Cairn
Contact us
Scoring grid 2026.10.9: 62 criteria, no AI

Every morning, your sites' score and what moves it.

Cairn reads your sites the way Google, Bing and AI answer engines read them. A score out of 100 on SEO, GEO and AIO, the issues to fix in order, and for every point gained or lost, the reason.

Request access Understand the method
  • No AI in the score
  • One check a day, from 5 a.m.
  • Up to 10 competitors compared

A score with no AI, so no model bias.

No language model reads or judges your pages. Cairn applies 62 verifiable, written, public criteria: the same site on the same day gets the same score.

Rules, not opinions

An AI that scores changes its mind from one run to the next and favors what resembles it. A written rule does not. Each criterion states what it checks, why it matters, how much it weighs and the fix that resolves it. Here is one, as Cairn applies it.

SEO, high severity, established evidence

Canonical URL

  • A single absolute tag pointing to the page itself1
  • A tag pointing to another address0.5
  • Missing, multiple, relative or pointing to an error page0

The fix: Give each page a single <link rel=canonical> tag with an absolute address (https://…) that points to the page itself.

Source: How to specify a canonical URL, Google Search Central.

Lean by design

No model to run, no browser to launch. Cairn reads the HTML and a few text files, like a search crawler, once a day.

1
check per day per site
0
calls to a language model
0
browsers launched
6
connections at most, at the same time

Every rule says where it comes from

So you don't chase points that may bring nothing, each criterion carries its evidence level; the most uncertain weigh little.

Established documented by the search engine or a standard
35
Reported observed in published studies
22
Uncertain effect plausible, no public evidence, low weight
5

A blocking issue doesn't get lost in the average

Good details must not hide a locked door. Homepage excluded from the index or search engines blocked: the overall score can no longer exceed 40, whatever happens elsewhere.

The method, plainly

Three readings of the same site, combined into a score out of 100. SEO weighs the most: nothing gets cited without first being found.

SEO search engines

40%

Can Google and Bing crawl the site, index it and present it well?

33 criteria

GEO AI answer engines

30%

Can answer engines (ChatGPT, Claude, Perplexity, Copilot) read the site and quote passages from it?

12 criteria

AIO brand and trust

30%

Is the brand recognized, and do the pages carry it?

17 criteria

What happens every morning

  1. 1

    Read

    From 5 a.m., the homepage and 8 pages in full, 30 more in a skim, up to 200 sitemap addresses verified. Then robots.txt and AI crawler access.

  2. 2

    Check

    Each criterion returns a score from 0 to 100. What does not apply, or could not be measured that day, is left out of the calculation instead of counting as zero.

  3. 3

    Weigh

    Criteria weigh according to their importance, pillars according to their weight. A blocking issue caps the score.

  4. 4

    Explain

    The issues to fix in order, with the fix and the points it would recover, and the reason for every change since the day before.

What the score is not. It measures verifiable best practices; it predicts neither a ranking nor a citation. No one knows those formulas, and Cairn does not claim to guess them.

Read the method

Every point gained or lost has a reason.

Every morning, Cairn compares the check with the previous day's, criterion by criterion. When the scoring grid grows, the step in the curve is not mistaken for a decline of the site.

Example: a site's score tracked over 90 daysThe score goes from 58 to 72 out of 100. canonical fixed (+4.1), sitemap returns 404 (−2.6), sitemap restored (+2.6), structured data (+5.3), new scoring grid (−1.4), titles rewritten (+3.8), AI crawlers allowed (+2.2).60708090+4.1 canonical fixed−2.6 sitemap returns 404+2.6 sitemap restored+5.3 structured data−1.4 new scoring grid+3.8 titles rewritten+2.2 AI crawlers allowed90 days agothis morning
Example: a site tracked for 90 days. Each cairn marks a day when the score moved, and the sign says why. The day the scoring grid changed is counted separately: it is not the site that declined.
site
What the site fixed, or let slip. The only points that are earned.
grid
What a change in the method shifted. Counted separately, on the same day.
measurement
What a measurement that became possible, or went silent that day, added or removed.

September

63.8 → 70.6 +6.8

site +8.2 grid −1.4
  • Valid structured datasite+5.3
  • Page titlessite+2.9
  • A single URL for the homepagegrid−1.4

Connected to your assistant

Cairn speaks MCP. The assistant working on a site's code reads its recommendations, makes the fixes, then runs a new check to verify. The AI acts on the results; it does not produce the score.

  • Recommendationsthe issues in the order to fix them
  • Run a checka new measurement after a release
  • Site reportscores, changes, failing criteria
  • Competitorsstrengths, weaknesses, what's new

Frequently asked questions

Another question? Write to us.

Why not have an AI score the pages?

Because a score must be verifiable. A language model answers differently from one run to the next, cannot say why, and favors what resembles what it has read. Cairn's criteria are written rules: anyone can redo the measurement by hand and get the same score.

Does the score predict my ranking or my citations?

No. It measures the share of best practices the site follows, those described by Google, Bing, reference tools and published studies. A good score opens the doors; it does not guarantee you will walk through them.

What does Cairn read on my site?

What a search crawler reads: the homepage and a sample of pages drawn from the sitemap, robots.txt, the sitemap itself, headers and structured data. Public sources are added: PageSpeed Insights for speed, Wikidata and Common Crawl for presence in open databases.

How many resources does a check use?

Very few: one check a day, HTML and text files, never more than six connections at once. No language model, no GPU, no browser launched on our side; speed is measured once by PageSpeed, on Google's servers.

And when the method changes?

The scoring grid has a version number. On the day it changes, the difference it causes is attributed to the grid, not to the site: your curve does not show a decline you did not cause.

See your sites' score, and what would raise it.

Cairn currently tracks the sites of SubAlpine and its clients. Write to us: we add your sites, and we read the first score together.

contact@subalpine.it Already have access? Sign in