Agent readiness, one check at a time

Agent readiness on Webflow

What Webflow gives you by default, where structured data and root text files need custom code or a redirect, and the fixes in order of effort.

Checks B1, B2, A4

Agent readiness on WordPress

Server rendering is the hard part done. What the plugin layer adds and breaks on A5, B1 and A6, and why a CDN in front is the usual real cause of a block.

Checks A5, B1, A6

Agent readiness on Shopify

What a Shopify theme gives you by default on B2, C5 and A4, where the platform makes root files and headers awkward, and the fixes in order of effort.

Checks B2, C5, A4

Agent readiness on Next.js App Router

What the App Router gives you by default, where a Next site loses points on A7, B5 and B4, and the fixes in effort order, grounded in the stack we run.

Checks A7, B5, B4

Certificate renewals and the HSTS header that quietly vanished

An HSTS header lives wherever TLS terminates, and moves with it. How a routine ops change removes it without an error, and how a re-scan with a diff catches it.

Checks D1

Why "should an agent cite this site" is a trust question

The trust pillar asks whether an agent should cite a site. That is a question about provenance, and provenance is files, headers and fields that exist or not.

Multi-page stability, the template drift that breaks trust

Check D7 wants every crawled page to return 200, none over four seconds, and the same navigation on all of them. How it measures each and what drift looks like.

Checks D7

security.txt in ten minutes

What RFC 9116 requires, the Expires field that quietly invalidates the file a year later, how check D6 reads it, and how we serve ours from a rolling window.

Checks D6

Identity consistency, one name across title, Open Graph and schema

How check D5 compares the title element, og site name and schema name, why they drift apart on ordinary sites, and how defining the name once stops it.

Checks D5

A changelog is a freshness signal

For a product site whose pages rarely change, a dated changelog is the cheapest way to score on D4, if each entry carries a date a crawler can read.

Checks D4

Freshness signals, lastmod, the 90-day window and visible dates

How check D4 scores freshness, any lastmod, a fifth of sitemap URLs modified in 90 days, a machine-readable date on a crawled page, and how ours stays true.

Checks D4

Privacy, terms, shipping and returns. The policies an agent looks for, by site type

Check D3 wants a linked privacy policy, plus shipping and returns for shops or a terms page for everyone else. Why linked matters more than existing.

Checks D3

Verifiable identity, a contact page and a sameAs that resolves

Check D2 wants a discoverable contact route and Organization schema with a sameAs link. How the scanner looks, and why a dead sameAs is worse than none.

Checks D2

HSTS in one header, and why mixed content still fails you under render

Check D1 awards four points for valid TLS, a Strict-Transport-Security header and no http references in the page. What each part measures, and how we set ours.

Checks D1

How C5 redistributes for non-commerce sites, and why

The arithmetic that lets a SaaS or content site score out of the same 100 as a shop, exactly how C5's four points move to C1 to C4, and how the report shows it.

Checks C5

On-site search that an agent can drive by URL

How check C6 finds your search form, why a GET form with a query parameter scores where a POST form or a JS combobox does not, with our header as the example.

Checks C6

Guest checkout is an agent feature

The four signals check C5 looks for on a shop, what each means for an agent completing a purchase, and how a login-gated cart shows up in the evidence.

Checks C5

WebMCP, what is real, what is proposed, and what to build now

WebMCP is a proposal, MCP has clients, agents.json has no governing spec and OpenAPI is settled. Which parts of C4 are a bet, and what to build first.

Checks C4

Building an MCP server for a marketing site in an afternoon

What our hosted MCP server exposes, why it is stateless and unauthenticated, and which read-only tools a marketing site would offer a model.

Checks C4

Publishing an OpenAPI spec for a site that has no API

A read-only OpenAPI 3.1 document over data you already render earns C4 credit and is genuinely useful to an agent. What to put in it, and where to link it.

Checks C4

agents.json, what it is, what to put in it, and what is still unsettled

Our /agents.json annotated field by field, why its endpoints list stayed empty until they existed, how check C4 tests for it, and which parts are a bet.

Checks C4

Buttons that are buttons, tab order, focus and keyboard traps

How check C3 counts clickable divs and positive tabindex on the homepage, why the keyboard-trap point is awarded untested, and how to test tab order by hand.

Checks C3

Breadcrumbs and URL hierarchy as a map for agents

Check C2's fourth point wants a breadcrumb trail or a URL hierarchy on an interior page. What the scanner looks for, and how our page header supplies both.

Checks C2

Click here, read more, and the contextless link problem

Check C2 fails a homepage where 5% or more of links say click here, read more or nothing. How the scanner counts, and how to write link text a machine can use.

Checks C2

Links that are links, click-handled divs and the missing href

Why navigation built on onClick is invisible to a plain fetch, how C2 counts anchors and clickable divs, and how our header keeps every link in the HTML.

Checks C2

The type=url trap, and why our scan form does not use it

An input with type=url rejects a bare domain like yourcompany.com, which is what the placeholder invites, so the form never submits. What we do instead.

Checks C1

Forms an agent can fill, labels, input types and a real submit button

What C1 measures on a form, why a placeholder is not a label, which input types the scanner expects, and why Enter-to-submit leaves an agent nothing to press.

Checks C1

Structured data across templates, not page by page

Schema that survives a redesign is built once, in typed builders, and emitted by the layout and each template rather than each page. How B1 and B2 read it.

Checks B1, B2

Duplicate content through an agent's eyes

Tracking parameters, trailing slashes, case, pagination and print views turn one page into several. What rel=canonical does for a machine that fetches once.

Checks B7

Canonical hygiene, one host for www, non-www, http and https

Every site has four URLs for one homepage. How check B7 probes the www and http variants, which statuses pass, and how rel=canonical is scored across pages.

Checks B7

Alt text as a data problem, not a compliance problem

How check B6 counts content images, why icons under 50px are excluded, what an empty alt costs, and how to audit coverage from templates, not page by page.

Checks B6

Clean text ratio, or how much of your page is actually page

What the 0.10 and 0.05 thresholds in check B5 mean, why script and style contents are excluded since rubric v1.0.1, what inflates markup, and how to measure it.

Checks B5

Open Graph for agents, not just for Slack previews

The three Open Graph tags are the cheapest summary a machine reads. How checks B4 and D5 score them, and how we generate a per-page image on the server.

Checks B4, D5

Title and description lengths, and why the ranges exist

Why check B4 wants a unique title of 15 to 70 characters and a description of 50 to 160 on every page, and how a site name template eats a third of the budget.

Checks B4

Landmarks, main, nav and footer, and why div soup is an agent bug too

How check B3 looks for main, nav and footer on every crawled page, how C3 treats clickable divs and positive tabindex, and why the WCAG fix is the agent fix.

Checks B3, C3

One h1 per page, and the other heading rules agents care about

Why check B3 wants exactly one h1 and no skipped heading level on every crawled page, what a skip does to extraction, and the component patterns that cause it.

Checks B3

Product and Offer schema that matches the page

What a correct Product and Offer block looks like, how checks B2 and C5 test for it, and why a price that disagrees with the page is worse than no schema.

Checks B2, C5

Which schema types your site type actually needs

Organization or WebSite first, then Product and Offer, Article, FAQPage or SoftwareApplication by site type. How check B2 decides what kind of site you are.

Checks B2

JSON-LD that parses, the five mistakes that score zero

Trailing commas, escaped quotes, an empty graph, the wrong script type and client-side injection leave check B1 nothing to read. How the scanner counts pages.

Checks B1

Time to first byte for bots, 1.5 seconds is the bar

Check A5 loses a point when the median bot fetch takes 1.5 seconds or more. Where the time goes, how to measure it with curl, and the fixes in order of effort.

Checks A5

Server-rendered parity, measuring the gap yourself

A Node script that computes check A7's raw-to-rendered text ratio, how to run it across your key routes, and what the number says about your rendering mode.

Checks A7

CAPTCHA challenges and agents, what a challenge page costs you

A challenge on every request is a closed door to an agent. How check A6 detects challenge markup, and how to scope challenges to login, checkout and signup.

Checks A6

Cookie walls are interstitials, the consent banner problem

Why check A6 treats a consent dialog as an interstitial, the markup it looks for, why a cookie-only wall scores one of two, and how going cookieless removes it.

Checks A6

XML sitemaps for agents, the three points you are probably dropping

How check A3 scores a sitemap, exists, parses, referenced from robots.txt, why the first sitemap declared is load-bearing for D4, and how ours is built.

Checks A3, D4

llms-full.txt, and when to publish the whole site as one file

What an llms-full.txt is for, how check A4 scores it, what belongs in it, how large is too large, and why ours comes from the same registry as the sitemap.

Checks A4

How to write an llms.txt that earns its place

The llms.txt convention, why A4 gives three points for three described links and one for a bare list, and a complete example file for a small SaaS company.

Checks A4

Google-Extended, what it controls and what it does not

Google-Extended is a robots.txt control token, not a crawler. What it switches off, what it leaves alone, three common misconfigurations, and what A2 records.

Checks A2

Your robots.txt is perfect and you are still blocking agents

A vendor-neutral tour of the edge layer, bot rulesets, bot scores, rate limits, JavaScript challenges and slow origins, and how each shows up in the A5 probes.

Checks A5

ClaudeBot and Claude-User are different things

Anthropic's crawler and its user-triggered fetcher share a prefix and get blocked together by accident. The robots.txt for each intent, and its A2 score.

Checks A2

Blocking GPTBot but not OAI-SearchBot, the distinction that matters

OpenAI documents three tokens so training, indexing and user fetches are decided separately. The robots.txt that opts out of training only, and its A2 score.

Checks A2

The six AI crawlers and what each one is for, a decision matrix

Which of the six A2 tokens a shop, a SaaS product, a publisher and a services firm should allow, what each combination scores, and what each block gives up.

Checks A2

robots.txt for the agent era, a practical rewrite

A line-by-line rewrite of a typical legacy robots.txt. What checks A1, A2 and A3 read from it, the REP rules a parser applies, and a finished file to adapt.

Checks A1, A2, A3

Agent readiness is not SEO, and not GEO either

Where SEO, generative engine optimisation and agent readiness overlap, where they diverge, and why ranking and being usable for a task are different problems.

Cloudflare blocks AI crawlers by default, before it reads your robots.txt

Cloudflare's default bot rules turn GPTBot, ClaudeBot and PerplexityBot away before robots.txt is read. How to check in thirty seconds, and what to change.

Checks A2, A5

Who is actually fetching your site, a field guide to AI user agents

The six crawler tokens check A2 scores, what each one is for, training, search indexing or a fetch on a user's behalf, and what blocking each one costs.

Checks A2, A5

Reading your scan report, a check-by-check walkthrough

What each section of an AgentFriendlyRank report means, what the evidence blocks show, how fixes are ranked, and how to decide what to do first.

The two fetches, what an agent sees versus what your browser sees

Most serious agent readiness findings live between the raw HTML and the rendered DOM. How check A7 measures the gap, and how to measure it yourself with curl.

Checks A7

The three tiers of fixes, hours, weeks and architecture

What the rubric's Low, Medium and High effort ratings mean in practice, which of the 27 checks fall into each tier, and what the work costs.

What agent readiness actually means, in one page

The four questions the rubric asks of a site, find, understand, act and trust, and why passing every SEO test says nothing about any of them.

Why most sites score zero on machine endpoints, and why that is the opportunity

Check C4 asks for an agents.json, an MCP manifest, API docs or an OpenAPI spec. Almost nobody publishes one, which makes it the cheapest differentiation.

Checks C4

Why we score out of 100 and publish the whole rubric

The case for a public, versioned scoring standard over proprietary AI visibility scores nobody can check, and why we never rewrite a historical score.