Key Takeaways

  • Resend captures over 60 percent of email infrastructure recommendations inside Claude Code by engineering specifically for autonomous agent scrapers.
  • Interactive JavaScript elements block AI models; coding agents cannot interact with dynamic pricing calculators or interactive sliders.
  • Exposing raw Markdown endpoints for marketing pages and documentation allows LLMs to ingest technical details without parsing clutter.
  • Resend writes explicit text tables alongside interactive tools so scrapers extract pricing tiers directly.
  • AI models like Claude 3.5 Opus demonstrate strong recency bias, favoring tools with updated plain-text documentation and structured llms.txt files.

The Method

Traditional search optimization targets human eyeballs scanning a Google results page. Answer engine optimization targets language models generating code. When Swyx pointed out that Claude Code recommends Resend “60% plus of the time,” (play from 6:45) Rocha explained that this dominance was built by stripping away the front-end tricks modern web designers love.

First, eliminate interactive UI dependencies from your data layer. Most developer tools show pricing through dynamic sliders. A user drags a handle from 10,000 to 50,000 emails per month, and a JavaScript script calculates the monthly cost. Rocha realized this breaks agent discovery completely: “The problem was, agents don't really drag a JavaScript widget.” (play from 1:01:55) If an LLM cannot execute the script or manipulate the DOM slider, it sees an empty state or hallucinated numbers. Rocha solved this by embedding complete, static text tables beneath the visual UI. As Rocha put it, “I spell out the same data in text so that LLMs can actually read that thing.” (play from 1:04:49)

Second, deploy raw Markdown endpoints across your entire domain. Developer agents do not want HTML tags, tracking pixels, or CSS stylesheets. They want raw tokens. Rocha applied this logic to documentation and expanded it to the entire marketing footprint: “Same with the markdown versions of the docs. I'm going to do markdown versions of every single page on the website, on the marketing website.” (play from 1:00:30) By maintaining clean Markdown alternatives accessible to automated crawlers, Resend ensures that models consume clean API parameters and copy-paste code snippets with zero parsing overhead.

Third, monitor crawler recency bias. As Rocha noted with Anthropic models, “At least with Opus 4.5 there was a recency bias in terms of the technology and where they were fetching data.” (play from 1:06:29) Models prioritize freshly updated sources that answer explicit developer queries in plain text. Resend pairs its Markdown routes with an llms.txt index file and explicit FAQ blocks that match the prompt phrasing developers type into terminals.

Where This Breaks Down

This strategy works when your primary customer is a technical developer relying on an agentic coding assistant to scaffold an application. It fails when your product relies on visual differentiation or enterprise procurement committees.

If you sell enterprise software where purchase decisions depend on multi-page PDF case studies, security questionnaires, and sales calls, stripping your site to Markdown does nothing for revenue. In that world, an AI agent recommending your API route does not sign the six-figure contract.

Furthermore, pure Markdown endpoints make content scraping trivial for your competitors. When you expose your pricing structures, API limits, and architecture in plain text, rival products can parse your entire product catalog in minutes. You trade proprietary layout protection for raw agent discoverability.

What to Do With This

Audit your pricing page in a headless browser with JavaScript disabled. If your tier limits, monthly fees, or API rate caps disappear when scripts do not run, write an unstyled HTML table containing every numerical tier and publish it directly into the page source before Friday.