The agent successfully assembled a comprehensive explanation of Deepgram's products, pricing, target audience, and competitive positioning, but had to work around an inaccessible primary website. Deepgram.com itself is a heavily JavaScript-rendered Next.js application that does not expose readable content in HTML responses, forcing the agent to rely entirely on web search results and third-party comparison articles (Gladia, G2) rather than Deepgram's own documentation or official pricing pages. The agent's final answer is well-structured and substantive, but sources external commentary rather than primary sources.

• ›The official Deepgram website (deepgram.com/, /pricing, /product/speech-to-text) consistently returned Next.js boilerplate HTML with no content—the agent could not extract information from Deepgram's own pages despite multiple attempts.
• ›Web search [10] and [11] surfaced comparison articles and competitor analyses as the primary discovery mechanism; the agent then fetched [13] (Gladia's Deepgram pricing article) and [16] (Gladia's alternatives list), which contained substantive competitive data and pricing breakdown that became the backbone of the answer.
• ›Search result [17] identified deepgram.com/product/speech-to-text as a canonical product page, but when fetched in [19], it returned the same unreadable Next.js shell, so it was never used; the agent cited it in sources despite not extracting content from it.
• ›Pricing details ($0.0043–$0.0092/min, $4.50/hour for Voice Agent API, free tier credit amounts) came from third-party sources (Gladia, BrassTranscripts), not from Deepgram's official pricing page, which the agent could not access.
• ›Competitive positioning (vs. AssemblyAI, Speechmatics, ElevenLabs, Gladia) and detailed feature comparisons were sourced from Gladia's third-party articles, not from Deepgram's own documentation or messaging.
• ›The agent correctly identified the primary friction point in its response: 'The website is heavily JavaScript-based' and explicitly called out inability to access Deepgram's own feature matrix, model documentation, and add-on pricing details.