Can AI Agents Use Your Website? Run the Agent-Readiness Audit

Why Does Agent-Readiness Suddenly Matter?

Because the traffic mix has already tipped. According to WorkOS's analysis of web traffic trends, automated requests — crawlers, indexers, and a fast-growing new category of agents that browse on behalf of a human — now make up the majority of activity on large parts of the web, with automated requests reported above half of HTML traffic in some measurement windows. The exact percentage moves depending on how you count, and honest reporting should say so, but the direction is not in dispute: the share of your visitors that are software, not people, is climbing.

The important distinction is what kind of software. Old-school crawlers scrape content to train models or build a search index. The new category is different: an agent visiting a page, reading it, filling a form, comparing options, and completing a transaction — standing in for a customer who delegated the errand. We wrote recently about how AI agents are becoming your customers rather than just researching on their behalf, and that shift is exactly what Is Agentic is built to measure.

Here is the uncomfortable part. Most of the design and SEO effort of the last decade optimized for a human clicking through a rendered page and for Googlebot, which now executes JavaScript. An autonomous agent often does neither well. Many agents fetch raw HTML and do not run a heavy client-side render. They can't “see” content that only appears after a framework hydrates. They give up on a soft-404 that returns a 200 status with an error page. They can't find your hours if they live inside a chat widget. This is the same structural problem we described when we argued that your website is built backward for AI search — except now the stakes are a lost transaction, not just a lost citation.

What Is the Is Agentic Tool, Exactly?

Is Agentic is a free scanner from Vercel. The scoring methodology comes from Ora, an agent-experience research company spun out of era labs. Vercel ships the tool and the interface; Ora defines and runs the checks. That division matters when you cite it — the number is Ora's methodology, the product is Vercel's.

The tool fetches your public site the way an agent would and runs 118 automated checks. It then rolls those checks into a 0–100 score and a letter grade. Ora is explicit that a score is only trustworthy “when the methodology is public, the checks run live, and the number is ranked against everyone else” — and all three are true here, which is what separates this from a vanity badge.

Under the hood, Ora describes a weighted model: essential checks draw from an 80-point pool, recommended capabilities from a 20-point pool, and emerging signals can add up to a 5-point bonus. On top of the web scan, the tool ships a read-only API, a command-line interface, and an MCP server so that AI coding agents can pull your score and even suggest fixes — none of which require an API key or billing account. As of launch, Ora confirms there are no paid plans, subscriptions, or per-report fees.

For a sense of how demanding the grading is: a public scan of aws.amazon.com published on the tool scored 57 out of 100 — a C on Ora's scale. If Amazon's cloud homepage lands in the middle of the band, the typical Northeast Indiana small-business site built on a page-builder theme is not starting from a comfortable place.

What Do the Four Dimensions Actually Measure?

The 118 checks are grouped into four dimensions, each weighted differently. The weighting tells you where Vercel and Ora think the real friction is — and it is not where most site owners spend their time.

Note where the weight sits. Usability at 40 points and Access at 30 points together carry 70 of the 100 available points. Discovery — the dimension closest to classic SEO and structured data — is only 20 points, and Payments is a slim 10. The message is blunt: being findable is table stakes; being usable and reachable is where sites win or lose. A site can have perfect schema markup and still fail because its core content only appears after a JavaScript render, or because its “book now” button opens a modal an agent can't trigger.

The letter bands make the scale concrete: A+ is 95–100, A is 86–94, B is 70–85, C is 48–69, D is 28–47, and F is 0–27. A “C” is not a passing grade in a world where an agent will simply pick the site it can use.

How Do I Run the Audit on My Own Site?

This is the part worth doing before you finish reading. The tool is genuinely self-serve.

The fastest path is the web interface: go to is-agentic.com, enter your domain, and read the graded report. If you or your developer prefer the terminal, the command-line version needs no install:

npx is-agentic yourbusiness.com

Add --json if you want machine-readable output to hand to a developer or feed into your own tooling:

npx is-agentic yourbusiness.com --json

A few honest caveats. The public API is rate-limited (Vercel documents roughly 120 requests per client IP per 60-second window), so this is a periodic audit, not something to hammer. And a letter grade is a starting diagnostic, not a full architecture review — a passing score means an agent can traverse your site, not that every workflow is optimized. Read the individual failed checks, not just the headline number. The check IDs — content-no-js, agent-friendly-404, json-ld, sitemap, metadata-completeness — each map to a specific, fixable defect.

When you get your report, group the failures into three buckets: content that doesn't render without JavaScript (an Access problem), missing structured data and sitemaps (a Discovery problem), and buried or agent-hostile conversion paths (a Usability problem). That grouping is your remediation roadmap.

What Does Remediation Actually Involve?

The good news for business owners is that almost nothing on this list requires reinventing your site. It requires fixing fundamentals that also happen to help human users and traditional search. In our experience, the highest-leverage fixes are these:

Server-render the content that matters. If your hours, services, pricing, and contact details only exist after a client-side framework hydrates, an agent may never see them. Server-side rendering or static generation of core content is the single biggest Access win. This is standard web development work, not a rebuild.

Return honest status codes. A missing page should return a real 404, not a 200 with an error message painted on it. agent-friendly-404 exists because soft-404s quietly poison an agent's navigation. Clean status codes are a one-time infrastructure fix.

Add and validate structured data. JSON-LD schema.org markup for your organization, local business, hours, services, and FAQs gives an agent an unambiguous machine-readable summary of who you are and what you offer. This is the same markup that strengthens AI-search visibility, which is why we fold it into our answer-engine optimization work.

Publish a real sitemap and complete metadata. These are cheap, and Discovery points are the easiest 20 on the board.

Make the conversion path reachable without a widget. If the only way to book, quote, or contact you is a third-party chat bubble, an agent likely can't use it. Provide a plain, linkable path — a real form, a real phone number in the markup, a real booking URL. This is the Usability bucket, and it is where the points and the revenue both live.

The pattern here is the one we keep returning to: the work that makes you legible to AI agents is largely the same work that makes you legible to AI search. That overlap is why the Google search-box redesign is resetting web strategy for mid-market sites, and why treating agent-readiness and AEO as one engagement, rather than two, is the efficient move.

What a Failing Score Costs a Fort Wayne Business

Picture a well-run Fort Wayne home-services company — HVAC, plumbing, electrical — with a clean, modern-looking site built on a popular page-builder theme. It looks great to a human. But its service area, its emergency hours, and its “request a quote” flow are all rendered client-side, and the only contact shortcut is a chat widget that loads after everything else. Run npx is-agentic against it and you would likely see a low Access score and a Usability score gutted by the invisible quote flow.

Now put it in context. A homeowner in Auburn or across Allen County, furnace out on the first cold night, tells an assistant to “find a company that does emergency HVAC near me tonight and book it.” The agent visits three local sites. On two, it reads server-rendered service pages, finds a linkable booking path, and completes the request. On the page-builder site, it sees a mostly blank render and a chat bubble it can't operate — so it never surfaces that business to the homeowner at all. The company didn't lose a ranking; it lost the job, and it will never see the lost lead in its analytics because the agent bounced before any human ever arrived.

This is not hypothetical friction for a rural market — it is arriving fastest in exactly the professional-services, home-services, healthcare, and financial-services categories that make up Northeast Indiana's business base. The upside is symmetrical: the same remediation that recovers the score recovers the booking, and it compounds with search visibility. We have already seen that Fort Wayne AI-search traffic converts when the fundamentals are right, and agent-readiness is the same fundamentals aimed at a new kind of visitor. Locally, being early here is a real edge — most competitors haven't run the scan yet.

Run Your Scan, Then Fix What It Finds

You can get a letter grade for your own site in the next five minutes at is-agentic.com, and you should — it costs nothing and it tells you exactly where the agentic web is bouncing off your business. What comes after the diagnosis is where we help.

Cloud Radix pairs answer-engine optimization with web development to close the gap: server-rendering the content agents can't see today, cleaning up status codes and sitemaps, adding the structured data that makes your business machine-legible, and exposing booking and contact paths an agent can actually complete. It is the same infrastructure story we tell about why AI is replacing the interface — the front door to your business is increasingly an agent, and the door needs to open. If your scan came back a C or worse, that is a fixable engineering problem — tell us your domain and grade and we'll map the remediation.

Frequently Asked Questions

Sources & Further Reading

• MarkTechPost: marktechpost.com/2026/08/23/vercel-introduces-is-agentic — Vercel introduces ‘Is Agentic’, a free agent-readiness scoring tool that audits public websites using Ora's 100+ checks.
• Vercel: is-agentic.com — Is Agentic by Vercel: the AI agent readiness score tool and web scanner.
• Ora: ora.ai/blog/is-agentic-with-vercel — Ora on partnering with Vercel to launch is-agentic.com, including the scoring methodology.
• Vercel: is-agentic.com/scan/aws.amazon.com — Public scan showing aws.amazon.com scores 57/100 for agent readiness.
• WorkOS: workos.com/blog/ai-agent-web-traffic-what-developers-need-to-change — AI agents now make up the majority of web traffic: what developers need to change.
• Schema.org: schema.org/docs/gs.html — Getting started with JSON-LD structured data vocabulary.