Scan my site free

this is a sample

A real measurement of a mid-sized industrial supplier, with the company's identity removed. We do not publish anybody's audit under their own name — including for our own marketing. Built from the report's own components and the same copy, so it cannot drift away from what you would receive.

example-supplier.com

needs work

Your website is being skipped, not rejected.

Nothing here is broken in the way a developer would call broken. Your pages simply do not give an answer engine what it needs to name you — so it names somebody else. Every item below is a specific reason why.

41score
  • Access34
  • Structure39
  • Content47

based on 19 page(s) · rubric v2

GraidLab Readiness Score

Measures how much of this site is implemented against the GraidLab rubric for access, structure and content, weighted 30 / 30 / 40. It is not a probability that an AI assistant will cite or recommend you, and it is not a score any search engine or AI provider recognises. What those systems actually did with your site is measured separately, further down.

36

issues found

+58.4

points recoverable

12/19/5

critical / important / minor

start here

Let the AI crawlers in

One small file at the root of your domain decides which machines are allowed to read you. Right now it is not saying what you think it is saying.

10 min · +8.2 ptsdeveloperopen it →

Chosen for what it returns against what it costs, not for how alarming it is. Do this one, then work down the plan below.

ai visibility

What happens when your buyers ask.

5%

3 / 58

answers citing you

10%

6 / 58

answers naming you

measured on 12 buyer question(s) × 58 run(s)

how this was measured
engines asked
perplexity (34×), google_ai_overview (24×)
measured on
2026-08-12
language
cs-CZ
questions · repeats each
12 · 4–5×
cited
The answer linked to your domain.
named
The answer used your brand name, with or without a link.

Every run is an independent question put to a live engine, so results move between runs — that is why each question is asked several times and the result is reported as a frequency rather than a yes or no. Repeats can differ per question where an engine failed to answer and the run was not retried. This is a sample of a system that changes, not a fixed measurement of your site.

cited runs per question

  • 0/5fasteners supplier with material certificates
  • 1/4bulk fixings for construction companies
  • 0/5where to buy DIN 933 bolts in bulk
  • 2/5industrial fastener distributor central europe

cited instead of you

  • wuerth.cz9×
  • fabory.com6×
  • haberkorn.cz4×

your plan

Three sittings, and what each one is worth.

  1. This week — under half an hour each

    2 to fix · about 20 min · +12.6 pts

    4154

    1. 01Let the AI crawlers in+8.2 · 10 min
    2. 02Write titles that answer, not titles that label+4.4 · 10 min

  2. This month — an afternoon at a time

    6 to fix · about 11.3 h · +35.3 pts

    5489

    1. 01Describe the business in structured data+8.0 · 42 min
    2. 02Make it obvious who is behind this+6.8 · 45 min
    3. 03Show when the page was last true+5.1 · 1.3 h
    4. 04Write so that a machine can quote you+4.1 · 1.5 h
    5. 05Answer the questions people actually ask+5.2 · 3 h
    6. 06Serve the words without JavaScript+6.1 · 4 h

readiness today 41 · projected, all three done 89

The projection is arithmetic on the same rubric, not a forecast: each point is what the score recovers when that item reaches full marks. It says nothing about how often an assistant will cite you — that is measured, not projected, and it is above.

what to fix

Everything holding you back.

Shortened. The report this came from carried 36 findings across 17 jobs; these are 8, chosen to show the range rather than the length. Everything else about them — the evidence, the fixes, the way you group and sort them — is exactly what you get.

Every finding says what its rule rests on: documented means a provider has published that it affects crawling, indexing or inclusion, and links to where they say it. Best practice means established web practice that no provider has tied to AI visibility. Experimental means an emerging convention whose adoption is not settled.

Can AI reach you at all?

2 to do · about 4.2 h · +14.3 pts

Before anything else can matter, a machine has to be able to open your pages and read what is on them. These are the things standing in the way.

  1. critical

    Let the AI crawlers in

    +8.2 pts

    the problem

    documentedsource

    One small file at the root of your domain decides which machines are allowed to read you. Right now it is not saying what you think it is saying.

    1. Your robots.txt refuses crawlers that build the indexes answer engines retrieve from. While that stands, no amount of good content will get you into their answers — they cannot read the page in the first place.

      4 of 8 AI crawlers blocked

    the fix

    Your robots.txt blocks four of the crawlers that answer engines retrieve from, including the one behind ChatGPT's search. While that stands, nothing else in this report can help you — those engines cannot read the page in the first place.

    1. 01Open /robots.txt and remove the Disallow lines that name OAI-SearchBot, Claude-SearchBot, PerplexityBot and Googlebot.
    2. 02Add the block below, which permits retrieval while leaving your decision about model training untouched.
    3. 03Confirm with curl that the file no longer refuses them.

    where: https://example-supplier.com/robots.txt

    User-agent: Googlebot
User-agent: OAI-SearchBot
User-agent: Claude-SearchBot
User-agent: PerplexityBot
Allow: /
    • done whenrobots.txt lets Googlebot, OAI-SearchBot, Claude-SearchBot and PerplexityBot reach every page you want quoted.
  2. critical

    Serve the words without JavaScript

    +6.1 pts

    the problem

    documentedsource

    Most answer engines do not run your JavaScript. Whatever appears only after the page boots is, as far as they are concerned, not on the page at all.

    1. affects 14 pages

      Most of your text only appears after the browser runs code. Google Search does render JavaScript, but the crawlers behind the answer engines largely do not — so where a visitor sees a full page, several of the systems this report measures see a nearly empty one.

      only 31% of the text is readable without JavaScript

    the fix

    Two thirds of your text appears only after the browser runs JavaScript. Google Search renders it; the crawlers behind the answer engines largely do not, so most of what a visitor reads is invisible to them.

    1. 01Ask your developer which rendering mode the site uses. On most stacks this is one setting.
    2. 02Server-render or pre-render at minimum the body text, the headings and the internal links.
    3. 03Check by disabling JavaScript and reloading: the words a buyer needs should still be there.
    • done whenTurn JavaScript off and reload: the main text, the headings and the links are all still there.

Does it understand what your pages are?

2 to do · about 52 min · +12.4 pts

A machine reads your pages without the visual design you see. These are the places where nothing tells it what it is looking at.

  1. critical

    Describe the business in structured data2 to fix

    +8.0 pts

    the problems

    One script tag, written once and dropped into your template, tells a machine what your company is, what this site is and where the reader is standing. Everything below goes into that one block.

    1. 01No structured data+5.4 ptsbest practice

      affects 12 pages

      Nothing on these pages tells a machine what they are — a company, a product, an article. Without it a reader has to infer the answer from your prose, and gets it wrong often enough to matter.

      0 structured data blocks

    2. 02No Organization markup+2.6 ptsdocumentedsource

      affects 12 pages

      Nothing on your site formally states your company's name, address and identity in a way machines read. That is what connects your site to a recognised business.

    the fix

    01No structured data

    Your pages carry no structured data at all, so nothing states in machine-readable form that you are a company, what you sell, or how to reach you. Adding it is one block of code in your template.

    1. 01Paste the block below into the <head> of every page, replacing the placeholder values with your own.
    2. 02Check it with Google's Rich Results Test before you deploy.
    3. 03Add Product markup to product pages afterwards — Organization first, because it is the one that establishes who you are.

    where: inside <head>, on every page

    <script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Your company",
  "url": "https://example.com",
  "description": "Bulk supplier of industrial fasteners…",
  "areaServed": "CZ",
  "contactPoint": {
    "@type": "ContactPoint",
    "telephone": "+420…",
    "contactType": "sales"
  }
}
</script>

    02No Organization markup

    The same block covers this one. Once the Organization entry above is in place, fill in the two fields that turn it from a company that exists into a company an assistant can describe.

    1. 01Add sameAs with the URLs of your company profiles — LinkedIn, the trade register, anywhere your name is already confirmed.
    2. 02Add address with the real registered address. This is the field that decides whether you appear in answers about your region.

    done when

    • Each page carries structured data describing what that page is, and the site-level Organization and WebSite entities exist once on the homepage.
    • The homepage carries one Organization block with your name, URL and contact details, and it parses as valid JSON-LD.
  2. critical

    Write titles that answer, not titles that label

    +4.4 pts

    the problem

    documentedsource

    The title and the description are the two lines a machine reads before it decides what a page is about. Fixing them is one pass through your pages.

    1. affects 2 pages

      This page has no title. It is the single most important line for both search results and AI answers, and yours is blank.

      title is 0 characters

    the fix

    Two of your news articles ship with no <title> element at all, so search engines and answer engines have nothing but the on-page heading to name them by.

    1. 01Add a title element to both article templates, built from the article's own headline.
    2. 02Keep the company name at the end, not the start — the distinctive part should come first.

    where: inside <head> on the article template

    <title>New production plant to create 200 jobs | Example Supplier</title>
    • done whenThe page returns a title element.

Is there anything here worth quoting?

4 to do · about 6.6 h · +21.2 pts

An assistant answers by lifting a passage from somewhere and naming who wrote it. These are the reasons it is lifting from somebody else.

  1. critical

    Make it obvious who is behind this

    +6.8 pts

    the problem

    best practice

    An assistant will not stake an answer on a site that does not say who wrote it, who runs it, or how to reach them. Cheap to fix, and it gates everything else.

    1. A reader who lands here cannot quickly tell who you are, what you sell and who it is for. If a person cannot, an assistant certainly cannot.

      assessed 1 out of 3

    the fix

    Nowhere on your homepage does one plain sentence say what your company sells and who it sells to. A machine reading the page has to infer it from your slogan, and it infers wrong.

    1. 01Put one sentence directly under the main heading: what you supply, to whom, and where. Write it the way you would say it on the phone.
    2. 02Repeat the same wording in the page title and the meta description, so all three agree.
    3. 03Remove the slogan from the first screen, or move it below the sentence. It reads as decoration to a machine, and it is what is currently being quoted.

    where: directly after your <h1>

    <p class="lede">
  We supply industrial fasteners and fixings in bulk to construction
  and manufacturing companies across Central Europe, from stock,
  with material certificates.
</p>
    • done whenThe first sentence under your main heading names what you supply, to whom and where — and the page title says the same thing.
  2. important

    Answer the questions people actually ask

    +5.2 pts

    the problem

    best practice

    Assistants answer questions. A page written as a brochure gives them nothing to lift; a page written as an answer gives them a paragraph they can use whole.

    1. affects 9 pages

      The page describes what you offer but does not answer what a buyer needs to know to decide. Those unanswered questions are being answered by someone else.

      3 of 104 headings are a question a buyer would ask

    the fix

    Your pages are written as descriptions rather than as answers. Three of a hundred and four headings phrase a question a buyer would actually type, and assistants answer by lifting the passage that answers the question they were asked.

    1. 01Write down the six questions your sales people answer on every first call.
    2. 02Put each one on the relevant page as a heading, with a direct answer in the first paragraph under it.
    3. 03Answer in the first two sentences. Everything after that is context, and context is not what gets quoted.
    • done whenFor each question your buyers ask before buying, a heading on your site asks it and the answer is directly underneath.
  3. important

    Show when the page was last true2 to fix

    +5.1 pts

    the problems

    Undated text is treated as text of unknown age, and unknown age loses to a dated page every single time — even when yours is the better answer.

    1. 01No publication or update dates+3.1 ptsbest practice

      affects 12 pages

      Nothing says when this was written or updated. Undated text is text of unknown age, and a reader with no way to date it has no way to judge whether it still holds.

      no publication or update date on the page

    2. 02Content has not been updated+2.0 ptsbest practice

      Nothing on the site has been updated in a long time. A page with no sign of maintenance gives a reader — human or machine — no way to tell whether it is still true.

      nothing updated for 780 days

    the fix

    01No publication or update dates

    None of your pages carry a published or updated date, visibly or in the markup, so nothing on the site can be dated by anything reading it.

    1. 01Add a visible 'last updated' line to article and reference pages.
    2. 02Mirror it in the markup with datePublished and dateModified so it can be read without parsing prose.

    02Content has not been updated

    Nothing on the site has changed in over two years. A page with no sign of maintenance gives a reader — human or machine — no way to tell whether it is still true.

    1. 01Pick the five pages that matter commercially and bring their figures up to date.
    2. 02Show the date you did it. An updated page nobody can date reads the same as an abandoned one.

    done when

    • Pages that have a date show it, both on the page and in the markup.
    • The most recently updated page carries a date from the last few months.
  4. important

    Write so that a machine can quote you

    +4.1 pts

    the problem

    best practice

    A quotable sentence is short, concrete, and still true when it is lifted out of the page. These are the places where yours are none of those things.

    1. affects 9 pages

      The text is short on checkable facts — numbers, dates, specifics. A claim with a figure in it can be verified and attributed to you; an adjective cannot.

      0.4 concrete facts per 100 words

    the fix

    Your pages make claims without numbers behind them. An answer engine quotes the source that says something checkable, and there is very little here it could quote.

    1. 01Take the three claims you make most often and put a number on each: how many items in stock, what delivery time, what minimum order.
    2. 02Replace every 'wide range' and 'many years' with the actual figure.
    3. 03Put those figures in the body text, not in an image — text in images is invisible to the engines.
    • done whenYour three most repeated claims each carry a number.

Yours would be about your site.

Same structure, your pages, your numbers, and the questions your buyers actually ask.

Scan my site free

Start free. No card until you have seen your score.