Site Health, Redirects & Canonicalization

Technical health is the part of SEO nobody notices until it stops working. A site the search engines struggle to fetch and interpret cannot be written out of the problem, and it cannot be linked out of it either. This guide covers the running maintenance that keeps the base clean: health score, redirects and canonicalization.

What a health score is

You cannot improve what you do not measure. So tools like Ahrefs and Semrush have built crawlers that imitate the search engines', work through your site, and summarize its technical health as a single number.

The calculation is plain:

Health score = (1 - (URLs with errors / all URLs)) × 100

The number is the share of addresses without faults. With 2,000 URLs where 500 have errors, the sum is (1 - (500 / 2,000)) × 100 = a health score of 75.

It is not a ranking factor. It is an indicator, and its value lies in being comparable over time. When it falls, you know something has broken, often alongside a release.

Why the technical side comes first

A poor health score cannot be balanced out with excellent content or a strong link profile. If the page cannot be fetched, if it gets canonicalized away, or if the metadata is missing, the quality of the writing is beside the point. That is why we go through the technical layer before touching the content plan, both in new engagements and after any significant development work.

The faults that most often pull the number down

The list of possible errors is long, and in practice the same handful keeps recurring.

• HTTP 4XX. The request fails, usually a link to a page or an image that has been removed.
• Image faults. The file is too heavy, in the wrong format, or cannot be fetched from the server.
• Missing metadata. The crawler finds a page with no page title or meta description, or with either one implemented incorrectly.
• JavaScript faults. Unused files, or files the browser cannot load.
• Missing HTML attributes. Canonical and hreflang above all, the two that declare duplicates and language versions.

Five fixes that move the number most

1. Keep the images in proportion

Images with oversized file weight load slowly and drag the whole page down with them. Save in a compressed format such as WebP or JPEG before uploading, and make sure the dimensions match the space on the page. A product image displayed at 200 × 200 pixels has no business being 3,000 pixels wide.

A CDN helps as well. The images then get served from a server near the visitor instead of from your own, which shortens load time noticeably. The effect shows up directly in your Core Web Vitals.

2. Clean up the source code

A common fault is the crawler finding elements in the code that should not be there: a reference to a deleted image, a call to a function nobody uses any more, an unused JavaScript file still sitting on the server. Go through the code at regular intervals, ideally as an item in your release routine.

3. Give every page its metadata

Page titles and meta descriptions are not read by the visitor as such, but they decide how you appear in the results and they help the engine interpret the page. Missing, they pull the number down. Every publishing tool has the fields, so this is about routine, not technology. How to write them is covered in titles and meta descriptions.

4. Keep canonicals and hreflang consistent

The canonical attribute declares that several pages hold the same or similar content, and which of them should be indexed. Hreflang is used when the same content exists in several language versions, for instance a Danish and a Norwegian variant.

Three faults are particularly common:

• Canonicalizing to the wrong page. Paginated pages in a listing should point at themselves, not at page one. Point them at page one and everything past the first page is blocked from being indexed.
• Contradictory signals. The canonical says one thing, the sitemap another and the internal linking a third. The engine then picks for itself, and it does not always pick what you wanted. See crawling and indexing.
• Incomplete hreflang. Every variant has to point at all the others and at itself, or the whole set gets ignored.

5. Invest in well-built code

Newer sites on a modern codebase often rank better, for the simple reason that good code produces fewer faults. It is rarely the most enjoyable investment to argue for internally, and it pays for itself in every subsequent effort.

Redirects: the most common source of lost traffic

A redirect sends the visitor and the search engine on from one address to another. The rule is short: use a 301 when the move is permanent, and a 302 only when it genuinely is temporary. A 302 left in place for a year is one of the most expensive pieces of carelessness in technical SEO.

Three things to keep in order:

• Avoid chains. A leads to B which leads to C. Point A straight at C. Every extra hop costs load time and signal strength.
• Never to the front page in bulk. A removed product page should go to the nearest relevant category page. Send everything to the front page and it usually gets treated as a soft 404.
• Map before a migration. In a move or a new site structure, every old URL needs a documented destination before release, and the list has to be verified afterward.

One thing worth saying: a 404 is not always a fault. A page that never had traffic or links can perfectly well return 404. The addresses that need a 301 are the ones with incoming links or a history of traffic.

Robots.txt and the self-inflicted blocks

Robots.txt governs what crawlers may fetch. Set up wrongly it is among the most serious technical faults there is: a Disallow: / at the top of the file stops the crawl of the entire site and takes the traffic with it.

Two things are worth checking regularly. That no important sections are blocked by accident, often a leftover from the staging environment. And that your rules do not unintentionally shut out AI crawlers you actually want to be fetched by, if visibility in AI search is a goal.

HTTPS is not up for negotiation

Every page belongs on HTTPS. The traffic is encrypted, the browser stops warning your visitor, and you do not pay the trust penalty an insecure marker costs. What the certificate does and how to get one is covered in what an SSL certificate is.

Prioritize, do not do everything

A crawl of a larger site returns hundreds of entries, and they are not worth the same. We work in this order: first what blocks indexing, then what hits pages with traffic or revenue, then what hits many pages at once, and cosmetics last.

Getting the mapping done properly is what a technical audit is for. And a pass over the technical base is included in every free SEO analysis.