What Top Agent-Traffic Sites Still Need for Agentic Commerce (Including Grok Bot)
Robauto Team · Published · Last updated
HTML edition with tables and JSON stats: Robauto’s top-~100 readiness median is 18/100. Live scans of marketontology.com, amazon.com, and robauto.ai — plus a Grok Bot–ready checklist with source links.
Robauto’s public Brand Leaderboard ranks sites by agent visits. Separately, the Agent Readiness Index runs a 25-check scan across the most-visited sites on that network. The headline from the latest index run is simple: agents are already showing up, but most of those sites still score in the teens on the surfaces that make agentic commerce possible.
This post is not a roast. Plenty of operators deliberately keep AI crawlers and shopping agents out — that is a legitimate product and risk decision. The sharper question: if you want Grok Bot, ChatGPT agents, Copilot, and other assistants to cite you, support customers, or eventually checkout on your behalf, what should a top-traffic site actually publish?
What the top-100 pulse shows
As of September 19, 2026, Robauto’s Brand Leaderboard (sorted by 7-day agent visits) is led by sites such as marketontology.com, thearknewspaper.com, and topleftdesign.com. The long tail includes publishers, SaaS brands, retailers, and infrastructure companies.
| Metric | Value | Source |
|---|---|---|
| Domains attempted | 99 | Agent Readiness Index |
| Scored | 79 | Agent Readiness Index |
| Blocked | 20 | Agent Readiness Index (block = access, not quality) |
| Median score | 18 / 100 | Agent Readiness Index |
| Mean score | 17 / 100 | Agent Readiness Index |
| Min / Max | 2 / 29 | Agent Readiness Index |
| Index last run | 2026-09-02 | Agent Readiness Index |
{
"agent_readiness_index": {
"attempted": 99,
"scored": 79,
"blocked": 20,
"median": 18,
"mean": 17,
"min": 2,
"max": 29,
"last_run": "2026-09-02",
"source": "https://robauto.ai/agent-readiness-index"
}
}
| Rank | Domain | Agent visits (7d) | Leaderboard |
|---|---|---|---|
| 1 | marketontology.com | 5,130 | Brand Leaderboard |
| 2 | thearknewspaper.com | 1,874 | Brand Leaderboard |
| 3 | topleftdesign.com | 1,823 | Brand Leaderboard |
So even among sites that already attract agent traffic, readiness for machine-readable identity, MCP/A2A discovery, and checkout protocols is still early.
Two leaderboard scans, two intentional postures
We re-ran live Agent Scans on September 19, 2026. Treat these as snapshots, not verdicts.
| Domain | Score | Signal Basics | Technical | Agent Integration | Agentic Commerce | Live scan |
|---|---|---|---|---|---|---|
| marketontology.com | 18 | 4/6 | 0/4 | 0/10 | 0/5 | Open scan |
| amazon.com | 20 | 2/6 | 0/4 | 0/10 | 0/5 | Open scan |
| robauto.ai (reference) | 98 | 6/6 | 4/4 | 9/10 | 5/5 | Open scan |
{
"live_scans_2026_09_19": [
{"domain": "marketontology.com", "score": 18, "tiers": {"signal_basics": "4/6", "technical": "0/4", "agent_integration": "0/10", "agentic_commerce": "0/5"}, "scan": "https://robauto.ai/scan?url=marketontology.com"},
{"domain": "amazon.com", "score": 20, "tiers": {"signal_basics": "2/6", "technical": "0/4", "agent_integration": "0/10", "agentic_commerce": "0/5"}, "scan": "https://robauto.ai/scan?url=amazon.com"},
{"domain": "robauto.ai", "score": 98, "tiers": {"signal_basics": "6/6", "technical": "4/4", "agent_integration": "9/10", "agentic_commerce": "5/5"}, "scan": "https://robauto.ai/scan?url=robauto.ai"}
]
}
marketontology.com — #1 by agent visits, score 18
Passes Signal Basics many sites still miss: HTTPS, an open robots.txt for AI agents, a sitemap, and an llms.txt brief. That helps explain leaderboard visibility.
Still missing for agentic commerce: server-rendered homepage text for non-JS crawlers (scan reported 0 words in raw HTML), Link headers / API catalog / OpenAPI, agent card and MCP server card, Content-Signals, and every Agentic Commerce check (x402, ACP, UCP, AP2, MPP). Label: “Early — wide open upside.” Full report: robauto.ai/scan?url=marketontology.com.
amazon.com — on the leaderboard, score 20, deliberate robot policy
The scan finds HTTPS and some server-rendered text, and Robauto pixel traffic is present. It also reports that robots.txt blocks GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and OAI-SearchBot site-wide — so those agents cannot cite the site through that policy.
That is not a failure of “not trying hard enough.” It is a large retailer choosing which agents may read the storefront. If your strategy is closed-by-default for generative crawlers, your readiness score will stay low by design — and that may be exactly what you want. Full report: robauto.ai/scan?url=amazon.com.
robauto.ai — reference implementation, score 98
For contrast, Robauto’s own domain scores 98 (“Fully agent-ready & verified”): Signal Basics, Technical Groundwork, and Agentic Commerce all complete; Agent Integration nearly complete (one OAuth discovery warning). That includes llms.txt, OpenAPI, A2A agent card, MCP server card, Content-Signals, x402, ACP, UCP, AP2, and MPP. Use it as a checklist when a site wants agents to both read and transact. Report: robauto.ai/scan?url=robauto.ai.
Why Grok Bot belongs in this conversation
On Robauto’s live AI search dataset (90-day window, updated hourly), grok_bot accounts for about 10.8% of classified AI-agent visits — second only to Meta AI in that mix. Network-wide AI agent views are about 20.26% of page views over 90 days.
| Engine / class | Share of AI visits (90d) | Source |
|---|---|---|
| Meta AI | 31.2% | AI Search Data |
| grok_bot | 10.8% | AI Search Data |
| baiduspider | 9.7% | AI Search Data |
| Claude | 3.4% | AI Search Data |
| Microsoft Copilot | 3.2% | AI Search Data |
| ChatGPT | 2.2% | AI Search Data |
| Grok / xAI | 1.7% | AI Search Data |
| All AI agent views | 20.26% of total page views | AI Search Data |
{
"ai_search_90d": {
"ai_traffic_percentage": 20.26,
"ai_agent_views": 251177,
"total_page_views": 1239904,
"top_engines": [
{"engine": "Meta AI", "share_percentage": 31.2},
{"engine": "grok_bot", "share_percentage": 10.8},
{"engine": "baiduspider", "share_percentage": 9.7}
],
"source": "https://robauto.ai/ai-search-data"
}
}
If Grok Bot and peer assistants are already in the traffic mix, sites that want them to help with research, support, or purchases need machine-readable inventory, policy, and — when commerce is the goal — a checkout rail agents can discover.
A constructive ladder for top sites that want agents
| Step | Tier | Ship this | Spec / docs |
|---|---|---|---|
| 1 | Policy | Decide robots.txt posture on purpose (welcome or block) | Content Signals |
| 2 | Signal Basics | HTTPS, sitemap, llms.txt, server-rendered text, metadata | llms.txt |
| 3 | Technical Groundwork | Link headers, api-catalog, OpenAPI, optional Markdown negotiation | RFC 9727 |
| 4 | Agent Integration | A2A agent card, MCP server card, skills, auth | A2A, MCP |
| 5 | Agentic Commerce | ACP, UCP, x402, AP2/MPP when you mean to sell | ACP, UCP, x402 |
| 6 | Verification | Robauto pixel for the final identity points | Agent Scan |
You do not need every protocol on day one. You do need an explicit path from “agent visits the site” to “agent can complete a supported task.”
How to read the leaderboard without weaponizing it
- Discoverable but not transactable — fix Signal Basics, then add an agent card / MCP / commerce surface.
- Closed on purpose — keep the robot policy; ignore vanity scores.
- Mid-migration — prioritize server-rendered inventory and one clear agent entrypoint.
The index maximum of 29 among scored top sites (September 2 run) shows how early this layer still is. Being #1 on agent visits is not the same as being ready for agentic commerce — and that is an opportunity, not an insult.
What to do this week
- Open robauto.ai/leaderboard and note whether you appear.
- Run robauto.ai/scan?url=yourdomain.com (or the
agent_scanMCP tool) and read the four tiers. - If you want Grok Bot and peers to help customers, fix Signal Basics first, then publish an agent card or MCP server card.
- If you do not want agents, tighten robots.txt deliberately and move on — a low score in that case is policy working as designed.
Bottom line
Across Robauto’s ~100 most-visited sites, Agent Readiness still clusters near 18/100, while agent traffic on the same network is already material. Leaders on the visit chart often have partial Signal Basics and little or no agentic commerce surface. Sites that intentionally block AI crawlers will look “unready” in this framework — and that can be the right call. Sites that want Grok Bot and agent checkout should treat the 25-check ladder as a build backlog: invite the agents you want, publish machine-readable inventory, then open a commerce protocol when you are ready to sell.
Sources
- Robauto Brand Leaderboard
- Robauto Agent Readiness Index
- Agent Scan — marketontology.com
- Agent Scan — amazon.com
- Agent Scan — robauto.ai
- Robauto AI Search Data
- llms.txt
- Model Context Protocol
- A2A Protocol
- Agentic Commerce Protocol
- UCP overview
- x402 docs
- RFC 9727
- Content Signals
Where can an agent get this data live?
Robauto MCP — Agent Intelligence & Brand Discovery Infrastructure. A public, stateless Streamable HTTP endpoint at https://mcp.robauto.ai/mcp, callable with no key. Setup and the full tool list are in the Robauto MCP directory.