AI Site Grade
stripe.com — AI Site Grade
Stripe's documentation site serves AI crawlers a 2-5KB plain-text shell while browsers receive 400-1200KB of rich HTML, severely limiting AI extraction of its most authoritative technical content.
Stripe's AI visibility is strong on the main site but critically undermined by content-starvation on docs.stripe.com, a missing llms-full.txt, and stale cold knowledge about its AI narrative.
- Findings
- 9
- Evidence checks
- 22
- Completed
- 2 Jul 2026
Analysis
Stripe's AI-Visibility Audit
The most striking finding is that Stripe's documentation — arguably the most valuable content for AI training and retrieval — is served as a thin 2-5KB plain-text shell to every major AI crawler, while browsers receive 400-1200KB of rich HTML. This is a deliberate server-side UA detection that strips the docs site of its full content for AI bots, despite the robots.txt explicitly allowing all crawlers.
Crawler Access
On stripe.com, all 11 tested AI bots (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, anthropic-ai, Bytespider, Applebot-Extended, Perplexity-User) receive a full 200 response with ~585KB of content — identical to a browser. The robots.txt at stripe.com/robots.txt contains zero AI-specific directives; no GPTBot, ClaudeBot, or Google-Extended rules exist. The wildcard rule (User-agent: *) allows /docs and /docs$ while disallowing a handful of utility paths. The llms.txt at stripe.com/llms.txt is present and substantial (~64KB), offering a structured, link-rich content map across all product lines — a rare and commendable implementation. However, the referenced llms-full.txt returns a 404 (JS-rendered error page), breaking the promise of a single-file documentation dump.
The Docs Subdomain Gap
On docs.stripe.com, the robots.txt uses a non-standard Content-Signal: ai-train=yes, search=yes, ai-input=yes directive — an experimental signal that most crawlers do not parse. Despite this permissive stance, every AI bot receives only 2-5KB of plain text (headings and link lists) while a browser gets 400-1200KB of full HTML with code examples, interactive API explorers, and prose. For example, docs.stripe.com/payments returns 5.3KB to GPTBot versus 434KB to a browser. This is not a block — it is a content-starvation pattern that severely limits what AI models can extract from Stripe's most authoritative technical content.
Cold-Knowledge Gap
The LLM's prior knowledge about Stripe is strong and accurate — co-founders, valuation, core products (Payments, Connect, Billing, Atlas, Terminal, Issuing, Radar), and even the 2023 layoffs. However, the cold knowledge is stale on recent product expansions: it mentions nothing about Stripe's aggressive push into agentic commerce, stablecoin/crypto infrastructure, AI agent payments, or the $1.9T in 2025 payment volume prominently featured on the homepage. The site now positions itself as "building the economic infrastructure for AI" — a narrative entirely absent from the model's prior.
Schema Posture
The homepage carries rich Organization JSON-LD with founders, locations, social profiles, and contact info — but zero WebSite schema with potentialAction (SearchAction) for site search, and no SoftwareApplication schema for any of the products. The /pricing page uses FAQPage schema with 5 well-structured Q&A entries. The /customers and /newsroom pages carry no schema at all. Given Stripe's developer audience, the absence of TechArticle or HowTo schema on documentation pages is a missed signal for AI answer extraction.
External Signals
Stripe's DNS TXT records reveal verification tokens from Anthropic, Apple, Cursor, ElevenLabs, Docker, Atlassian, and Canva — indicating deep partnerships with AI-native companies. The homepage prominently features case studies with Hertz, URBN, Instacart, Le Monde, and AI-native startups (Lovable, Gamma, Runway, Supabase, ElevenLabs, Browserbase, Decagon). The newsroom is JS-rendered (only 17 words extracted via plain GET), meaning AI crawlers get almost no content from Stripe's own press announcements.
Findings
AI crawlers receive stripped-down plain text on docs.stripe.com High
Every major AI bot receives only 2-5KB of plain text (headings and links) from docs.stripe.com, while browsers get 400-1200KB of full HTML with code examples and interactive content. This content-starvation pattern severely limits what AI models can extract from Stripe's most authoritative technical documentation.
What to change: Remove the server-side user-agent detection that truncates content for AI crawlers on docs.stripe.com, or serve the full HTML to all bots.
llms-full.txt returns a 404 error High
The llms.txt file at stripe.com/llms.txt references an llms-full.txt file, but that URL returns a 404 (JS-rendered error page), breaking the promise of a single-file documentation dump for AI consumption.
What to change: Create and serve a valid llms-full.txt file with the full documentation content, or remove the reference from llms.txt.
LLM cold knowledge lacks Stripe's AI narrative Medium
The LLM's prior knowledge about Stripe is strong but stale on recent product expansions: it mentions nothing about agentic commerce, stablecoin/crypto infrastructure, AI agent payments, or the $1.9T in 2025 payment volume. The site's positioning as 'building the economic infrastructure for AI' is entirely absent from the model's prior.
What to change: Ensure that key AI-related product pages and announcements are crawlable and contain clear, structured content that AI models can index.
Newsroom page is JS-rendered, yielding minimal content for AI crawlers Medium
The newsroom page at stripe.com/newsroom returns only 17 words via plain GET, indicating heavy JavaScript rendering. AI crawlers that do not execute JS receive almost no content from Stripe's own press announcements.
What to change: Implement server-side rendering or provide a static fallback for the newsroom content to ensure AI crawlers can access press announcements.
Homepage lacks WebSite schema with SearchAction Medium
The homepage carries rich Organization JSON-LD but no WebSite schema with potentialAction for site search, which helps AI models understand site navigation and search capabilities.
What to change: Add WebSite schema with SearchAction to the homepage and other key pages.
Product pages lack SoftwareApplication schema Medium
Despite offering numerous software products (Payments, Connect, Billing, etc.), no product page uses SoftwareApplication schema, which would help AI models classify and extract product details.
What to change: Add SoftwareApplication schema to product pages with appropriate properties (name, description, applicationCategory, etc.).
Customers and newsroom pages carry no schema markup Low
The /customers and /newsroom pages have no structured data, missing opportunities for rich snippets in AI-generated answers.
What to change: Add appropriate schema (e.g., Article for newsroom, ItemList for customers) to these pages.
docs.stripe.com robots.txt uses non-standard Content-Signal directive Low
The robots.txt at docs.stripe.com includes a non-standard 'Content-Signal: ai-train=yes, search=yes, ai-input=yes' directive that most AI crawlers do not parse, potentially causing confusion.
What to change: Replace the non-standard directive with standard User-agent and Allow/Disallow rules, or ensure the directive is widely supported.
Main robots.txt has no AI-specific directives Low
The robots.txt at stripe.com contains no rules for GPTBot, ClaudeBot, Google-Extended, or other AI crawlers, relying solely on the wildcard rule. While this currently allows access, it provides no fine-grained control.
What to change: Consider adding explicit directives for major AI crawlers to ensure intended access and prevent future accidental blocks.
What's working
- llms.txt file is present and substantial — Stripe hosts a ~64KB llms.txt at stripe.com/llms.txt, offering a structured, link-rich content map across all product lines — a rare and commendable implementation that helps AI crawlers discover key resources.
- All AI crawlers receive full content on stripe.com — All 11 tested AI bots receive a full 200 response with ~585KB of content on stripe.com, identical to a browser, ensuring the main marketing site is fully visible to AI models.
- Pricing page uses FAQPage schema — The /pricing page includes FAQPage schema with 5 well-structured Q&A entries, helping AI models extract pricing-related answers directly.
- Homepage carries rich Organization JSON-LD — The homepage includes Organization schema with founders, locations, social profiles, and contact info, providing authoritative entity data to AI models.
- LLM has strong prior knowledge about Stripe — The LLM's cold knowledge accurately covers co-founders, valuation, core products (Payments, Connect, Billing, Atlas, Terminal, Issuing, Radar), and even the 2023 layoffs, indicating strong baseline visibility.
- DNS TXT records show verification tokens from AI-native companies — Stripe's DNS TXT records include verification tokens from Anthropic, Apple, Cursor, ElevenLabs, Docker, Atlassian, and Canva, indicating deep partnerships with AI-native companies that may boost external signals.
- Homepage features case studies with AI-native startups — The homepage prominently showcases case studies with AI-native companies like Lovable, Gamma, Runway, Supabase, ElevenLabs, Browserbase, and Decagon, reinforcing Stripe's AI ecosystem positioning.
Track stripe.com across AI search
This is one snapshot. Open the interactive report to inspect evidence, or grade another site free.