sendbird.com — AI Site Grade

Analysis

Sendbird — AI-Visibility Audit

The homepage at sendbird.com carries a noindex, nofollow robots meta tag, meaning Google and other search engines are explicitly instructed not to index the site's primary entry point — a rare and aggressive stance for a company positioning itself as an enterprise AI platform.

Crawler Access Split

The sendbird.com robots.txt is a study in contradictions. GPTBot and Google-Extended are disallowed from the entire domain, while OAI-SearchBot, ChatGPT-User, and PerplexityBot are explicitly allowed. Yet compare_bot_access shows every AI bot tested — including GPTBot and Google-Extended — receives a 200 with full content (97KB) from the Netlify-hosted homepage. The robots.txt rules are not enforced at the server level, meaning training crawlers (GPTBot, Google-Extended) can and do ingest the full page despite the explicit Disallow: / directive. The delight.ai subdomain tells a different story: hosted on Cloudflare, it blocks 8 of 11 AI bots at the WAF layer (403 "Your request was blocked"), including OAI-SearchBot, PerplexityBot, ClaudeBot, and ChatGPT-User, despite its own robots.txt explicitly allowing them. Only Google-Extended, Applebot-Extended, and Browser pass through. The delight.ai robots.txt is a sophisticated AEO document with commented sections distinguishing "Answer Engine Optimization" from "AI Training" bots — but the Cloudflare WAF overrides the intent.

Cold-Knowledge Gap

The LLM's cold knowledge describes Sendbird as an in-app chat, voice, and video API provider competing with Twilio and PubNub, with customers including DoorDash, Reddit, and Hinge. The site itself has undergone a complete repositioning: the homepage H1 is now "The AI customer experience platform," and the primary product is delight.ai — an "AI concierge" and "Agent-as-a-Service" platform with memory, omnipresence, and autonomous resolution. The legacy communications API (Chat, Calls, Desk) is relegated to a secondary navigation item. The cold model knows nothing about delight.ai, Agent Memory Platform (AMP), For You Conversations (FYC), Trust OS, or the AI concierge concept — a gap of roughly 18 months of product evolution. The model's knowledge of "Sendbird Desk" as a current product is stale; the site now routes Desk under delight.ai.

Schema and Content Posture

The homepage carries no WebSite or Organization JSON-LD schema — only a fragmented @graph block with an Article type labeled "Fork page" (an internal CMS artifact) and a bare Organization block. The Organization schema is present but buried inside a graph wrapper on subpages like /pricing/chat and /about. The delight.ai subdomain has rich schema: FAQPage with 4 questions, SoftwareApplication with an aggregateRating of 4.6/5 from 116 reviews, and Service type — all absent from the primary sendbird.com domain. The homepage has only 271 words of visible text, suggesting a heavy JS-rendered shell that may produce thin content for crawlers that don't execute JavaScript. The /ai page is a canonical duplicate of the homepage (canonical: https://sendbird.com) and also carries noindex, nofollow.

External Signals

The DNS TXT records reveal integrations across the AI ecosystem: openai-domain-verification, anthropic-domain-verification, cursor-domain-verification, and bird-domain-verification (Twitter/X). The llms.txt file at sendbird.com is 281KB — unusually large and comprehensive, containing documentation links, blog listings, and full blog post excerpts. This is a strong positive signal for AI discoverability, though its value is undermined by the homepage's noindex directive and the delight.ai WAF blocks. No external review sites (G2, Reddit) surfaced in search results, suggesting limited third-party discourse about the brand's AI pivot.

Findings

1. Homepage carries noindex, nofollow meta tag High

   The sendbird.com homepage includes a robots meta tag instructing search engines not to index or follow links on the page, which is a rare and aggressive stance for an enterprise AI platform.

   What to change: Remove the noindex, nofollow meta tag from the homepage to allow search engines to index the site's primary entry point.

2. Robots.txt disallows GPTBot and Google-Extended but server serves full content High

   The sendbird.com robots.txt disallows GPTBot and Google-Extended from the entire domain, yet the server returns a 200 with full content to all AI bots tested, including those disallowed. This means training crawlers can ingest the page despite the explicit directive.

   What to change: Either enforce the robots.txt rules at the server level (e.g., via Netlify configuration) or remove the disallow directives for GPTBot and Google-Extended to align intent with behavior.

3. Delight.ai WAF blocks AI bots despite robots.txt allowing them High

   The delight.ai subdomain's robots.txt explicitly allows OAI-SearchBot, PerplexityBot, ClaudeBot, and ChatGPT-User, but the Cloudflare WAF returns 403 for these bots, blocking 8 of 11 AI bots tested. Only Google-Extended, Applebot-Extended, and Browser pass through.

   What to change: Update the Cloudflare WAF rules to allow the AI bots that are permitted in robots.txt, or align robots.txt with the actual access policy.

4. LLM cold knowledge lacks Sendbird's AI repositioning and delight.ai High

   The LLM's cold knowledge describes Sendbird as an in-app chat API provider, with no awareness of the delight.ai AI concierge platform, Agent Memory Platform, or other recent product innovations. This gap spans approximately 18 months of product evolution.

   What to change: Increase AI-visible content about delight.ai and the AI platform through structured data, blog posts, and external citations to update LLM knowledge.

5. Homepage has only 271 words of visible text Medium

   The sendbird.com homepage contains only 271 words of visible text, suggesting a heavy JavaScript-rendered shell that may produce thin content for crawlers that do not execute JavaScript.

   What to change: Increase the amount of static HTML content on the homepage to ensure crawlers can extract meaningful information without JavaScript execution.

6. Homepage lacks WebSite and Organization JSON-LD schema Medium

   The sendbird.com homepage has no WebSite or Organization JSON-LD schema. The only schema present is a fragmented @graph block with an Article type labeled 'Fork page' and a bare Organization block, which is an internal CMS artifact.

   What to change: Add proper WebSite and Organization JSON-LD schema to the homepage to improve structured data for search engines and AI crawlers.

7. /ai page is a canonical duplicate of homepage and also noindex Medium

   The /ai page at sendbird.com/ai has a canonical URL pointing to the homepage and also carries noindex, nofollow, making it effectively invisible to search engines and AI crawlers.

   What to change: Remove the noindex directive from /ai and set a self-referencing canonical to allow it to be indexed as a distinct page about AI capabilities.

8. No external review sites surfaced in search results Medium

   Web searches for Sendbird on G2, Reddit, and other review platforms returned zero results, indicating limited third-party discourse about the brand's AI pivot.

   What to change: Encourage customers to leave reviews on G2, Capterra, and other platforms, and engage in relevant online communities to build external signals.

9. Rich schema on delight.ai not replicated on sendbird.com Medium

   The delight.ai subdomain has rich schema including FAQPage, SoftwareApplication with aggregateRating, and Service type, but these are absent from the primary sendbird.com domain, missing an opportunity to boost AI visibility for the main brand.

   What to change: Add similar rich schema (FAQPage, SoftwareApplication, Service) to the sendbird.com homepage and product pages to improve structured data coverage.

What's working

• Large and comprehensive llms.txt file — The sendbird.com llms.txt file is 281KB, containing documentation links, blog listings, and full blog post excerpts, which is a strong positive signal for AI discoverability.
• DNS verification records for multiple AI platforms — DNS TXT records include openai-domain-verification, anthropic-domain-verification, cursor-domain-verification, and bird-domain-verification, indicating proactive integration with the AI ecosystem.
• Delight.ai has rich FAQPage and SoftwareApplication schema — The delight.ai subdomain includes FAQPage with 4 questions, SoftwareApplication with an aggregateRating of 4.6/5 from 116 reviews, and Service type, providing excellent structured data for AI crawlers.
• Sitemap available with 80 URLs — The sendbird.com sitemap is accessible and contains 80 URLs, helping crawlers discover site content.
• Active blog with recent posts — The Sendbird blog at /blog contains recent posts, including one about delight.ai, providing fresh content for AI crawlers.
• Detailed comparison page against Stream — The /stream-comparison page provides a head-to-head comparison with Stream, offering substantive content that can help AI crawlers understand Sendbird's positioning.
• Customer stories page with brand logos — The /customers page lists customer logos and stories, providing social proof and content for AI crawlers.
• Pricing page with detailed plan information — The /pricing/chat page contains 2732 words of detailed pricing and plan information, offering rich content for crawlers.