When a consumer asks an AI agent a direct question about price, they buy nearly half the time. That’s the finding from Conviva’s new A2A Commerce Report, and it’s a signal invisible to every traditional product analytics platform on the market.
A2A Commerce Is Scaling Quickly, Data Finds
Agent-to-agent (A2A) commerce is not a future state. Conviva analyzed full-census behavioral data from e-commerce sites processing more than 100 million sessions per week and found a channel already generating tens of millions in revenue at zero acquisition cost.
On a single site in the dataset, 478 confirmed purchases per day were directly attributed to AI platform referrals. At an average order value of $250, that’s approximately $44 million in annualized revenue generated with no paid media, no retargeting, and no spend on ChatGPT or other AI referral sources. Daily AI-referred sessions on that same site rose from 11,000 on May 4 to a stable floor of 13–14K by June 1, a 25% increase in under 30 days and on pace to exceed 20,000 per day by Q3 2026.
Most retailers have no dashboard for this channel, no attribution line, and no optimization roadmap.
Why AI-Referred Buyers Convert at 3.5x the Site Average
AI-referred visitors behave differently from organic traffic because they arrive with a decision already made. They’ve completed their research inside an AI platform, selected a specific product, and come to the site to complete the transaction. These visitors convert 3.5x better than the site average. If a third-party agent interacts directly with a brand’s own agent (known as agent-to-agent commerce), conversion jumps to 4x the site average.
Conviva found that 83% of AI-referred visitors landed directly on a product page because the AI recommended a specific item. 92% of those sessions involved a single page, and 100% of AI-attributed purchases in this dataset happened within the same session the visitor arrived in. The buying decision was formed before the buyer reached the site, which means the product page’s job is to confirm the decision, not create it. Every element that introduces doubt or adds friction works against conversion in this cohort.
The Highest-Converting Signal in the Dataset: Price Intent
Nexa, Conviva’s analytics agent, found a behavioral pattern within agent sessions: when a buyer engages an on-site agent and asks about price, they convert at 45.5%. That’s the highest-converting segment in the entire dataset, outperforming every other source, cohort, or session type measured.
Price intent expressed in natural language to an agent is one of the strongest purchase signals in e-commerce, and it has been invisible until now. Standard event-based analytics tools track clicks, pageviews, and time on site. Agent tools focused on model outputs and latency aren’t built to track a consumer’s experience across the AI agent, app, and website, or to correlate the sequence of those interactions with whether a transaction completes. Humans can manually review individual agent conversations, but not at the scale we’re talking about. Only agentic analytics, like Nexa, can identify patterns where an agent is involved.
Conviva tracks every agent interaction as a stateful, sequential behavioral pattern, which means seeing not just that a session occurred but the intent expressed within it and how that pattern predicts conversion downstream.
What to Do With This Signal
The A2A channel is real, it’s growing, and it’s currently unoptimized by most retailers. Four actions prioritized by revenue impact:
- Optimize product pages for decided buyers. 92% of AI-referred sessions involve a single product page visit. Remove discovery overlays and cross-sell carousels from the initial viewport for AI-referred sessions and test a stripped-down variant: specs, delivery confidence, social proof, and a clear call to action.
- Activate on price intent in real time. When the on-site agent detects a price question, route that session to a live offer, loyalty prompt, or discount immediately. Real-time agent topic classification, detecting the pattern as it happens rather than after the session ends, is what separates a curated experience from a missed conversion.
- Track AI platforms as formal acquisition channels. ChatGPT and other AI referral sources need their own CVR dashboards and revenue attribution lines. As the source mix diversifies beyond ChatGPT, source-level visibility becomes essential to protect the return on AI optimization investment.
- Invest in AI discoverability before it becomes paid. Structured product data and AEO/GEO optimization determine which products AI systems recommend. AI platforms have not introduced paid placement tiers yet, and that window is finite.
Key Takeaways
- AI-referred visitors convert at 3.5x the site average. When those agents interact with a brand’s own agent (A2A) conversion is 4x the site average. ChatGPT desktop peaked at 7x average conversion, the highest in the dataset.
- 100% of AI-attributed purchases happen within the same session, confirming that buying decisions are made inside the AI platform before the buyer reaches the site.
- Buyers who ask an on-site agent about price convert at 45.5%, the highest-converting behavioral segment in the dataset.
- On a single site, the A2A channel is estimated to generate approximately $44M in annualized revenue at zero customer acquisition cost.
- On a single site, AI-referred sessions grew 25% in under 30 days and are on pace to exceed 20K per day by Q3 2026.
FAQ
Q: What is agent-to-agent (A2A) commerce?
A: A2A commerce describes the channel where an external AI agent such as ChatGPT researches and recommends a product for a consumer. In A2A commerce, the external agent engages with the brand’s agent to find more information, like availability, size and color options, price, and more. Two agent layers are involved: the referring platform and the on-site agent.
Q: Why do AI-referred visitors convert at higher rates than organic traffic?
A: AI-referred buyers have already completed their research inside the AI platform before arriving at the site. 83% land directly on a specific product page, 92% view only one page, and 100% of purchases in Conviva’s dataset happen within the same session. The buying decision is pre-made, so the product page needs to confirm it rather than create it.
Q: What is the conversion rate when a buyer asks an on-site agent about price?
A: Buyers who engage an on-site agent with a price question convert at 45.5% in Conviva’s dataset, the highest-converting segment measured and nearly 45x the 1% site average.
Q: What does Conviva measure that traditional analytics tools can’t see in A2A sessions?
A: Traditional product analytics tools track events: clicks, pageviews, time on site. They can’t connect a consumer’s behavior across an external AI agent, a website, and an on-site agent, or correlate the sequence of interactions with whether a purchase completes. Conviva captures full-census, stateful, sequential behavioral data across all three surfaces, making patterns like price intent visible and actionable in real time.
Q: Is the A2A channel currently being tracked by most retailers?
A: Most retailers do not have AI referral sources broken out as formal acquisition channels, have no CVR dashboards for ChatGPT or other AI platforms, and have no real-time visibility into on-site agent behavioral patterns. Conviva’s report documents a channel that is generating significant revenue with no paid acquisition spend and is largely untracked by the retailers benefiting from it.