Methodology

How the Free B2B AI Visibility Checker works

We read your website the way AI agents do: page code only, no browser, nothing typed or submitted. Here is every check, how we score it, and what this version can't see.

Research: website readability (35 points)

When an agent researches your category, can it read and understand you?

1.1

AI search crawler access

7 pts

Your robots.txt rules for OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot and Perplexity-User, on your homepage and key pages. Training bots (GPTBot, ClaudeBot, Google-Extended, CCBot) are reported, never scored.

Pass
None blocked
Warning
Some blocked, or key pages blocked
Fail
All blocked
1.2

Homepage access

7 pts

We request your homepage as a normal browser, then as ChatGPT-User, Claude-User and Perplexity-User, and compare what comes back. A block page served with a 200 counts as blocked.

Pass
Every agent gets the same page as the browser
Warning
Some agents blocked or challenged
Fail
All agents blocked while a browser gets in
1.3

Content without JavaScript

7 pts

Words of page content in the raw HTML of your homepage and up to 5 key pages (a product or service page, a customer story, a blog post, your about page). Navigation and footer text are not counted.

Pass
Homepage 200+ words and a typical key page 150+
Warning
Homepage under 200, or key pages under 150 (the report says which)
Fail
Homepage under 50 words, or an empty JavaScript app shell
1.4

Accidentally hidden pages

4 pts

noindex in a meta robots tag or X-Robots-Tag header.

Pass
None on the homepage or key pages
Warning
On a key page
Fail
On the homepage
1.5

Page essentials

3 pts

Title, meta description, exactly one main heading and a canonical URL on the pages we read.

Pass
90%+ present
Warning
60 to 89%
Fail
Under 60%
1.6

Company identity

3 pts

Organization structured data (JSON-LD) on your homepage.

Pass
Name, URL and a logo or official profiles
Warning
Name only
Fail
None
1.7

Sitemap

2 pts

The sitemap listed in robots.txt, or /sitemap.xml. For a sitemap index, we read up to 3 of the sitemaps it lists.

Pass
Valid, linked from robots.txt, updated in the last 180 days
Warning
Not linked, stale, or one of the sitemaps we read is broken
Fail
Missing or invalid
1.8

Response speed

2 pts

Time until your homepage starts responding: the median of the 5 homepage requests we make, measured from the Cloudflare location running the check.

Pass
Under 800 ms
Warning
800 to 2000 ms
Fail
Over 2000 ms or no response
1.9

AI guide files

+2 bonus pts

A valid llms.txt, and markdown versions of your pages linked with rel="alternate" type="text/markdown".

+1 each. Bonus points never cost you points.

Evaluate: product access, MCP and API (15 points)

When an agent evaluates your product, can it try it? Only scored when your site shows product signals: developer docs, an API, an app login or an MCP server.

2.1

Discoverable

4 pts

MCP discovery files on your own domain: /.well-known/mcp/server-card.json and /.well-known/mcp.json. We look only where an agent arriving at your site would look.

Pass
A discovery file points to your MCP server
Warning
Your site mentions an MCP server, but nothing points agents to it
Fail
No MCP server findable from your site
2.2

Working

4 pts

The MCP handshake: initialize, initialized, then a request for the tool list. We never call a tool.

Pass
Valid response within 5 seconds, or a correct login request
Warning
Slow or partial
Fail
Error response
2.3

Understandable

3 pts

The tool list: does each tool have a real description and an input schema? If your server requires login, we read your server card instead.

Pass
1 to 40 tools, 80%+ clearly described
Warning
Partly described
Fail
No descriptions
2.4

Connectable

2 pts

If your server requires login, whether it tells agents how (OAuth protected-resource metadata that we fetch and validate).

Open, or login with valid metadata. Bonus points never cost you points.

2.5

Documented

2 pts

A developer docs link on your homepage, and an OpenAPI spec we can fetch and validate (from your api-catalog or a homepage link).

Pass
Both
Warning
One
Fail
Neither

Reach out: agent conversion (50 points)

When an agent is ready to act for its buyer, can it reach your sales team?

3A

Find the way in

8 pts

A contact, demo or trial page linked from your homepage, or found in your sitemap or at a common address like /contact.

Pass
Linked from the homepage
Warning
Exists, but not linked from the homepage
Fail
None found
3B

Let in

10 pts

Your contact page requested as a browser and as the three AI agents, plus your robots.txt rules for that page.

Pass
Every agent gets the page
Warning
Some agents blocked or challenged
Fail
All agents blocked, or no contact page at all
3C

See the form

12 pts

A real contact form in the page code. Newsletter, search and footer forms do not count. We name the form tool when the form is loaded by a script (HubSpot, Marketo, Pardot, Typeform, Jotform).

Pass
Contact form in the page code
Warning
Form inside an embedded frame whose page has a real form: browser agents can use it, page-code readers can't
Fail
Form only built by JavaScript (on the page or inside the frame), or no form
3D

Understand the form

5 pts

Field labels and the number of required fields.

Pass
All labelled, 7 or fewer required
Warning
Some unlabelled, or 8 to 10 required
Fail
Mostly unlabelled, over 10 required, or no visible form
3F

Agent-friendly route

+3 bonus pts

A visible email address, ContactPoint structured data, and contact details in llms.txt or a contact tool on your MCP server.

+1 each. Bonus points never cost you points.

3G

Readable proof

5 pts

A customer story or case study page with 300+ words of page content.

Pass
Readable case study
Warning
Thin, image or PDF only, or behind a form
Fail
None found
3H

Self-serve path

3 pts

Only if your homepage offers a free trial or sign-up: whether an agent can reach and read the sign-up page.

Pass
Reachable, with a form or clear page
Warning
Loads only with JavaScript
Fail
Blocked

How we score

  • Pass earns full points, a warning earns half, a fail earns none.
  • Each step scores the points earned out of the points that applied, weighted Research 35, Evaluate 15, Reach out 50.
  • If your site shows no product signals, Evaluate is not scored and its weight moves to the other two steps (about 41 and 59).
  • Checks that don't apply, and checks we couldn't complete because of an error on our side, are left out. They never cost you points.
  • Bonus points add up to 5. The score is rounded and capped at 100.
  • 80 to 100 is Shortlist-ready, 50 to 79 is At risk, 0 to 49 is Off the shortlist.
  • If an agent can't reach your sales team at all (no contact page and no email, all agents blocked from your contact page, or no visible form and no other route), the score can't go above 79.
  • If we can't reach your homepage, or fewer than 60% of checks complete, we don't give a score.

How we request your pages

  • Our own user agent: VertoDigital-AIVisibilityChecker/1.0 (+https://vertodigital.com/b2b-ai-visibility-checker/tool/how-it-works). It follows your robots.txt.
  • For the access tests only (your homepage and contact page), single requests with the user agents the AI companies publish for ChatGPT-User, Claude-User and Perplexity-User.
  • At most 30 requests to your site per check, one at a time, 10 seconds per request, 60 seconds in total.
  • Requests go through the public internet like any visitor's, so your firewall and bot rules apply.

What we never do

  • Submit a form or type into a field.
  • Call a tool on your MCP server or log in anywhere.
  • Run a browser or execute your JavaScript.
  • Use AI to decide a result. The same site gives the same result.

What this version can't see

  • Content that only appears after JavaScript runs. Agents that drive a real browser may see it.
  • What happens when a form is submitted, including bot checks that only run on submit.

Public reports, private comparisons and removals

  • Each company gets one report page, updated when it is checked again. Reports are not indexed by search engines.
  • Comparisons with a competitor live on a private link, are never published and are deleted after 12 months.
  • Request an update or removal at any time.

Methodology version 2026-10-07.1

The check shows where agents drop off. The Assessment shows what it's costing you.

Our Pipeline Readiness Assessment is a complimentary 5-day diagnostic across paid efficiency, organic and AI search visibility, and pipeline data, with a 30-minute debrief from a specialist.