How AI reads Stripe

stripe.com Jun 8, 2026 6 min read Basic Web Presence

Short

34/ 100

AEO Level 2Basic Web Presence

The short answer

We scanned Stripe, the global payments infrastructure company trusted by millions of businesses worldwide, and measured an AEO score of 34 out of 100 — level 2. Structured data is the clear bright spot at 80, but agent interfaces and identity auth both sit at zero, meaning AI agents cannot discover or interact with Stripe programmatically.

What AI sees

When an AI agent visits Stripe's homepage today, it finds solid JSON-LD markup but almost no infrastructure built for agent-to-agent interaction.

Stripe's homepage delivers clean semantic HTML and strong structured data — the 80 score in that category reflects real investment in schema markup. An AI crawler can extract core business identity, product categories, and pricing tiers from the page. That is roughly where the good news ends. There is no sitemap.xml (the endpoint returns 404), so agents cannot enumerate the full content surface. The robots.txt file is silent on AI crawlers — GPTBot, ClaudeBot, and PerplexityBot receive no explicit guidance. No RFC 8288 Link headers point agents toward useful resources. For a company whose core product is an API, the absence of a discoverable API catalog is the most striking gap: agents querying about payment integration receive incomplete, surface-level answers.

Stripe homepage captured by the AIPUSH AEO scanner — JSON-LD structured data scores 80 but no agent discovery headers are present

Stripe robots.txt as retrieved during the scan — no AI crawler user-agent rules or Content-Signal directives found

Where it loses points

Agent interfaces and identity auth both score zero — Stripe publishes no MCP server card, no API catalog, and no agent skills index, leaving the agentic layer completely dark.

How to fix it

Three targeted additions would materially close Stripe's AI-readiness gap and position it as a first-class citizen in agentic commerce stacks.

Publish an MCP Server Card

Goal

Expose an MCP Server Card at the well-known path so AI agents can discover Stripe's capabilities and transport endpoint automatically.

Issue

No MCP server card exists at /.well-known/mcp/server-card.json, leaving agentic clients with no structured way to negotiate a connection.

Fix

Create a JSON file at /.well-known/mcp/server-card.json declaring serverInfo, transport type, and capability scopes such as payments.create and subscriptions.read. This single file unlocks Stripe as a first-party callable tool for any MCP-compatible AI agent or orchestration framework.

Add an RFC 9727 API Catalog

Goal

Publish a machine-readable catalog of Stripe's APIs so automated discovery tools and AI agents can find and understand every integration point.

Issue

No API catalog exists at /.well-known/api-catalog, which is especially damaging for a developer-first company whose entire value proposition is its API surface.

Fix

Serve /.well-known/api-catalog as application/linkset+json with a linkset array covering the core APIs — Payments, Billing, Connect, and Treasury at minimum — and include links to OpenAPI specifications so agents can resolve schemas automatically.

Declare AI crawler rules in robots.txt

Goal

Give AI crawlers explicit, per-user-agent directives so Stripe controls exactly which content trains models and which content feeds answer engines.

Issue

The current robots.txt contains no rules for GPTBot, ClaudeBot, PerplexityBot, or any other AI crawler, creating a policy vacuum that leaves citation and training behavior up to each crawler's defaults.

Fix

Add named user-agent stanzas for every major AI crawler with Allow directives for documentation and pricing pages Stripe wants cited. Append a Content-Signal directive declaring preferences for ai-train, search, and ai-input so aggregators respect the intent.

Common questions

Why does Stripe score only 34 out of 100 on AEO despite being one of the most developer-friendly platforms on the internet?

Stripe's structured data is genuinely excellent — schema markup scores 80 — so the technical foundation is solid. The drag comes entirely from the agent-facing layer: no MCP server card, no API catalog, no AI crawler policy, and a missing sitemap. These gaps are fixable in days but collectively cut the overall score almost in half.

Does the missing sitemap.xml affect how AI answer engines respond to questions about Stripe's products and pricing?

Yes. Without a sitemap, AI crawlers rely on link-following alone to discover content, which means documentation pages, pricing tiers, and integration guides may be missed or indexed incompletely. Answer engines that cannot confirm canonical URLs tend to cite with lower confidence, which reduces Stripe's presence in AI-generated financial and developer answers.

How would publishing an MCP server card help Stripe in a world where AI agents execute financial workflows autonomously?

An MCP server card makes Stripe natively callable by any MCP-compatible AI agent — think expense automation, invoice reconciliation, or subscription management driven by a user's AI assistant. Without it, agents must rely on unstructured documentation scraping, which is slower, error-prone, and increasingly unsupported as the agentic commerce ecosystem matures.