PlaybooksUpdated November 4, 20268 min read

Agent-Ready Website Checklist: 27 Checks Before You Publish

Before you publish, run 27 checks across three groups: Discovery (can crawlers like GPTBot, ClaudeBot, and Google-Extended fetch your pages), Data Quality (are titles, schema, pricing, and links clear), and Actionability (can an assistant answer, quote, and act). These checks make your site easier for AI systems to read; they do not guarantee citations, rankings, or traffic.

Key takeaways

27 checks in three groups: Discovery (9), Data Quality (10), and Actionability (8), the same categories the audit scores.
Start with robots.txt: allow the major AI user-agents, and publish llms.txt at your domain root.
Publish pricing, use one clear H1 per page, and add JSON-LD for Organization, Service, and FAQPage.
Sensitive agent actions, payments, emails, record changes, require explicit human approval, authentication, and scoped permissions.
The free audit runs the whole list in about 30 seconds and hands back a prioritized fix list with copy-ready code.

How to use this checklist

This is the pre-publish pass we run on every site: 27 checks grouped into the same three categories the free audit scores, Discovery (can machines find and fetch you), Data Quality (is what they fetch accurate and readable), and Actionability (can an assistant answer a buyer, and can a compatible agent act safely). Work top to bottom. A failed check is not an emergency; it is a task with a known fix.

Everything here is testable in a browser, in view-source, or with a single curl command, no search console access, no paid tooling. Where a check deserves a full walkthrough, we have written one: the llms.txt guide and the schema guide cover those steps in depth.

Time: 45-90 minutes by hand for a five-to-ten-page site, or about 30 seconds in the free audit.
Order matters: Discovery failures hide Data Quality work, if a crawler cannot fetch the text, the rest of the list never gets tested.
Score: Discovery is worth 40% of the audit, Data Quality 40%, Actionability 20%. Fix in that order.

Discovery: can machines find and fetch you (9 checks)

Discovery covers access. If these nine fail, nothing downstream matters, an assistant that cannot retrieve your page cannot quote it. Check robots.txt first; it is the single most common place a site silently blocks the readers it wants to reach.

1. robots.txt allows the major AI user-agents: GPTBot, OAI-SearchBot, ClaudeBot, Google-Extended, and PerplexityBot.
2. robots.txt does not block the CSS and JavaScript your pages need to render.
3. sitemap.xml exists, is referenced from robots.txt, and lists only canonical URLs that return 200.
4. llms.txt is published at your domain root and links to your key pages and any markdown mirrors.
5. Core pages, home, services, pricing, FAQ, render their main content without JavaScript (test with view-source or curl).
6. Every page sets a canonical URL, and your key pages self-reference it.
7. No important page carries noindex or a Disallow rule left over from staging.
8. Every important URL returns 200 directly, no redirect chains, no soft-404s.
9. Every important page is reachable within a few clicks of the homepage, with descriptive anchor text.

Data Quality: is what they fetch clear and correct (10 checks)

Discovery gets a crawler to your page; Data Quality decides whether what it reads is usable. This is where most sites lose points, not because anything is broken, but because the facts are ambiguous, duplicated, or buried behind a form. Consistency beats volume here: one company name, one set of service names, one price.

1. Each page has a unique title tag of roughly 50-60 characters that leads with what the page actually is.
2. Each page has a meta description of roughly 150-160 characters that states the answer, not a slogan.
3. Exactly one h1 per page, phrased the way a buyer would ask the question.
4. JSON-LD Organization markup on the homepage with consistent name, logo, and sameAs links.
5. JSON-LD Service (or Product) markup on every offer page, with a price or price range where you publish one.
6. JSON-LD FAQPage markup on pages that carry a visible FAQ, matching the visible text exactly.
7. Every meaningful image has descriptive alt text; decorative images use empty alt.
8. Internal links use descriptive anchors ("pricing for team audits"), not "click here".
9. Pricing is published, a number, a range, or a clear "from $X", on a crawlable page.
10. Content has depth: each key page answers the full question in plain language instead of three lines and a contact form.

Actionability: can an assistant or agent act on it (8 checks)

The last group asks a different question: once an assistant has read your site, can it do something useful with it, answer a buyer, quote your terms, hand off to a human, or call a scoped tool? Actionability is weighted at 20% of the audit because it is the layer you add last, not the one to start with. Eight checks:

1. Documentation, onboarding, and help content are discoverable from the navigation or footer, not hidden behind login or a JavaScript-only search box.
2. FAQs cover the buying questions: scope, timeline, pricing logic, security, and what happens after a form is submitted.
3. Agent signals are in place: llms.txt points to docs and machine-readable resources, and any MCP endpoint is declared where agents can find it.
4. Each page has one primary conversion path, audit, booking, or contact, visible without scrolling twice.
5. Contact and booking details exist as plain text and HTML, not only inside a JavaScript widget.
6. Any pre-authorized agent action is written down: what it may do, what it may not, and which steps require human approval.
7. Security and policy material, privacy, terms, data handling, is crawlable and linked from the footer.
8. Response expectations are stated: what happens next, and how fast, after a human or an agent initiates contact.

What this checklist does not promise

A clean checklist means your site is legible to AI systems. It does not mean those systems will cite you, rank you, send traffic, or generate leads, no one can promise that, and we do not. Preparedness improves the odds of being understood; it does not determine what an assistant chooses to say about you.

Nor is the checklist a substitute for having something worth quoting. Clear structure carries your expertise to the model; the expertise still has to be there. If your pages say the same generic things as every competitor, being machine-readable only makes the sameness easier to detect. The checklist clears the mechanical barriers, the substance is still your job.

Or run all 27 checks in 30 seconds

You can walk this list by hand, or run the free Agent-Ready Website Audit: about 40 checks in roughly 30 seconds, a score across Discovery, Data Quality, and Actionability, and a prioritized fix list with copy-ready code you can paste straight in. Every run produces a shareable report URL, and the full report unlocks with a name and email.

Fix the priority-1 items yourself, most are an afternoon: robots.txt rules, missing schema, buried pricing. If you would rather hand it over, the Agent-Ready Website Build starts at $2,500 and ships design, copy, schema, and llms.txt in one pass, with the full build running $15,000-$25,000 for larger sites. Either way, re-run the checklist after you publish, these checks drift quietly.

Frequently asked questions

How long does the agent-ready checklist take to run manually?

A first pass takes 45-90 minutes for a typical five-to-ten-page site: `curl` for `robots.txt` and `llms.txt`, view-source for schema and titles, and a browser with JavaScript disabled for the rendering check. The free audit runs the same checks, plus about a dozen more, in roughly 30 seconds.

Is every check a Google ranking factor?

No. Some checks overlap with standard SEO hygiene, titles, canonicals, internal links, but others, such as `llms.txt` and MCP signals, are about AI-system preparedness rather than search rankings. Passing the checklist means AI systems can read you correctly; it does not place you in any result.

What is the single most common failure?

Fetching: robots.txt rules left over from staging, or core content that only appears after JavaScript runs. If a crawler cannot get the text, the rest of the checklist never gets a chance to matter, which is why Discovery sits at the top of the list and carries 40% of the audit score.

Do I need an MCP server to pass?

No. MCP signals are one of the eight Actionability checks and matter only when you want compatible agents to call scoped tools on your behalf. A site that passes the other 26 checks, clear pricing, real docs, correct schema, is already ahead of most.

Put this guide to work on your site.

Run the free audit for an instant score and prioritized fix list, or book a 30-minute call and we scope the work for you.

Run free audit Book a design tour session

We do not guarantee AI citations, rankings, traffic, leads, revenue, or purchases.

Playbooks · 8 min read

How to Audit Your Website for AI Readiness (Free Method)

Measure AI readiness with a weighted audit: Discovery 40%, Data Quality 40%, Actionability 20%. Manual steps, plus the free 30-second audit with a fix list.

Foundations · 6 min read

llms.txt: How to Create One (With a Template)

llms.txt is a curated markdown file at your site root for AI clients. What it is, when it helps, a copy-ready template, and honest limits on its impact.

Foundations · 7 min read

AEO vs SEO: What Changes in the AI Search Era

SEO earns rankings in search results; AEO makes your content understandable and usable by AI assistants and agents. Where they overlap, where they diverge.