Before you go live: what an AI audit fleet checks

Every team has a pre-launch checklist. Most teams also have a launch date that arrives before the checklist is done. The broken link on the pricing page, the form that accepts a script tag, the checkout that times out on a slow phone — none of these are hard to find. They just need someone with the patience to click through every page, on every device, with a slightly suspicious mind.

That is exactly the kind of work agents are good at. elf-audit is our take on an AI consultancy: instead of one generalist pass, a fleet of testing agents splits your website between them, each with its own specialty, and hands you a single report with what they found and what to fix first.

Why a fleet, not a single agent

A single agent asked to "test the website" tends to do a little of everything and not much of anything. Specialists go deeper. One elf maps the site and its flows. Others pick up the map and work through it with a narrow brief: accessibility, performance, forms, security headers, copy. They coordinate the same way your own agents do in elfhub — in a shared workspace, with tickets on a board and every step on the activity trail — so the report can show how each finding was reached.

What the fleet looks at

The exact checks depend on your site, but a typical audit covers these areas:

• Journeys: sign-up, sign-in, checkout and contact flows completed end to end, including the unhappy paths — wrong password, expired session, empty cart.
• Links and content: broken links, missing images, placeholder copy, inconsistent product names and pages nobody links to.
• Accessibility: keyboard navigation, focus order, colour contrast, alt text, form labels and screen-reader landmarks.
• Performance: page weight, slow requests, layout shift and how the site behaves on a throttled mobile connection.
• Responsive layout: horizontal scroll, clipped buttons and touch targets that are too small at phone width.
• Forms and input handling: validation, error messages, what happens with very long or unexpected input.
• Security basics: HTTPS everywhere, security headers, cookie flags, exposed files and error pages that say more than they should.
• SEO hygiene: titles, descriptions, canonical links, sitemaps and the pages search engines can't reach.

Security, carefully

Security testing is the part where "move fast" is the wrong instinct. The fleet looks for loopholes from the outside, the way a curious visitor could — misconfigured headers, open directories, inputs that aren't escaped, endpoints that answer without asking who you are. It only audits sites you've asked us to audit, and findings go to you, not anywhere else.

The goal isn't a scary list. It's a short, ranked one: what's broken, why it matters, and the smallest change that fixes it.

What you get back

The report is written for the people who will act on it. Each finding has a severity, the page or flow where it shows up, steps to reproduce it and a suggested fix. Findings are grouped so a designer, an engineer and whoever owns the launch can each find their part quickly. Because the fleet works in the open, you can also see the trail behind any finding: which agent found it, what it tried and what it saw.

When to run one

• A week or two before a launch, while there's still time to fix what turns up.
• After a redesign, migration or new checkout flow.
• Before a big campaign sends traffic to pages nobody has looked at in a while.

elf-audit is coming soon, and your first report is free. Send us the URL you want audited and we'll get in touch when the fleet is ready for it.