# 97% of Websites Are Invisible to AI Agents

SiteSpeakAI ran two analyses of the live web: can AI answer the questions your buyers ask, and can an AI agent act on your site at all. We scanned close to 10,000 websites. Both point the same way: the web was built for human eyes, not for AI.

## Method

Aggregates from SiteSpeakAI's public scanner tools, February to July 2026.

- Answerability: 9,846 unique websites, 52,158 model-graded answers (0-100). We trained a temporary AI assistant on each site and asked the questions a real buyer asks. Failed crawls scoring under 10 are excluded from the reported mean.
- Agent-readiness: 216 websites scored 0-100 across four weighted categories: content discoverability (30%), AI agent tools (30%), content quality (25%), technical readiness (15%).

Scores are model-assigned and are not a certification.

## Finding 1: The agentic web does not exist yet

Across 216 sites, the average score for AI agent tools was 5 out of 100. 97.4% of sites expose nothing an AI agent can act on: no WebMCP, no MCP endpoint, no documented API. The same sites average 98 out of 100 on technical basics like HTTPS, mobile and speed.

Category averages (0-100):

- AI agent tools: 5
- Content discoverability: 77
- Content quality: 65
- Technical basics: 98

The web aced the last era's checklist and has not started this one. That gap is the opening: the web is a blank field for whoever makes their site agent-actionable first.

## Finding 2: AI can describe you, but it cannot tell customers what you cost

Across 9,846 sites and 52,158 answers, AI answered buyer questions at a mean of 67 out of 100 on successfully-scanned sites. The average hides where it breaks. Ranked by what AI answered worst:

| Buyer question | Avg answer score | Sites tested |
| --- | --- | --- |
| Cancellation or refund terms | 33 | 1,340 |
| How do I subscribe or is there a newsletter | 38 | 3,300 |
| How much does it cost | 43 | 8,100+ |
| What integrations or APIs does it support | 45 | 375 |

What AI answered best: how it works, step by step (83), who it is for (82), and what makes it different (81).

AI has fully absorbed your marketing story and cannot reliably state your price, refund policy, or how to sign up: the exact facts a buyer needs to say yes.

## Finding 3: You are either legible to AI, or you are not

Plotting every answerability score gives two walls, not a bell curve. Roughly 37% of scans scored below 50 and 40% scored 75 or higher, with a thin valley between them and a large cluster near zero (sites AI could barely read at all). There is very little "sort of readable." A site is either structured so AI can lift clean answers out of it, or it is a wall of design and JavaScript that AI slides off.

## The two gaps are separate problems

Answerability and agent-readiness are not the same failing. A site can score well on answerability, with clean crawlable content, and still score zero on agent tools because it exposes no way for an agent to act. Getting AI to understand you and letting an agent use you are two different pieces of work. Most sites have done neither, a rare few have done the first, and almost none have done the second.

## Why this happens

Pricing lives in a JavaScript table AI cannot parse. Refund terms are buried in a PDF or a legal page nothing links to. There is no llms.txt, no structured data, nothing an agent can call. It is all there for a human who clicks around, and invisible to the AI summarising it for them.

## How to fix it

1. Put the commercial facts in plain, crawlable text: pricing, refund and cancellation terms, and how to start, as real HTML rather than a widget or PDF.
2. Publish an llms.txt so AI knows what your site is and where the key pages live.
3. Make your site agent-actionable. A SiteSpeakAI chatbot with public MCP enabled registers WebMCP tools automatically, so an agent can look up pricing, book, or hand off to a human instead of hitting a dead end.
4. Check where you stand with the free AI Score scanner that produced this data.
5. Track whether it is working. Getting AI-readable is step one; knowing whether AI assistants now mention and cite you is step two, which is what MentionScout measures across ChatGPT, Gemini, Perplexity and more.

Run the free scanner: https://sitespeak.ai/tools/ai-score-my-site