated or summarized at scale.
Q: How close are we to “buy in agent” experiences where shoppers don’t even visit your site before purchasing?
Brittny: Closer than most people realize. Perplexity is already experimenting with it. Shopify is forming tight integrations with ChatGPT. Once these flows are normalized, conversion may happen entirely inside the agent. But that raises new challenges for brands: inventory accuracy, pricing consistency, fulfillment handoff. One bad checkout experience—say, ordering one item and receiving 15—could turn off a shopper for good.
Caitlyn: Expect early adoption with lower-cost, low-stakes items. As comfort builds and trust deepens, agent-led purchasing will grow. The infrastructure just needs to catch up.
Q: In a multi-channel world, should brands still invest in their own .com or focus on winning inside agents?
Brittny: Your brand site still matters, just not the way it used to. No one cares about animation or hover states anymore. They care about whether your PDP content can be parsed by an agent. So yes, invest in brand.com. But optimize it for machine readability, not just visual design.
Q: How should brands measure ROI on GEO?
Brittny: This is the question we get all the time. One simple model we use is: if shoppers are 4 to 7 times more likely to convert after interacting with generative content, and your AOV stays constant, you can estimate the lift. But that’s just the starting point.
Caitlyn: Treat GEO like you’d treat a media campaign. Set a flight window. Isolate your tactics. Measure the lift in-store or online. And remember: not all conversions are immediate. GEO can drive long-tail impact—people don’t always click, but they remember.
Q: What does agentic commerce look like in the long run?
Brittny: Picture your own Jarvis: an always-on assistant that knows your preferences, limits, timing, and priorities. That’s where this is heading: personal agents transacting across the internet on your behalf. We’re not there yet, but the foundation is already being built.
Final thought: Don’t panic. Focus.
As Caitlyn put it in her closing remarks: focus matters most. Whether you’re cleaning data, refining PDP content, or choosing a tech partner, the best approach to agentic commerce is to double down on clarity, credibility, and control—because that’s what machines are trained to trust.
Watch the Rithum LIVE sessions on demand here.
Talk to our team
