How we differ from other audits
This page lists what you actually get and how to verify it. The numbers are real and can be re-counted at any time, and every item on the list is something you can see on our own site. What we do not have yet is written down too.
checks, each with its own explanation and score impact
pages per scan on a paid plan, 30 on the free one
automated tests, so a fixed bug does not come back
guide articles in Lithuanian and English
What we do differently
Twelve things you can verify in a few minutes. The right column is what to ask another auditor about the same thing.
| What we do | What to ask the others |
|---|---|
| The report also lists what was not checked. A check that never ran looks the same on screen as one that passed, so we show that list separately. | Is there a place in their report that says what went unchecked? |
| If we crawled too few pages, we do not calculate a score. One page out of two hundred says nothing about a website. | What score does their tool show after crawling a single page? |
| We tell a closed port from a blocked one by measuring: at the same time we check a control port that has to answer. | How do they tell a closed port from a filtered one? |
| We have fixed seventeen of our own mistakes, and wrote a test for each so it cannot come back. | Do they have a list of mistakes they found and fixed? |
| When we change the score scale, we neither recalculate old reports nor compare them with new ones. Otherwise a changed scale would look like a changed website. | Does their score from last year still mean the same thing today? |
| The same problem lowers the score once, not across three separate lines. | How many times does their report punish the same missing page? |
| AI-suggested text is checked by the same rules we apply to a client site: sentence length, topic, repeated headings, and whether any invented numbers appeared. | What checks their AI-written text before a person sees it? |
| A new page is graded by the same checks before it is published, and its three versions are ranked by score. | Does their content tool grade its own draft? |
| Findings are grouped into work items by code, not by a model. Run it twice and you get the same list in the same order. | Run their tool twice — is the list the same? |
| We also read what JavaScript builds on the page. On one real WordPress site that revealed 144 % more text than the server sent. | What does their scanner see inside an empty block that JavaScript fills? |
| We join Search Console data with what the crawl found. That shows pages which have impressions in Google but answer with an error or are closed to indexing. | Does their audit know which page has impressions and returns a 404? |
| We do not claim to have things we do not have. What is still missing is listed at the bottom of this page. | Does everything declared in their site schema actually exist? |
What you get every week
An audit shows how a site looks today. But it changes without you knowing: after a system update, after an editor's fix, after a plugin adds noindex, or when server settings change. So we keep checking instead of checking once.
The first scan
Five scores from 0 to 100, and findings together with the URLs where the problem is. Up to 30 pages, no signup.
The weekly email
On Monday night we rescan stale domains, and at 10:00 we send the email. If nothing changed during the week there is no email: otherwise you would soon stop opening it.
We show the change
Every score comes with the change against the previous scan of the same domain, so you see at once whether it improved or dropped.
The history is kept
Scores, domain and findings are kept indefinitely, so a year later you can see how the site changed month by month. A free scan stays fresh for a day, a paid one for two weeks.
What else the report gives you
A list of work items
Sixty findings are grouped into 10-15 work items. The same problem across twelve pages is one job, not twelve separate tasks.
A fix that is ready to use
Text you can paste, JSON-LD with a note on where it goes, or a task together with how to check that it is done.
A file for the developer
Every technical work item in one markdown file. You can drop it into Jira or Notion and the code blocks stay intact, unlike copying out of a PDF.
PDF and XLSX
The PDF can be forwarded to a client or printed. The XLSX can be filtered and sorted. Both are built from the same data, so they cannot disagree.
Building a new page
A generated page is immediately run through the same checks we use to grade client sites. You see the score and the findings before publishing.
Search Console data
With Search Console connected, half of the crawled pages are picked by impressions, so the audit starts with the pages people actually see.
Standards we apply to ourselves
All of these can be verified on our own site from the outside, with a browser or your own scanner.
A markdown version
With an Accept: text/markdown header the same URL returns markdown instead of HTML, so an AI agent does not have to pick the text out of the layout. We add no separate .md URLs, to avoid duplicating the same pages in the index.
llms.txt and llms-full.txt
The list of URLs is assembled by code from the same source as the navigation and the sitemap, so it cannot go out of date.
Content-Signal in robots.txt
It states what the content may be used for: search, AI answers and training. The file holds only directives and the sitemap, with no comments.
Link headers
describedby points to llms.txt, and alternate points to the markdown version. This is described in RFC 8288.
auth.md
It describes how an agent registers and signs in, so it does not have to guess.
WebMCP
The scanning tool is declared on the form itself, so an agent can start a scan without separate instructions.
Schema.org data
Organization and WebSite on every page, and on the pricing page a Service with three Offer nodes. The prices in the schema and on screen come from one source.
RSS in both languages
Search Console accepts RSS as a sitemap — Google has recommended it for finding new content faster. The feeds are separate, because one feed declares one language.
What we do not have yet
These will appear once they are real. Claiming they already exist would be the same thing we treat as a mistake on client sites.
- agent-card.json — we have no agent of our own yet, so we do not declare a card for one.
- aggregateRating in the schema — there are no ratings yet, and stars without them would be an invented number.
- An empty sameAs list — we have not created social profiles yet, and a link to one that does not exist would answer 404.
- No automatic rescans on the free plan — one scan is about 35 requests to a server, so we do not make them unasked.
Start with what search and AI know about you today
The first scan is free and needs no account. The report stays available at its link.
Scan a domain →Updated