The GEO leaderboard: who's winning AI visibility in 2026

Agent Integration scored 0/13 across all 118 domains in every industry category. Zero. Not one company in our dataset has deployed agent-specific endpoints, structured tool responses, or machine-readable capability declarations. The agentic web is being built, but the companies that will populate it aren't preparing for it.

The expanded leaderboard

118 domains across 12 industries. Here are the top 10; the full dataset includes F-grade domains like hm.com (F/14), adidas.com (F/15), and freshworks.com (F/17).

Industry benchmarks

Pattern 1: Bot access is solved — AI visibility isn't

Every domain achieved the maximum 75/75 on Bot Access. Technical infrastructure is no longer the constraint. But AI Visibility ranges 12–18/68 median — LLMs can reach your content but don't know what it means. The gap between accessibility and understandability is 57 points wide.

Pattern 2: AI companies are getting worse, not better

OpenAI dropped to D/57 (was C/62). Anthropic fell to D/48 (was C/52). Vercel crashed from #1 at B/82 to #14 at C/70 — partly because Vercel's own robots.txt now includes 'ai-train=no', an explicit AI training block that penalises its score. These are the companies building the models — yet they're among the worst at making their own content discoverable to those same models.

Pattern 3: The within-category canyon

In Sportswear: Nike B/80 vs Adidas F/15 — a 65-point gap. In B2B SaaS: Twilio B/86 vs Freshworks F/17 — a 69-point gap. Nike and Twilio built API-first infrastructure; Adidas and Freshworks use marketing-page-first architectures. The structural choice, made years before GEO existed, determines whether you're invisible to agents.

What separates B from F

• Agent integration endpoints (0/13 across all 118 domains) — no company has started; first mover advantage is completely unclaimed
• Entity linking to knowledge graphs — Wikipedia/Wikidata sameAs on Organisation schema; GEO Presence median is 18–30/46
• llms.txt implementation — AI Discovery plateaus at 12–15/20 median without it (a property of our rubric, which weights llms.txt, rather than of measured crawler behaviour); Shopify has it, Anthropic doesn't
• Render gap — AI Visibility at 12–18/68 reflects JavaScript-heavy pages that serve thin HTML shells to crawlers

What nobody has done yet

No domain scored above 0 on Agent Integration. The category covers /.well-known/agent-card.json (A2A protocol), /.well-known/oauth-protected-resource (RFC 9728), and MCP discovery endpoints. These are a day's work each. Zero companies have shipped any of them. Being the first in your industry category to hit 1/13 is a genuinely differentiating signal right now.

Where to start if you're C, D, or F

1. Deploy llms.txt at your root with current sitemap references — worth 12–15 points on THIS scorecard. That is a score movement, not a measured gain in AI visibility: our rubric rewards llms.txt for rarity, and no published evidence yet ties it to retrieval or citation. Cheap to add; do not treat it as a growth lever. Check llmstxt.org for current format.
2. Add entity links: Wikipedia and Wikidata sameAs on your Organisation and Product schema — boosts GEO Presence from the 18–30 range toward the 40s.
3. Implement structured product schema if you sell anything — Commerce scores are 0–2/2 across all 118 domains; this is table stakes.
4. File agent integration as a roadmap item — /.well-known/agent-card.json (A2A) and /.well-known/oauth-protected-resource (RFC 9728) are the next frontier. First movers will hold these positions.

Run your domain at hidden-layer-blogs.pages.dev to see where you rank.