Zuko Launches 'Agent Score' to Test Whether AI Agents Can Complete Your Forms and Checkouts
Key Takeaways
- Form analytics veteran Zuko has launched Agent Score, an audit tool that tests whether AI agents can complete a site's online forms and checkouts
- It scores agent-readiness out of 100, pinpoints the fields and steps where automated traffic drops out, and returns a video plus a list of technical fixes
- As agent-driven purchasing surges, forms built for humans risk quietly losing a new category of "machine customers"
What Zuko Announced
Manchester, United Kingdom (Newsfile Corp. - June 8, 2026) - Leading form analytics and session replay platform Zuko has launched a tool to test whether AI agents can complete forms and checkouts.
www.newsfilecorp.comOn June 8, 2026, Manchester-based form analytics and session replay platform Zuko launched Zuko Agent Score, a tool that tests whether AI agents can successfully complete a business's online forms and checkouts. It identifies the points where automated traffic drops out before converting.
Zuko, originally known as Formisimo, has spent more than a decade helping businesses understand why users abandon online forms and checkouts. The new tool applies that same lens to AI agents, a category of traffic that existing analytics tools were never built to handle.
What Agent Score Measures and Returns
At the heart of Agent Score is a simple idea: run an actual AI agent through your existing forms and checkouts, then observe where it fails. Rather than relying on a static checklist, the tool exercises the flow live and watches what happens.
The output starts with a score out of 100 showing how ready the form is to accept business from AI agents. Alongside it comes a detailed analysis covering accessibility, navigation, error handling, and data input. The tool also returns a video recording of the agent working its way through the form, plus a list of technical recommendations and fixes to improve the experience.
Alun Lucas, Managing Director at Zuko, framed the gap clearly: "Most online forms and checkout flows were designed for humans, they don't consider how AI agents interact with websites." He added that "as AI agents begin handling more tasks on behalf of customers, enterprises need more visibility into where those interactions break down, which is exactly why we've launched this new tool."
Why Agent-Readiness Is Suddenly a Question
The backdrop is an explosion in agent-driven traffic. Citing industry data, Zuko notes that AI agent traffic is growing at more than 7,000% per year.
That figure aligns with independent research. HUMAN Security's 2026 State of AI Traffic & Cyberthreat Benchmark Report found that agentic AI traffic grew 7,851% year over year. The multiplier reflects rapid early adoption from a very small 2024 base, but it signals how fast the shift is moving. The same report notes that 2.3% of agentic activity is already happening on checkout pages.
What matters is that AI agents perceive the web in a fundamentally different way than people do. They don't parse CSS layouts or appreciate animations; they depend on the underlying structure to operate. Missing labels, empty buttons, broken heading hierarchies, and absent alt text are structural problems that break agent navigation even when humans never notice them. Accessibility has quietly become a business-critical development priority.
Many companies are racing to show up in AI-powered discovery. Far fewer are asking whether the destination can actually transact once an agent arrives. The distance between digital presence and the ability to respond when an agent tries to buy is exactly the gap Agent Score targets.
A New Kind of Drop-Off
What is happening here is a different species of abandonment than the classic cart drop-off. Human abandonment is tangled up with psychology, price, and friction. Agent abandonment is more mechanical and rooted in structure.
JavaScript-dependent validation, CAPTCHAs, unexpected error messages, or buttons that can't be reached by keyboard alone may be trivial obstacles for a human but hard stops for an agent. The trickier issue is that these failures often happen silently. An agent may loop on the same action and get stuck, or push through with the wrong data, leaving nothing behind but a server log that says traffic arrived and never converted.
Including a video recording in the output reads as a remedy for that invisibility. The most dangerous agent failures are the ones that look exactly like success: a wrong address entered and submitted, or the wrong item selected, can still register as a clean completion. Pairing a quantitative score with a watchable replay lets a human verify what actually happened.
What This Means for Merchants
For e-commerce operators, the message is clear. Conversion is no longer driven only by human customers; AI agents acting on a customer's behalf are becoming buyers in their own right, and forms need to assume that.
On the payments side, standards like OpenAI's Agentic Commerce Protocol and Google's Universal Commerce Protocol are already building the rails for agents to interact with merchant catalogs and checkouts. But whether a site's front-end forms can actually accept that agent once it tries to buy is a separate question. A real gap remains between protocol support and the UI implementation customers and agents touch.
A practical first step is to audit whether your key forms and checkouts can withstand agent interaction: do fields carry proper labels and ARIA attributes, is error handling legible to an agent, can the flow be completed with keyboard navigation alone? Because these overlap heavily with human accessibility improvements, the investment pays off twice. Lead-generation forms and subscription sign-ups, where agents are most likely to act on a user's behalf, are sensible places to start.
The piece teams tend to overlook is making the fix durable. Sites change constantly with campaigns, inventory, and payment updates, and an agent flow that passed in one release can break silently in the next. Whether a diagnostic like Agent Score can be folded into a regression suite that runs on every deploy will largely determine long-term readiness.
Conclusion
Zuko's launch marks the moment the agentic commerce conversation comes down from payment protocols to front-end implementation quality. Standards can usher agents to your door, but if your forms drop them on the threshold, that traffic is lost revenue.
It is no small thing that a company which spent a decade making human cart abandonment visible is now turning the same lens on machine customers. Merchants are entering a phase where agent-readiness is best treated not as an eventual nice-to-have but as a measurable metric tied directly to conversion. The next question is one of speed: how quickly teams can turn these diagnostic findings into concrete fixes.
