The Mirror Website Playbook: Turn AI Visibility Into Inbound

Most websites are designed for the wrong audience. They're built to impress humans on the first scroll, while the actual buyer journey now starts inside an AI assistant that wants a clean, parseable answer — not a hero video.

Rebuilding your live site for AI is slow, expensive and risky. A mirror site sidesteps all of that. This playbook walks through the full operating model, from week one to compounding inbound, distilled from twelve months of running mirrors in production for B2B service firms and SaaS companies.

What a mirror site is

A mirror site is a parallel, AI-optimised version of your content. Same business, same offer, same positioning — but stripped down to the form that AI assistants and time-starved buyers actually want.

• Lives on a subdomain or sub-path you control.
• Reuses your brand, but with cleaner typography and tighter copy.
• Ships with llms.txt, AGENTS.md, structured FAQs and a smart intake flow.
• Iterates weekly — content changes don't need a design review or a deploy.

Why mirror instead of rebuild

The case for a mirror, in one table:

The shape of a mirror site

A mirror is small on purpose. Most live in 8–15 pages. The discipline of "what would we cut?" is most of the value.

Notice what's missing: long About pages, multi-step onboarding flows, blog archives. Those can live on your live site. The mirror's job is to convert the buyer who arrived already informed.

Week-by-week build plan

Week 1 — Audit & inventory

• Crawl your existing site and list every public page.
• Pull the top 50 brand-relevant prompts from AI assistant logs.
• Map each prompt to the page that should answer it (or note "no page exists yet").

Week 2 — Draft the mirror

• Stand up the subdomain (we use mirror. as a convention).
• Write tight versions of your 8–15 canonical pages. Plain language, one purpose per page.
• Publish llms.txt and AGENTS.md.

Week 3 — Wire up intake

• One smart form. Budget, timeline, fit — three questions, not twelve.
• Route enquiries straight into your existing CRM. No new system to learn.
• Set up assistant-source tracking so you can see what's working.

Week 4 — Ship and observe

• Soft-launch the mirror. Don't redirect anything from your live site yet.
• Let assistants discover it organically; most index within 7–14 days.
• Watch the questions visitors and assistants ask. The questions are your roadmap.

The improvement loop

A mirror site that doesn't iterate stops working. The discipline is a weekly loop:

1. Listen — which questions came up, which pages got cited, what assistants got wrong.
2. Draft — write the missing answer or sharpen the existing page.
3. Ship — push the change. No design review, no PM rituals.
4. Measure — did mentions go up? did enquiry quality improve?

This is unsexy work. It also compounds. In our cohort, mirrors running a disciplined weekly loop deliver ~3× the inbound of mirrors that ship once and forget:

Metrics that matter

Most marketing dashboards track the wrong things for AI inbound. Sessions and bounce rate are misleading when most visitors arrive from a one-shot answer. Track these instead:

Pitfalls to avoid

• Cloning the live site verbatim. The mirror should be tighter, not identical. If it's a copy, it adds no value.
• Treating it as a side project. Mirrors work when one person owns them. Without ownership, the loop dies in week three.
• Skipping the standards files. Without llms.txt and AGENTS.md, you're invisible to half the channel.
• Optimising for the wrong metric. Sessions don't matter. Booked calls do.
• Never asking visitors what they wanted. The single best source of new pages is the questions your intake form keeps surfacing.

FAQ

Will a mirror site hurt my SEO?

No, if you set canonicals correctly. The mirror complements the live site rather than competing with it. We handle this for clients automatically.

Do I need new branding?

No. Mirrors use your existing logo, colours and tone. The difference is in copy density and page purpose, not visual identity.

Can my agency run this?

They can — but most agencies aren't set up for weekly iteration. The operating model is closer to a product team than a marketing campaign. If your agency is comfortable with weekly ships, great. If not, run it in-house or with a partner who is.

How long until it pays back?

Most cohort companies see the mirror cover its cost within 60–90 days from booked calls alone. The compounding effects on assistant share-of-voice show up over 6–12 months.