MachineSee
Introductory prices until 28 February 2027 · Full Report €79 · Fixes done €490 · Refund within 7 days if a report isn't usefulSee packages

How it works

We read your website the way machines do.

No eyes, no scripts, no login: only the code a crawler, an AI assistant or an agent receives. Then 42 checks, one score, and evidence for every result.

  1. 01 · ReadYour pages, as code

    Home page and five key pages, robots.txt, sitemap, 20 links. The Full Report reads 30 pages, 120 links and three competitors.

  2. 02 · Check42 checks, five areas

    The same list for every site. Each check is a pass, a fail or not applicable.

  3. 03 · ReportOne score, every result proven

    Each finding links to the page it came from, with the fix. Passes are listed too.

What a machine receives

A person sees a page. A machine gets a handful of facts.

If a fact is missing, wrong or added only by a script, the machine works without it.

Six facts a machine extracts from HTML, each with the element it comes from HTML AS RECEIVED <html lang="en"> <title>Boutique hotel in Southrop</title> <meta name="robots" content="noindex"> <script type="application/ld+json"> {"@type":"Hotel","telephone":"+44…"} <h1>Rooms, restaurant, spa</h1> <a href="tel:+441367…">01367 …</a> <img src="pool.jpg" alt=""> <div id="app"></div> ← added by script lime: a fact the reader can use red: a fact that works against the site grey: not delivered to a reader without scripts WHAT THE READER CONCLUDES Language: English (check U12) Subject of the page (U6) Do not index this page (A1) This is a Hotel; here is its phone (U3, U4, U5) What the page is about (U10) A number a phone can dial (C3) An image with no description (U13) Nothing: the text was never delivered (U1) Six facts a machine extracts from HTML, each under the line of code it comes from HTML ↳ WHAT A MACHINE CONCLUDES <html lang="en"> ↳ Language: English (U12) <title>Boutique hotel in Southrop</title> ↳ Subject of the page (U6) <meta name="robots" content="noindex"> ↳ Do not index this page (A1) <script type="application/ld+json"> {"@type":"Hotel","telephone":"+44…"} ↳ This is a Hotel; here is its phone (U3, U4, U5) <h1>Rooms, restaurant, spa</h1> ↳ What the page is about (U10) <a href="tel:+441367…">01367 …</a> ↳ A number a phone can dial (C3) <img src="pool.jpg" alt=""> ↳ An image with no description (U13) <div id="app"></div> ← added by script ↳ Nothing: the text was never delivered (U1) lime: a fact the reader can use red: a fact that works against the site grey: not delivered without scripts
Illustrative HTML. The codes (U12, A1…) are check identifiers from the list of 42 below.

Five areas, one score

What we check, and how much it counts.

  1. 30%

    Technical access

    Can crawlers and fetchers reach your pages at all?

    12 checks
  2. 20%

    Structure

    Is the business described in the code?

    11 checks
  3. 20%

    Content

    Is the text there without a browser?

    4 checks
  4. 15%

    AI discovery

    Can answer systems verify and cite you?

    7 checks
  5. 15%

    Agents

    Can software acting for a customer find the way to act?

    8 checks

Every finding says how we know

  • Observed in the page or file
  • Measured by a test we ran
  • Our interpretation of observed facts

And how urgent it is

  • Priority blocks a machine from reading or acting
  • Worth fixing gives an incomplete picture
  • Note small, but adds up

What we do not do

Honest limits.

  • No sampling of what ChatGPT, Gemini, Perplexity or Google say about you in the free check or Full Report. A person does that in Larger scope.
  • No running your JavaScript. We read the HTML as delivered, as some bots do.
  • No browser page-speed test. We measure how fast your server responds.
  • No automatic Google Business Profile check. A person does that in Larger scope.
  • No promises of rankings, mentions, traffic or revenue. Nobody controls those.
  • No invented findings. A healthy site gets a short report full of passes.

The details

Everything else, for the curious.

All 42 checks, with weights

Every site is run against the same list. Each check is reported as pass, fail (with the finding, its severity and the page it came from) or not applicable (for example, "phone numbers tappable" when no phone number is printed). Each check has a weight of 1 to 3 that feeds the score. The list is generated from the engine's code, so what you read here is what runs.

IDCheckWhat passing meansWeight
Technical accessibility · weight 30% · Can crawlers and fetchers reach the pages at all?
A1Home page may be indexedNo robots 'noindex' instruction on the home page.3
A2Search-engine crawlers allowedrobots.txt does not block Googlebot or Bingbot.3
A4Sitemap present and readableA sitemap is declared or found at /sitemap.xml and returns valid XML.2
A5Internal links resolveThe internal links tested return a page, not an error.2
A6Key pages may be indexedContact, about, services, pricing and review pages carry no 'noindex'.2
A7Canonical addresses stay on this domainNo page names another website as its official address.3
F1Served over HTTPSThe final address is https://.3
F2Mobile viewport on the home pageA viewport meta tag is present.3
F3Mobile viewport on other pagesEvery other page read has a viewport tag.2
F4Home page server response under 3 sHTML delivered within 3 seconds, measured once.2
F5Other pages respond under 3 sNo other page read took over 3 seconds.1
F6Home page HTML under 1.5 MBThe HTML itself is not unusually heavy.1
Machine-readable structure · weight 20% · Is the business described in the code, not only in pictures and prose?
U2Structured data is validEvery JSON-LD block parses and uses real schema.org types.2
U3Structured data describes the businessAn Organization, LocalBusiness or specific business type is declared.2
U4Address and telephone in structured dataThe business description carries address and telephone.2
U5Structured-data phone matches the pageThe machine-readable telephone is one shown to visitors.2
U6Home page title is descriptiveA title of 10 to 70 characters that is not a placeholder.2
U7Page titles are distinctNo two pages read share a title.1
U8Home page has a meta descriptionA description for search results and link previews.2
U9Key pages have meta descriptionsAt most one page read lacks a description.1
U10One main heading on the home pageExactly one H1 stating what the page is.2
U11Key pages have a main headingAt most one key page read lacks an H1.1
U12Language declaredThe html element carries a lang attribute.1
Content interpretation · weight 20% · Is the text itself available and readable without a browser?
U1Readable text without scriptsThe home page delivers its text in HTML, not only through JavaScript.3
U13Images have text descriptionsFewer than 30% of home-page images lack alt text.2
U15Business name consistentStructured data and site name agree on one name.1
C6Services or offer page linkedA page stating what is sold, linked from the home page.2
AI discovery readiness · weight 15% · Can answer systems verify and cite the business?
A3AI answer crawlers allowedrobots.txt does not block OAI-SearchBot, ChatGPT-User, PerplexityBot or Claude-SearchBot.3
U14Link-preview data completeOpen Graph title and image are present.1
C5Third-party trust signals presentReview-platform links, accreditations or review structured data on the pages read.1
F9Privacy policy or company detailsA privacy page or registration details on the pages read.2
F10Email on the business's own domainPublished addresses are not only free webmail.1
D1Profiles linked from structured dataThe business description lists its official profiles (sameAs).1
D2Last-updated dates for machinesThe sitemap or structured data carries modification dates.1
Agent readiness · weight 15% · Can a visitor, or software acting for one, find the offer, the price and the way to act?
C1A way to make contact from the home pageA phone link, email link, form, booking engine or contact link.3
C2Contact page reachable in its HTMLThe contact page carries a form, call link or email address in the HTML.3
C3Phone numbers are tappableEvery phone number printed on the home page is a tel: link.3
C4Clear call to action on the home pageA link or button that says book, enquire, contact, request or buy.2
C7Prices present in the pricing page HTMLWhere a pricing page exists, prices are in the HTML.1
G1Prices or a quote route on the pages readA price, or a request/quote/enquire action, is in the HTML of the pages read.2
F7Form fields labelledEvery form field has a label, aria-label, title or placeholder.2
F8Links have readable textFewer than three icon-only links without a label.1
How the score is calculated

Priority: blocks a system from reaching or acting on the site (a blocked crawler, a site without a mobile layout, a phone number nobody can tap). Worth fixing: machines get an incomplete or inconsistent picture (no structured data, a broken sitemap, half the images undescribed). Note: small, cumulative. The plan in the Full Report orders items by severity and then by effort, with a time estimate for a developer, instructions and the code to paste for each item.

The score. Each check carries a published weight of 1 to 3 (in the list of 42 checks). A category's score is the weight of its passed checks divided by the weight of its applicable checks, as a percentage; checks that do not apply are left out, never counted as passes. The overall score combines the five categories with the area weights (30 / 20 / 20 / 15 / 15). Bands: 85 and above Strong; 65 to 84 Sound, with gaps; 40 to 64 Weak; below 40 Not readable. Four checks are blocking: if the home page may not be indexed (A1), search crawlers are blocked (A2), the page has almost no readable text without scripts (U1) or the site is not served over HTTPS (F1), the band is capped at Weak whatever the number, because the site cannot be read or found. The score is computed from the ledger; no model writes it, and it does not measure traffic, rankings or what any assistant says. Two sites with the same score can differ in what to fix first; the plan, not the score, says what to do.

Why each area matters commercially
CategoryThe question it answersWhy it matters commercially
Technical accessibility (30%)Can crawlers and fetchers reach the pages at all? Indexing instructions, crawler rules, sitemap, links, canonical tags, HTTPS, mobile viewport, response time.If a system cannot reach or is told to ignore a page, nothing else about it matters. Google states its AI features have no requirements beyond being crawlable and indexable with text content. These are the cheapest problems to fix and the most expensive to leave.
Machine-readable structure (20%)Is the business described in the code? Structured data present, valid, naming the business with address and telephone that match the page; titles, descriptions, main headings, language.Search results, maps and assistants assemble their picture of a business from these fields. Missing or inconsistent facts are filled in from other sources, or left out.
Content interpretation (20%)Is the text itself available to a reader without a browser? Readable HTML without scripts, image descriptions, a page that states the offer, one consistent name.What a reader without scripts and without eyes can extract is the version of the business that gets passed on.
AI discovery readiness (15%)Can answer systems verify and cite the business? AI answer crawlers allowed; link-preview data; third-party trust signals; company details; a domain email; profiles declared; modification dates.Answer systems fetch pages at answer time and look for corroboration. We measure the observable prerequisites; we do not, and nobody can, measure inclusion.
Agent readiness (15%)Can a visitor, or software acting for one, find the offer, the price and the way to act? A contact route in the HTML, tappable numbers, a call to action, prices or a quote route, labelled forms, readable links.The visitor who is ready to call or book is the one you cannot afford to lose to a number that will not tap. Agents acting for users read the same HTML and need the same things.
What we read, and how we behave on your site

The free check fetches your home page, robots.txt, your sitemap and up to five key pages chosen from your own navigation (contact, about, services or rooms, pricing, booking, privacy), and tests up to 20 internal links. The Full Report extends this to 30 pages and 120 links, and runs the same checks on up to three competitors you name, the same day.

We identify ourselves with a named crawler, respect standard web ports and public addresses only, make at most a few dozen requests per site, and never fetch anything behind a login. If a site refuses our requests, we say so; we do not guess.

How findings are labelled

Every finding carries one of three labels. Observed: the fact is in the page or the file (a missing tag, a number printed as text). Measured: we tested it (a link returned 404, a page took 3.6 seconds to respond). Our interpretation: a reading of observed facts that a reasonable specialist could dispute (for example, that a page lacks a clear call to action). In the Full Report, text written with the help of an AI model is marked AI-assisted and may cite only the evidence we collected; anything it cannot cite is removed.

How we know the engine is right

Before any finding reaches a customer it has passed three gates. The engine's checks are deterministic: the same page gives the same finding, and each check has regression tests built from real sites where it was once wrong (a server that answered HEAD requests with 404, a robots.txt group that looked like a block and was not, a cart page flagged for not being indexed). Every Full Report is read by a person on the live site within one working day; corrections are emailed and the page updates. And every finding links to the page it came from, so you, or your developer, can check it yourself. We run the same checks on our own site.

Sources

The statements we make about how other systems read websites come from those systems' own documentation, read on the dates given.

  • Google Search Central, AI features and your website (read 13 Nov 2026): no additional requirements or special structured data for AI Overviews and AI Mode; pages must be indexable with text content; inclusion is never guaranteed.
  • Google Search Central, JavaScript SEO basics (read 13 Nov 2026): crawl, render queue, index; "not all bots can run JavaScript".
  • Google Search Central, Creating helpful, reliable, people-first content (read 13 Nov 2026): trust, sourcing, disclosure of AI-generated content.
  • OpenAI, Overview of OpenAI crawlers (revised 9 Dec 2025): GPTBot, OAI-SearchBot, ChatGPT-User and how robots.txt applies to each.
  • Anthropic, support article 8896518, Does Anthropic crawl data from the web: ClaudeBot, Claude-User, Claude-SearchBot.
  • Perplexity, crawler documentation: PerplexityBot and Perplexity-User.
  • Cloudflare, The crawl-to-click gap (29 Aug 2025) and Radar Year in Review 2025: crawl volumes and crawl-to-refer ratios, with the authors' caveats.
  • Ahrefs, AI search overlap (11 Aug 2025; 15,000 prompts): only 12% of links cited by ChatGPT, Gemini and Copilot were in Google's top 10 for the same prompt; Perplexity 28.6%. One study, one period.
  • BrightLocal, Local Consumer Review Survey 2026 (11 Feb 2026; 1,002 US adults): use of generative AI tools for local recommendations rose from 6% to 45% in a year. One commissioned US panel; question wording not published.
  • Directive (EU) 2019/882, the European Accessibility Act, applicable to services from 28 June 2025. We do not give legal advice; check your national transposition.
Your data

We store the address checked, the result, the time and a one-way hash of the requester's IP address. For a paid report we add the email and the competitors named. No cookies, no third-party analytics. Details in the privacy notice.

See it on your own site.

All 42 checks on your home page and five key pages, in about 20 seconds.

Run the free check →