sandboxaq.com — AI Site Grade

Analysis

SandboxAQ's homepage prominently features a Claude integration banner ("Use Claude to Access Our Large Quantitative Models") — yet the site has no llms.txt, no FAQ schema, and the cold LLM knowledge about the company is already 12-18 months stale, missing the entire LQM narrative that the site now leads with.

Crawler Access

All major AI crawlers — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, anthropic-ai, Applebot-Extended — receive a 200 with full HTML content identical to the browser baseline (62,527 bytes). The sole exception is Bytespider (ByteDance), which gets a 403 from Cloudflare. The robots.txt is a bare minimum: a single User-agent: * rule with no disallows and a sitemap pointer. No AI-bot-specific directives exist. The site runs on Cloudflare behind Webflow hosting, with a strict-transport-security header and frame-ancestors 'self' CSP. No llms.txt exists (returns 404). The sitemap.xml contains 676 URLs and is well-formed.

Cold-Knowledge Gap

The LLM's prior knowledge describes SandboxAQ as a "quantum sensing and cryptography" company spun out of Alphabet in 2022, listing AQtive Guard, AQBioSim, and AQNav as flagship products, and citing a $500M raise at $5B+ valuation in 2023. The actual site tells a fundamentally different story: the brand is now positioning itself around Large Quantitative Models (LQMs) as the core differentiator — "LLMs generate content. LQMs generate results." The homepage and solution pages lead with LQMs for drug discovery, materials, cybersecurity, and navigation, with a prominent Claude integration (Anthropic partnership) announced May 2026. The cold knowledge contains zero mention of LQMs, the Anthropic partnership, or the "Quantitative AI" framing that dominates the current site. The $500M funding figure and $5B+ valuation are not mentioned anywhere on the site.

Schema Posture

The homepage carries a BreadcrumbList and an Organization schema with logo, URL, and social profiles. The About page uses a LocalBusiness schema type (unusual for an enterprise AI company) with a Tarrytown, NY address. Blog posts use BlogPosting schema with datePublished and dateModified. However, no product-level schema (SoftwareApplication, Product, WebApplication) exists on any solution page. The drug discovery, AQNav, and LQM pages have zero JSON-LD. No FAQPage, HowTo, or TechArticle schemas are present anywhere despite the site containing definitional content about LQMs, comparison language ("LLMs vs LQMs"), and feature lists that would naturally map to structured data.

External Signals

The press room shows coverage from TechCrunch, WSJ, WIRED, Fox Business, CIO.com, and Nature. The DNS TXT records reveal an Anthropic domain verification token (anthropic-domain-verification-1qxf8t), confirming the partnership is production-integrated. An OpenAI domain verification token is also present. The site references NVIDIA collaboration and Google Cloud infrastructure. No negative signals, Reddit threads, or controversy surfaced. The blog is actively published (multiple posts per month through May 2026).

Findings

1. Cold LLM knowledge is 12-18 months stale, missing LQM narrative High

   The LLM's prior knowledge describes SandboxAQ as a quantum sensing and cryptography company, with no mention of Large Quantitative Models (LQMs), the Anthropic partnership, or the 'Quantitative AI' framing that now dominates the site. This gap means AI assistants cannot accurately represent the company's current positioning.

   What to change: Publish an llms.txt file and an llms-full.txt that describe the company's current focus on LQMs, the Anthropic partnership, and key solutions. Update the site's metadata and structured data to reinforce the LQM narrative.

2. No llms.txt file published High

   The site returns a 404 for llms.txt, meaning AI assistants have no structured, machine-readable overview of the company's content. This is a missed opportunity to guide LLMs with accurate, up-to-date information.

   What to change: Create and publish an llms.txt file at the root domain, following the llms.txt standard, to provide AI crawlers with a curated summary of the site's key pages and content.

3. No product or software schema on solution pages High

   Solution pages for LQMs, AQNav, and drug discovery contain zero JSON-LD structured data. These pages describe software products and services that would benefit from SoftwareApplication or Product schema, improving AI understanding and rich results.

   What to change: Add SoftwareApplication or Product JSON-LD schema to each solution page, including properties like name, description, applicationCategory, and offers.

4. No FAQPage schema despite definitional content Medium

   The site contains comparison language and feature lists (e.g., 'LLMs vs LQMs') that are natural candidates for FAQPage schema. Without it, AI assistants may not surface these explanations in rich results.

   What to change: Identify pages with Q&A-style content and add FAQPage schema with Question/Answer pairs.

5. About page uses LocalBusiness schema instead of Organization Medium

   The About page uses a LocalBusiness schema type with a physical address, which is unusual for an enterprise AI company and may confuse AI parsers about the company's nature. Organization or Corporation schema would be more appropriate.

   What to change: Replace LocalBusiness schema with Organization schema on the About page, and consider adding a separate LocalBusiness schema for the physical office if needed.

6. Cold knowledge cites stale funding and valuation not on site Medium

   The LLM's prior knowledge mentions a $500M raise at $5B+ valuation in 2023, but these figures are not present anywhere on the current site. This mismatch can lead to fabricated or outdated information in AI responses.

   What to change: Consider adding current funding or valuation information to the site (e.g., on the About or Press page) to align with external knowledge, or ensure llms.txt provides accurate context.

7. Bytespider crawler blocked by Cloudflare Low

   The ByteDance crawler (Bytespider) receives a 403 error, preventing its content from being indexed. While not a major AI crawler, this may limit visibility in certain AI products.

   What to change: If desired, allow Bytespider by adjusting Cloudflare WAF rules or robots.txt.

8. Robots.txt lacks AI-bot-specific directives Low

   The robots.txt has only a single User-agent: * rule with no disallows and no specific directives for AI crawlers. While this allows full access, it misses the opportunity to guide crawlers to the most important content or to set crawl rate limits.

   What to change: Consider adding specific directives for AI crawlers, such as allowing access to key pages and optionally setting crawl-delay.

What's working

• All major AI crawlers receive full HTML content — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and others get a 200 with full HTML identical to the browser baseline, ensuring AI assistants can index the site's content.
• Anthropic domain verification token present — DNS TXT records include an Anthropic domain verification token, confirming the Claude integration is production-ready and trusted by Anthropic.
• OpenAI domain verification token present — An OpenAI domain verification token is also present in DNS records, indicating readiness for potential OpenAI integrations.
• Press room features coverage from top-tier outlets — The press page lists coverage from TechCrunch, WSJ, WIRED, Fox Business, CIO.com, and Nature, providing strong external validation and backlinks.
• Blog is actively published with multiple posts per month — The blog contains frequent posts through May 2026, signaling an active content strategy that can attract AI crawlers and provide fresh material for indexing.
• Homepage includes Organization schema with social profiles — The homepage carries Organization JSON-LD with logo, URL, and social profile links, helping AI assistants correctly identify the company.
• Blog posts use BlogPosting schema with dates — Blog articles include BlogPosting schema with datePublished and dateModified, which helps AI assistants understand content freshness.
• Sitemap.xml is well-formed with 676 URLs — The sitemap is properly structured and contains a comprehensive list of URLs, aiding crawler discovery.