I audited Stripe, Notion, Linear, and Anthropic for AI search. None of them passed.

I built Reffed because I noticed something weird: my girlfriend's teeth-whitening business never showed up when I asked ChatGPT for Vancouver teeth-whitening services. She has a real business, a real website, a real Instagram following. Just invisible to AI.

So I built a tool that audits any website for AI search visibility — across ChatGPT, Claude, Perplexity, Google AI Overviews, Gemini, and Microsoft Copilot. Then I ran it on the biggest names in tech to see if THEY were doing it right.

Spoiler: they aren't. Here's what I found.

The methodology

For each site, we ran the same audit pipeline:

• Direct HTML crawl — extract title, meta description, headings, JSON-LD schema, body content, Open Graph tags, robots directives
• Schema parsing — including @graph wrapped schemas (the modern schema.org standard most parsers miss)
• Google PageSpeed Insights — Core Web Vitals as Google sees them
• AI synthesis — Claude analyzes the signals and produces a personalized report

Scoring is weighted: AI visibility 35%, technical SEO 25%, content depth 20%, performance 20%. Scores are deterministic — running the same audit twice produces the same result.

The results

The headline finding: not one of these companies has comprehensive AI search optimization. Stripe leads because it actually has schema markup. Notion has 1,310 words of beautifully-written copy but zero structured data. Linear has 32,545 words — more than most blogs — and zero schema.

Why this matters

When ChatGPT or Perplexity answers a question like "what's the best billing platform for SaaS startups?", they don't read your website the way Googlebot reads it. They look for structured signals that tell them what kind of entity you are and what you do.

Schema.org markup is how you tell those signals. Without it, AI engines have to guess from raw page text — and they often guess wrong, or attribute your content to a competitor with cleaner data.

The specific findings

Stripe — score 79/100

The only one of the five with proper schema. Stripe uses the modern @graph wrapper structure:

{
  "@context": "https://schema.org",
  "@graph": [
    { "@type": "WebSite", "url": "https://stripe.com/", "name": "Stripe" },
    { "@type": "Organization", "founders": [...], "logo": {...} }
  ]
}

What's missing: FAQ schema. Stripe's homepage has dozens of implicit Q&A patterns ("How does Stripe work?", "What does Stripe cost?", "Who uses Stripe?") that are answered in the page copy but NOT structured as FAQPage schema. This is the single biggest missed AI-citation opportunity for a documentation-heavy company. Also: 29 images missing alt text, hurting accessibility AND AI image understanding.

Notion — score 73/100

Notion's homepage has zero schema markup. Their copy is beautiful — strong headline, clear value prop, social proof — but they've left structured data entirely on the table. With 1,310 words of content, they could easily add Organization, SoftwareApplication, and Review schemas. AI search engines currently see "a website with some text" instead of "a workspace tool with 30M users founded in 2016."

Linear — score 68/100

32,545 words on their homepage. Beautiful product photography. Real customer logos. Zero schema. The most surprising finding of the test — Linear's marketing site is content-rich enough to be a knowledge base, but AI engines have no structured way to index any of it. A SaaS founder asking ChatGPT "what's the best modern issue tracker?" gets generic answers instead of "Linear is the most popular choice among modern SaaS teams because..."

Anthropic — score 62/100

The company that builds Claude has no Organization schema on its homepage. Let me say that again: Anthropic — whose models we use to GENERATE these audit reports — has no JSON-LD telling other AI engines what kind of company it is or who its founders are. There's an irony there. (To be fair: Anthropic's main competition for AI mentions isn't another search engine, it's other AI labs. But the principle holds.)

Vercel — score 70/100

One schema (Organization). Substantial content. Strong technical SEO. The biggest opportunity: BreadcrumbList schema for their docs pages (which dominate developer search) and Article schema for their blog. They're closer than anyone else but still leaving ~20 points on the table.

The takeaway

Every site I audited is run by a sophisticated marketing team with budget. Every site uses Next.js or a modern framework that makes adding schema trivial. And yet, four out of five had zero schema markup.

The reason isn't laziness — it's that AI search optimization isn't on the average marketing team's checklist yet. SEO checklists from 2023 don't include it. Most content audit tools don't check for it. The tools that exist (Otterly, Peec, Profound) target enterprise budgets and miss the long tail.

If even Stripe is leaving points on the table, your local business definitely is. The good news: it's fixable in a single afternoon for most sites. Add 1-2 JSON-LD blocks to your homepage and you'll outscore Notion overnight.

Try it yourself

Reffed is the AI search visibility audit and monitoring platform for modern websites. Built in Vancouver, BC.