AI agent accessibility
Find out whether agents can see you — and why they can't.
Infrastructure for agentic commerce
AI agents now read your catalogue before your customers do. Most platforms block them without knowing it, serve them an empty page, and measure nothing at all. Metavizor diagnoses, fixes and instruments.
An engineering firm led by practitioners: our consultants work in production, not in workshops.
Free test — 30 seconds, no sign-up
We query your page live using the user agents of GPTBot, ClaudeBot, PerplexityBot and Google, then show you exactly what they receive. Not an estimate: your server's actual response.
Accessibility score
This test detects blocks based on user agent and robots.txt. IP-range filtering — common with Datadome or Cloudflare — can only be observed in your own server logs. That is what the full audit is for.
Want to know what this test cannot see?
The full audit opens your server logs and WAF rules, template by template, and puts a cost on every fix.
23 Sept 2026
Google AI Mode rolled out in France
26%
of France's top 100 retailers have deployed an AI assistant
79%
of those assistants do nothing but redirect to a product page
The diagnosis
None of them show up in a marketing report. All of them are visible in your logs within the hour.
GPTBot, ClaudeBot, PerplexityBot and Google-Extended fall under the same Datadome, Cloudflare or Akamai rules as malicious scrapers. You are absent from generative answers, and nothing in your dashboards says so.
Agent crawlers do not execute your SPA. Where a browser shows a complete product page, they receive an empty container. The catalogue exists — it is simply unreadable.
Varnish, CDN, purge and variation rules: agent traffic breaks the hit-ratio assumptions your infrastructure was sized on, and drives origin load up with nothing in return.
LLM-driven traffic is segmented neither in your analytics nor in your CDN logs. With no numbers, no budget decision is possible — and the topic stays a corridor conversation.
Services
Areas 3 and 4 keep your production standing. Areas 1 and 2 make it readable by agents. They sell separately and reinforce each other.
Find out whether agents can see you — and why they can't.
Produce the number that unlocks the budget.
Hold up on the days that matter.
Use AI where it actually delivers: operations.
Method
We open your logs, your WAF rules and your templates. You get a factual report: what is blocked, on which templates, at what criticality, and what each fix costs. Usable as-is in a steering committee.
We execute the plan: access rules, server-side rendering of critical templates, structured data, feeds, instrumentation. We work in your repositories, with your teams, on your deployment pipelines.
The dashboard runs, alerts catch regressions, and a monthly review decides what comes next. The goal is that you no longer need us to observe — only to decide.
Questions
It is the most common objection, and a fair one. But blocking means disappearing: when a customer asks an agent where to buy something, only readable merchants get named. The right answer is neither "block everything" nor "open everything" — it is to choose: open the category and product pages that make you money, and close what has no business being there. That trade-off is decided rule by rule, and it is exactly what we work through.
One hour of reading server logs. Filter by user agent over the last thirty days and count the 403 and 429 responses. In most cases we have seen, the block had been in place for months without a single alert firing: bot protection does its job, it does not ask whether the decision is good for the business. The free test at the top of this page already gives you a first signal.
A Markdown text file at the root of your site describing your business and main pages for language models. It replaces neither robots.txt nor your sitemap: it complements them by giving context where the others only give permissions and addresses. It takes about an hour to put in place. Ours is at /llms.txt.
It is the second most common problem. Agent crawlers do not execute JavaScript: where your browser shows a complete product page, they receive an empty container. You do not need to rebuild everything — server-side rendering or pre-generating just the category and product templates is usually enough to restore readability.
This plays out in infrastructure, not in the CMS: reverse proxy, CDN, WAF rules, rendering layer. So we work platform-independently — Magento, Salesforce Commerce Cloud, PrestaShop, Shopify Plus, bespoke builds. What matters is your delivery chain, and we know how to read it.
The diagnostic takes five to ten days and gives you a costed picture immediately. Unblocking work shows an effect within days — the time it takes crawlers to come back. Server-side rendering and structured data take longer, typically two to eight weeks depending on your deployment pipeline.
A diagnostic takes five days. It tells you whether AI agents reach your catalogue, what stops them, and what the fix costs.
Request an audit