March 18, 2026 · 5 min read
OpenAI builds AI but scores 23/110 on AI discoverability. Stripe leads at 68/110. Here's what the data reveals about the state of AEO.
We used our AEO Checker to scan four of the biggest names in tech — Stripe, Shopify, Cloudflare, and OpenAI — across 7 dimensions of AI agent discoverability. Each site gets a score out of 110.
The question: are the companies building AI actually optimized for AI?
The answer might surprise you.
| Site | Score | Grade |
|---|---|---|
| shopify.com | 71/110 | Good |
| stripe.com | 68/110 | Good |
| cloudflare.com | 54/110 | Fair |
| openai.com | 23/110 | Poor |
The winner. Shopify nails the fundamentals: perfect structured data, a comprehensive llms.txt, and blazing performance. Where it falls short: no markdown content negotiation and incomplete robots.txt rules for AI bots. Their llms.txt file is the gold standard — detailed, structured, and clearly designed for LLM consumption.
Close second. Stripe's developer-first DNA shows: perfect structured data, a solid llms.txt, and fast responses. The gap vs. Shopify comes from content structure — Stripe's homepage is more visual and less heading-rich. Notable: despite having one of the most AI-integrated APIs in the world, no ai-plugin.json or OpenAPI spec is discoverable from the homepage.
Interesting split. Cloudflare is the only site that scores on "Markdown for Agents" — they pioneered the Accept: text/markdown content negotiation that lets AI agents get clean markdown instead of HTML. But they don't have a llms.txt file, which is ironic since their blog literally wrote the playbook on it. Cloudflare walks the walk on the engineering side but hasn't adopted the community standards yet.
The shocker. OpenAI — the company that started the AI agent revolution — scores 23/110. Zero structured data. Zero llms.txt. Zero content structure. Their homepage is essentially a sleek marketing splash page with minimal semantic markup. The only points come from performance (fast) and a partial robots.txt (GPTBot is mentioned, but other bots aren't explicitly allowed).
The 2 sites that scored "Good" (Shopify, Stripe) both have comprehensive llms.txt files. The 2 that didn't score well don't. A single file — 500+ characters of structured context about your product — is worth up to 15 points. It's the highest-ROI AEO action you can take.
JSON-LD schemas, OpenGraph tags, and meta descriptions account for 20 of 110 points. Shopify and Stripe both score perfect 20/20. OpenAI scores 0. When an AI agent processes your page, structured data is the first thing it extracts.
Only Cloudflare supports Accept: text/markdown — the emerging pattern where AI crawlers request markdown instead of HTML. This is still early, but Cloudflare's blog posts demonstrate a 40% improvement in AI comprehension when serving markdown. The others don't support it at all.
All four sites score only 8/15 on robots.txt. They mention some AI bots but don't explicitly allow all major crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended). In a world where new AI crawlers appear monthly, explicit allowlisting beats hoping for the best.
Despite Stripe having one of the most powerful APIs in fintech, there's no discoverable ai-plugin.json or OpenAPI spec from their homepage. Same for all four sites. AI agents can't use tools they can't find. If you have an API, /.well-known/ai-plugin.json should exist.
AEO (Agent Engine Optimization) is where SEO was in 2005: almost nobody is doing it, the standards are still forming, and early movers will have an enormous advantage.
The companies building AI aren't optimizing for AI. That gap is your opportunity.
Three things you can do today:
Scan your site for free. 7 checks, 110 points, actionable recommendations.
Scan Your Site →Methodology: All scans performed on March 18, 2026 using the AEO Checker tool (7 checks, 110 max points). Scores reflect homepage-level checks only. These companies may have better AEO on subpages (especially docs).