Site health

Site health checks the pages of your site for what search engines, AI assistants and AI crawlers need: titles and descriptions, headings, indexing, links, structured data and AI access. This page explains the Overview and Site health screens, lists every check with its severity, and explains when the checks run and how the score works.

To add a site and run a scan, see SEO & GEO. The scan itself is the same one Accessibility uses; see Scanning a site.

Overview

Card	What it shows
Site health	The site’s score out of 100, how many pages were checked and when the last scan finished. Green at 90 or above, amber up to 20 points below, red further below. See The score.
Problems to fix	How many problems there are, with counts by severity: high, medium and low. A problem shared by several pages (one duplicate title, one broken address) counts once. FYI items aren’t counted. Click the counts to open Site health.
AI search access	Open to AI search when robots.txt lets every AI search and user crawler in, Some AI search is blocked when it doesn’t, or Not checked yet before the first scan finishes. Below it, whether the site has an llms.txt: found, found but not in the usual format, or none.
Clicks from Google, 28 days	Clicks from Google search over the latest 28 days of Search Console data, with impressions, average position and the change in clicks against the 28 days before. The change only shows when the stored data covers those earlier 28 days. Click it to open Search. Before the first sync finishes it says “Search Console is connected; numbers appear after the first sync.”, and without Search Console it shows a dash and Connect Search Console. See Overview card.

Below the cards:

• Fix these first: the top 5 problems, high severity first, then the ones on the most pages. Each links to its card on Site health. The Pages column says how many pages it’s on, or “site” for a site-wide problem.
• AI crawlers: each AI crawler Startica knows, its operator, what it’s for, and whether your robots.txt lets it read the site’s start page (Allowed or Blocked). The card says when the crawler list was last checked. See AI crawlers and llms.txt.
• Coming next: the planned parts of SEO & GEO. See What’s coming.

Scan now and Check a page at the top start the same scans as on the Accessibility Overview.

The Site health screen

Site health lists every problem, grouped by area, in this order:

1. AI access
2. Indexing
3. Links
4. Titles, descriptions and headings
5. Structured data

Within each area, high severity comes first, then the problems on the most pages. Each problem is a card with:

• its title, a severity badge (high, medium or low) and how many pages it’s on;
• why it matters;
• How to fix;
• the pages it’s on, each a link that opens the page in your browser, with details where they help: the title and its length, the shared title, the headings found, the canonical address, where a noindex came from, which JSON-LD block can’t be read, which properties are missing, or which crawlers are blocked.

A broken link shows the address that doesn’t work, the HTTP status it answered with, and up to 10 pages that link to it (“and more” if there are others).

Problems shared by several pages are grouped so one fix covers them. For example, a title used on three pages is one card entry, “title” on 3 pages, and two different shared titles are two entries.

Good to know

Some findings are facts worth knowing, not problems. They’re listed last, under Good to know, with an FYI badge, and they never count against the score:

• a page asks not to be indexed, and isn’t in the sitemap;
• a page’s canonical points to another page;
• AI training crawlers are blocked.

Each is fine if it’s what you intended. The card says what to change if it isn’t.

When the checks run

Every page a scan checks is checked for search and AI at the same time as for accessibility, in the same browser load. The checks come in two kinds:

Kind	Checks	When they run
Page checks	Title, description, headings, noindex, canonical elsewhere, structured data on the page	As each page is checked. Check a page and the recheck buttons re-run them for those pages.
Site checks	Duplicate titles and descriptions, broken internal links, broken canonicals, orphan pages, missing sitemap, the home page’s structured data and social image, AI crawlers, llms.txt	When any scan finishes: a full scan, a page check or an issue recheck. They use the latest results of every page.

The sitemap is read fresh on every full scan. Which pages it lists matters for two checks: a noindex page in the sitemap, and orphan pages. If you change your sitemap, run Scan now to pick it up; a page check uses the sitemap from the last full scan.

robots.txt and llms.txt are fetched from the site when a scan finishes, so the AI crawlers table and the llms.txt status are as of the last scan.

Every check

Severities: high problems are likely to keep pages out of search or AI answers; medium ones hurt how pages are shown or understood; low ones are smaller improvements; FYI items don’t count.

AI access

Check	Severity	Reported when	How to fix
AI search or answer crawlers are blocked	high	robots.txt stops an AI search crawler, or a crawler that fetches pages people ask about, from reading the start page.	Allow the search and user crawlers listed. You can keep the training crawlers blocked.
AI training crawlers are blocked	FYI	robots.txt stops an AI training crawler from reading the start page.	Nothing, if that’s your choice.
No llms.txt	low	/llms.txt doesn’t answer with HTTP 200, or answers with a web page (HTML) instead of a text file: that’s usually the site’s catch-all page for addresses that don’t exist.	Publish /llms.txt: a # title, a one-paragraph summary, then links to your most important pages.
llms.txt doesn’t follow the expected format	low	/llms.txt exists but doesn’t start with a Markdown # title.	Start the file with # Your product name and a one-paragraph summary.

See AI crawlers and llms.txt for how robots.txt is read, and examples of both files.

Indexing

Check	Severity	Reported when	How to fix
Missing pages answer as if they exist (soft 404)	medium	An address that can’t exist on the site (Startica asks for one made-up address after each scan) answers HTTP 200 instead of 404.	Make unknown addresses answer HTTP 404; with a catch-all route, call the framework’s not-found page when nothing matches.
Page is in the sitemap but asks not to be indexed	high	The page’s robots meta tag or X-Robots-Tag header says noindex (or none), and the sitemap lists it.	Remove the noindex if the page should be found, or take it out of the sitemap.
Canonical points to a page that doesn’t work	medium	The page’s canonical points to another page on the site that the scan couldn’t load.	Point the canonical to a working page, usually the page itself.
Page is only in the sitemap: nothing links to it	low	The sitemap lists the page, but no checked page links to it. The start page is never reported.	Link to it from a related page or the navigation, or remove it from the sitemap if it’s obsolete.
No sitemap found	low	The last full scan found no page listed in a sitemap.	Publish /sitemap.xml and list it in robots.txt with a Sitemap: line.
robots.txt is a web page, not a text file	low	/robots.txt answers HTTP 200 with an HTML page, so crawlers find no rules and no Sitemap: line. Its rules are then treated as allowing everything.	Serve a plain-text /robots.txt with a Sitemap: line.
Page asks not to be indexed	FYI	The page says noindex and isn’t in the sitemap.	Nothing, if that’s intended. Otherwise remove the noindex.
Canonical points to another page	FYI	The page’s canonical address isn’t the page’s own address. Fragments (#…) and query strings are ignored when comparing.	Nothing, if the page is a copy or variant. Otherwise point the canonical to the page itself.

Links

Check	Severity	Reported when	How to fix
Broken internal link	high	A checked page links to a page on the same site that answers with HTTP 400 or higher.	Fix or remove the links, or redirect the old address to the right page.

Only links within the site are checked. Links to other sites aren’t followed.

Titles, descriptions and headings

Check	Severity	Reported when	How to fix
Missing page title	high	The page has no <title>, or it’s empty.	Add a title that says what the page is, in about 50 to 60 characters.
Several pages share the same title	medium	Two or more checked pages have the same title, ignoring upper and lower case and spaces at the ends.	Give each page its own title.
Missing meta description	medium	The page has no <meta name="description">, or it’s empty.	Add one of about 120 to 160 characters that says what the page offers.
No main heading (h1)	medium	The page has no <h1>.	Add one <h1> that names the page’s topic.
Title is too long	low	The title is longer than 60 characters.	Shorten it to about 60 characters, with the important words first.
Title is very short	low	The title is shorter than 10 characters.	Say what the page offers and for whom, in 30 to 60 characters.
Meta description is too long	low	The description is longer than 160 characters.	Shorten it to about 160 characters.
Several pages share the same meta description	low	Two or more checked pages have the same description.	Write a description for each page.
More than one main heading (h1)	low	The page has two or more <h1> elements.	Keep one <h1> and make the others <h2>.
Home page has no social preview image	low	The start page has no og:image meta tag.	Add og:title and og:image meta tags to the home page, and ideally every page.

Lengths count characters, after trimming spaces at the ends.

Structured data

Check	Severity	Reported when	How to fix
Structured data (JSON-LD) can’t be read	medium	A <script type="application/ld+json"> block on the page isn’t valid JSON. The page’s entry says which block (counting from 1).	Fix the JSON (a missing comma or quote is typical) and check it with a schema validator.
Structured data is missing required properties	low	A block of one of the types below is missing a property it needs. Reported once per type, with the pages it’s on.	Add the missing properties listed on the page.
Home page has no structured data about the product	low	The start page has no Organization, WebSite, Product, SoftwareApplication or LocalBusiness data.	Add a JSON-LD block to the home page: Organization (name, url, logo) and Product or SoftwareApplication (name, description, offers).

The properties Startica checks for each type are a practical subset of what search engines ask for, not every property they can use:

Type	Needs
Organization	name and url
WebSite	name and url
Product	name, and one of offers, review or aggregateRating
SoftwareApplication	name and offers
FAQPage	mainEntity
Article, BlogPosting	headline
BreadcrumbList	itemListElement

Blocks inside an @graph are checked too. Other types are recognized but not checked for properties.

The score

The score is Startica’s own measure of progress. It isn’t a ranking, a prediction of traffic, or a score any search engine uses.

For each page, the score starts at 100 and loses points for each problem on the page:

Severity	Points
high	15
medium	7
low	2
FYI	0

A page doesn’t go below 0. For example, a page with a missing description (7) and two h1s (2) scores 91.

The site’s score is the average of its checked pages’ scores, less the site-wide problems: those that belong to the site rather than a page (blocked AI search crawlers, no sitemap, no or malformed llms.txt), and broken internal links. Each costs 10 points if high, about 4.7 if medium and about 1.3 if low, and together they take off at most 30 points. The score is kept to one decimal place, shown as a whole number, and doesn’t go below 0.

Some notes:

• Only checked pages count. Skipped and failed pages have no score.
• A broken link is recorded on the address that doesn’t work. That address failed, so it has no page score of its own; each broken link counts against the site’s score instead (10 points, within the 30-point limit).
• The home page’s structured data and social image checks count against the home page’s score.
• The Overview colors the score against 90. Unlike Accessibility, SEO & GEO has no target score setting.

Fix and recheck

1. Start with Fix these first on the Overview.
2. Open the problem on Site health and read why it matters and how to fix it.
3. Fix it on the site and deploy the change.
4. Click Check a page on the Overview for a page you fixed. Its page checks run again, and the site checks are recomputed when it finishes.
5. For changes to your sitemap, or to fix problems across many pages, run Scan now.

Recheck page and Recheck N pages under Accessibility run the same checks, so a recheck there updates Site health too.

Every problem (not FYI items) is also on SEO & GEO → Opportunities, ranked against what Search and AI visibility found. For title and description, structured data and llms.txt problems, the AI can draft the fix there for you to copy. See Opportunities.

Limits

The checks look at what a page tells crawlers in its HTML and headers. They can’t tell:

• whether a title or description is good, only whether it exists and how long it is;
• whether the content answers what people search for;
• how search engines or AI assistants rank or cite the site;
• whether a crawler actually follows robots.txt or reads llms.txt.

Not checked yet: redirect chains, how many clicks a page is from the home page, broken links to other sites, and language alternates (hreflang).

Related

• SEO & GEO
• AI crawlers and llms.txt
• Opportunities: Site health problems ranked with the rest, and fix drafts
• Scanning a site
• Try the demo: sample results for fieldnote.example