Loading…
We Tested Marketing Incentives to AI Agents. Here's What Happened.
Grace Cummins
- Source
- Ramp
- Published
- Added to Yomu
Summary
Ramp tested whether AI agents would relay a tracked referral incentive to people researching spend-management software. Across roughly 50 marketing pages, Cloudflare Workers served bot-specific variants—Markdown, stripped HTML, or schema-heavy content—while unique URLs and logs attributed downstream activity. Markdown surfaced most reliably, and bot classification proved difficult: major assistants were labeled “AI Assistants,” OpenAI SearchBot cached aggressively, and DeepSeek required TLS fingerprinting plus ASN detection. Over 32 days, agent relays rose from roughly 40 at 2.5 weeks to nearly 10 times that total, with Claude responsible for most matches; Perplexity moved from vague references to one specific branded mention, while ChatGPT remained at zero. The experiment concludes that agents can relay offers, but model-specific content strategies, citation-aware page selection, and better attribution are still needed.
Context
The work responds to the expectation that agent-mediated B2B purchasing will grow, including Gartner’s estimate that $15 trillion in B2B purchases could be agent-mediated by 2028. The experiment asks whether agents can receive and relay a direct referral offer rather than evaluating a site through its human-oriented presentation.
Approach / What changed
Three content variants were served to detected AI bots across about 50 pages using Cloudflare Workers, while human visitors saw no change. The variants used Markdown, stripped HTML, or schema-heavy content, with balanced test groups, unique tracking, bot-traffic logs, and daily prospect-style queries across Claude, ChatGPT, Perplexity, and Gemini.
Takeaways
- Cloudflare classifies ChatGPT, Claude, and Perplexity crawlers as “AI Assistants,” so rules aimed only at “AI Search” can miss major platforms.
- Pages with higher existing AI citation volumes were much more likely to surface the embedded offer; low-citation pages produced no incentive mentions regardless of format.
- Anthropic’s crawler generated more traffic than all other named AI crawlers combined in the observed logs, while some high-volume agents required IP-based detection because their user agents resembled ordinary browsers.