Digital agency Stradiji has released an AI visibility study and audit that concludes most websites effectively close their doors to large language models. That is the headline claim. The short version for operators: if the finding holds, most sites are opting out — deliberately or by accident — of the systems that increasingly decide which brands AI assistants mention.

The announcement itself is thin on the numbers a buyer would want. Stradiji reports that a majority of audited websites restrict LLM access, but the public write-up does not specify how many sites the audit covered, which sectors they came from, or over what period the data was collected. Without a sample size, "most" could mean 51 percent of 40 sites or 90 percent of 4,000. The gap matters.

What does the study actually claim?

The core claim is simple. Most websites, when Stradiji's audit checked them, did not allow LLM crawlers or AI systems full access to their content. In practice, that usually means robots.txt rules, technical blocks, or site architectures that AI models cannot read.

The commercial logic behind the study is not hidden. Stradiji sells visibility work — auditing how sites appear, or fail to appear, in AI-generated answers. A study that finds widespread blocking creates a natural pipeline to the agency's paid services. That does not make the finding wrong; it makes it unverified by anyone with money on the other side of the conclusion.

Why should operators care?

Search behavior is shifting. A growing share of product and brand questions now get answered directly by AI assistants rather than by a page of blue links. If a site blocks LLM crawlers, it will not appear in those answers. For an e-commerce operator, that is the difference between being recommended and being invisible.

There is a counter-argument the study does not settle. Some publishers block AI crawlers on purpose — to protect content, control licensing, or preserve ad revenue from human visitors. Blocking is not always an oversight. The audit's framing, at least as publicly summarized, does not distinguish between sites that block LLMs by accident and sites that made a deliberate business choice to do so.

What is missing from the announcement?

A reader trying to act on this study hits several walls:

• No stated sample size for the audit
• No breakdown by industry, site size, or geography
• No publication date range for the data
• No methodology section in the public summary
• No indication of whether the audited sites were Stradiji clients or independent samples

Each of these omissions is standard disclosure for a credible visibility study. Their absence is the story as much as the finding itself.

What can an operator do with this?

The practical takeaway survives the caveats. Operators who want AI assistants to mention their brand should check their own robots.txt and crawler rules before assuming the problem belongs to everyone else. An internal audit takes hours, not weeks. If your site already allows major LLM crawlers, Stradiji's "most sites block AI" finding is a competitive detail, not a threat.

If it does not, the decision to open or stay closed is a business call — content protection versus AI visibility — not a technical emergency.

Stradiji's study points at a real shift in how sites get discovered. Until the agency publishes its sample and method, treat "most websites close doors to LLMs" as a plausible claim from an interested party, not a settled measurement.

The door may well be closed. The audit that says so has not shown its own numbers.