Agent-readiness · the next mobile-readiness
Claude in Chrome, ChatGPT's agent mode and Gemini already browse ordinary websites for their users — reading, comparing, filling forms, completing purchases. Research finds agents complete tasks on roughly 78% of accessible, well-structured sites versus 42% of inaccessible ones. Agent-readiness is 2010's mobile-readiness: early, measurable, and cheapest to fix before the traffic arrives.
Agents that already browse and buy for their users
What we measure
Every check says plainly how strong its evidence is — including the fashionable ones whose measured effect today is weak.
Semantic structure and labelled forms carry the most weight in your score — because the 78%-vs-42% completion gap is measured on exactly these.
Structured data (JSON-LD) is the difference between an agent knowing your price and guessing it.
Most robots.txt files take no position on AI crawlers. Silence is a policy by accident — we make it deliberate.
We include llms.txt as the cheap hedge it is — and tell you plainly that its measured effect today is weak. Evidence-weighted means exactly that.
What arrives
The fix pack is written for your site, in the exact formats agents and crawlers read. Your web person applies it in an afternoon; we retest and issue the updated score.
# yourshop.ie > Independent hardware shop, Dublin. Same-day collection, EU delivery. ## What we sell - Hand tools, power tools, fixings — [catalogue](/products) - Trade accounts — [terms](/trade) ## For agents - Prices are in EUR and include VAT - Stock status is live on each product page - Orders: [checkout](/checkout) · Contact: [hello@yourshop.ie](mailto:hello@yourshop.ie)
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Product",
"name": "18V cordless drill driver",
"sku": "DRL-18V-01",
"offers": {
"@type": "Offer",
"price": "129.00",
"priceCurrency": "EUR",
"availability": "https://schema.org/InStock"
}
}
</script>
# A deliberate stance, per crawler — not silence. User-agent: GPTBot Allow: /products Disallow: /account User-agent: ClaudeBot Allow: / User-agent: Google-Extended Allow: / Sitemap: https://yourshop.ie/sitemap.xml
The path
Free checkers hand you a score; the pack hands you the finished artefacts and the retest that proves the work happened.
Your site scored 0–100 across eleven checks — structure, interaction, access policy, trust signals — worst problems first, each with what an agent experiences and why it matters.
Not advice — artefacts: your llms.txt written, your JSON-LD entity drafted, your robots policy for GPTBot/ClaudeBot/Google-Extended, and a fix list your web person can apply in an afternoon.
Apply the fixes, we re-run the audit and issue the updated score. Maintained customers get re-audited as the agent landscape moves — a copied report goes stale; a subscription to currency cannot.
Founding pilot
Onboarding a small founding cohort by email, one site at a time.
Do it yourself
One-off
Kept current
Pilot guarantee: if the audit doesn't surface at least three concrete, fixable issues specific to your site, the fix pack costs nothing. Free checkers exist and we'll happily point you at them. The alternative at the other end: "AEO" agency engagements at $5,000–$25,000 plus monthly retainers — mostly SEO firms with new wording. Figures from published agency pricing; not our quotes.
Card payments are taken by Stripe on its secure checkout page — this site stores no payment details. Pilot invoicing by bank transfer is also available after scoping, with nothing owed until your order is confirmed in writing. Our solicitor's pre-launch review is ongoing; its wording lands on these pages the day it arrives.
Honest limits
The field is young; a vendor who won't say so is selling you the hype, not the readiness.
AgentProof measures observable agent-readiness signals and hands you implementation artefacts. It is not a guarantee of agent traffic, rankings or revenue.
No law requires any of these changes — this is opportunity, not compliance, and we say so on every report.
Every check is weighted by measured evidence, and we say plainly when a fashionable artefact's effect is weak today.