A scan of more than 1,500 organisations found that 94 per cent of websites could not accept an AI agent connection, according to Encore founder and chief executive Matthew O’Riordan’s Fortune commentary. The finding points to a growing gap between customers using assistants to complete online tasks and websites built to keep automated visitors out.
Watch Desk analysis
What happened
Encore says it modelled popular personal AI assistants across 25 industries, using public pages and no log-ins. The company says 26 per cent of sites blocked agents that identified themselves, while only 1 per cent offered customers a way to approve a limited task and withdraw access. Just 6 per cent had published a position on AI agents, according to the scan, dated 23 October.
O’Riordan argues that blanket bot defences can reject legitimate customer assistants alongside harmful automation. He recommends making rules machine-readable, offering a direct agent connection and letting customers delegate narrow tasks for limited periods. Encore builds AI agents for business interactions, so its results and recommendations come from a company with a stake in wider agent adoption.
Why it matters
If an assistant cannot complete a booking or comparison on one site, it may take the customer elsewhere without the business ever seeing the abandoned attempt. O’Riordan argues that this can also shape what an assistant learns to recommend in future. That is a plausible commercial concern, but this company-led scan measures access and published policies, not lost sales or long-term changes in recommendation behaviour.
The practical choice is not simply to let every bot through or slam the door. Businesses need ways to distinguish authorised customer tasks from unwanted traffic, while keeping customers in control of what they delegate.
Our read
The useful signal here is the mismatch: nearly all scanned sites were unprepared to accept an agent connection, and very few offered controlled delegation. The numbers deserve attention, not a victory lap for agent commerce. Encore’s published methodology and results make the claim testable; its commercial interest is worth keeping in view as the debate moves from access to adoption.
What to watch
- Whether Encore publishes the checks and weights behind its Agent Readiness Index in a form others can reproduce.
- Whether website operators offer customer-authorised, limited agent access rather than relying on broad bot blocks.
- Whether independent studies measure failed transactions and distinguish customer agents from harmful automation.
Discussion spark: Should businesses build a safe, limited way for customers’ AI assistants to act on their behalf, or keep automated agents out until the security case is clearer?
Sources and evidence
- Source update (27 October 2026, 12:30 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.