Home › Guides

Guides

Short, specific write-ups on the parts of technical SEO where the common advice is wrong or out of date.

All guides

Meta Description Length: What Gets Truncated

The number everyone quotes is 155 characters. The real rule is pixel width, and knowing the difference saves you from rewriting descriptions that were fine.

robots.txt Mistakes That Silently Deindex Pages

Five robots.txt errors that silently remove pages from Google — a trapped noindex, blocked CSS and JS, wildcards that over-match — plus the five-minute check.

Heading Structure: What H1–H6 Order Does

Headings are document structure, not a ranking trick. What H1-H6 order genuinely affects, what Google says it does not, and how to check a page in a minute.

JSON-LD @graph: Why One Block Beats Five

Most sites emit one JSON-LD block per schema type and never connect them. A single @graph with @id references fixes the duplication. One pattern, every type.

A Pre-Publish SEO Checklist for Any New Page

Nine checks that take under ten minutes and catch the mistakes that are invisible after publishing. Ordered so that the cheap ones come first.

What 71 Sites Actually Put in robots.txt

We fetched the robots.txt of 78 high-traffic sites. A quarter declare no sitemap at all, and five still ship a directive Google ignores. The fix takes one line.

What 61 Homepages Actually Ship in JSON-LD

We parsed the JSON-LD on 61 homepages. Half ship none in their HTML, and 14 still carry markup for a feature Google retired in 2024. The raw counts are inside.

What 62 Homepages Actually Ship in Open Graph

We read the Open Graph tags on 62 homepages. 90% have some, only 23% declare image dimensions, and two sites ship an og:image tag that carries nothing.

What 70 robots.txt Files Say About AI Crawlers

We read 70 robots.txt files and checked 22 AI crawler tokens in each. 63% never mention one, and the most famous crawler is blocked least often.

What 60 Sites Actually Do With hreflang

We parsed hreflang on 60 high-traffic homepages. Less than half declare it, and a third skip x-default — the tag Google recommends but never requires.

What 55 Homepages Actually Do With Canonical

We fetched the homepage of 78 well-known domains and read 55. Four in five declare a canonical; one in five ships a homepage with none at all.

llms.txt Adoption: 22 of 78 Big Sites

Half of 78 well-known sites return a clean 404 for /llms.txt. Of the 22 that serve one, 11 hide behind a redirect and 4 break their own heading.

What 59 Homepages Put in Their SERP Snippet

We fetched 78 well-known homepages and read 59. Every one ships a title; the median is 35 characters, and a fifth of the descriptions run long.

Whatsmyserp Alternatives: Pick by Job, Not Price

Two different jobs hide behind the phrase 'Whatsmyserp alternative', and most lists ignore the difference. Here is what each tool is actually for.

Why Google Won't Index Your Page: A Checklist

Indexing problems get diagnosed in the wrong order, so the fixes don't work. Here is the order that separates a code fault from a trust gate.

How AI Search Engines Choose What to Cite

AI search engines cite pages they can find, understand, and trust. Here is what makes a page citable, and how this site is built to be one. No tricks involved.

How Long Should a URL Slug Be?

Short, lowercase, and carrying the topic words: a good slug reads like a label, not a database key. Here is the working rule for writing one.

nofollow vs sponsored vs ugc: Which to Use

Paid links get sponsored, user-written links get ugc, and links you do not trust get nofollow. All three are hints now, and you can combine them.

How Many Internal Links Should a Page Have?

There is no number to hit. Link where a reader would want the next step, keep important pages within three clicks of the homepage, and fix orphans.

Image Alt Text: What to Write and How Long

Alt text describes what the image does for the reader, in one sentence. Decorative images get an empty alt, and keywords stay out of it. Edge cases included.

Sitemaps in the Wild: 55 of 78 Sites

We pulled 251 URLs out of 55 real sitemaps and requested each one: 1 in 8 answers with a redirect, and 25 of the 55 sitemaps were not at /sitemap.xml.

Redirect Chains in the Wild: 78 Domains

We walked http://bare-domain for 78 well-known sites: 37 take two hops to reach their own homepage, and 28 of those could have done it in one.

Heading Structure in the Wild: 53 Homepages

We parsed 1,696 headings from 53 well-known homepages: 16 have no single visible H1, and 22 open with something other than an H1. The fixes are usually trivial.

Favicons in the Wild: 61 Homepages

We fetched every icon declared by 61 well-known homepages: 248 tags, 6 sites with nothing larger than 32x32, and a .ico that is really a WebP.

Indexing Directives in the Wild: 59 Homepages

We read the robots meta tags and response headers of 59 well-known homepages: 25 declare anything, 12 of those change nothing, and one noindexes itself.

Link rel in the Wild: 54 Homepages

We read all 10,121 links on 54 well-known homepages: 5 sites put nofollow on an outbound link, 1 uses sponsored, and nobody uses ugc. Every link was counted.

Resource Hints in the Wild: 63 Homepages

811 resource hints on 63 homepages: preload and preconnect do the work, prerender is gone, and 81% of preconnects warm origins the HTML never mentions.

Images in the Wild: 55 Homepages

2,418 images on 55 homepages: 39.6% declare themselves decorative, 5.8% have no alt at all, and the first image is lazy-loaded on 13 sites. Full method inside.

HTML lang, Charset & Viewport: 60 Sites

We read the document metadata of 60 homepages: one declares no language, three declare their charset too late to count, four block pinch zoom outright.

Security Headers in the Wild: 60 Sites

60 homepages: 92% send HSTS but 7 ask for less than a year and one asks for zero; 62% send a CSP and 73% of those allow inline script. Measured, not assumed.

Forms and Labels in the Wild: 60 Sites

60 homepages, 146 form controls, but only 74 a screen reader actually reaches — and 2 of those have no name at all. Every failure mode is named inside.

HTML Weight and Render-Blocking Scripts

60 homepages weigh 51 MB. The median is 414 KB, 60% still block the parser on a script, and one site ships a single 12.8 MB JSON file inside its HTML.

How to Write a Title Tag That Gets Clicked

Front-load the topic, stay inside the pixel budget, and write the words a searcher would recognise as their own question answered.

UTM Parameters: The Complete Guide

Five parameters, three that matter, and one naming mistake that splits every campaign in your reports.

How to Create a Favicon

A favicon is the smallest brand asset you own and the one users see most. The sizes you need, Google's current rules, and how to ship one in an afternoon.

What these guides have in common

Each one covers something you can check and fix in an afternoon, and each one leads with the part of the conventional advice that is wrong or out of date. That constraint is deliberate. There is no shortage of material explaining what a title tag is; there is very little that tells you the 155-character rule is a pixel measurement in disguise, or that disallowing a page prevents its noindex from ever being read.

The other shared property is that none of them require a tool subscription to act on. If a guide's conclusion is "buy this product", it does not belong here.

Where to start

If a page has dropped out of search results, read the robots.txt mistakes guide first — it covers the failure mode that produces no error message anywhere, which is why it takes people weeks to find.

If you are publishing new pages and want them to earn clicks, read the meta description guide, then check each page with the SERP Preview.

If you are cleaning up an older site, the heading structure guide is the one that usually uncovers a template problem affecting every page at once.

What is not covered here

There is no guide on link building, keyword research tooling, or content calendars. Those topics are saturated with material that is either obvious or wrong, and this site is deliberately about things you can check and fix in an afternoon.

There is also nothing here about AI-generated content at scale. Not because it is taboo, but because the useful advice depends entirely on what you are doing with it, and a general guide would be worthless to everyone.

How to read a technical SEO claim

Most bad advice in this field has the same shape: a rule stated as absolute, with no mention of what it was measured on. "Titles should be 60 characters" is a measurement of one font at one size on one device. "Keyword density should be 2%" was never measured on anything. When a rule has no conditions attached, that is the signal to check it rather than adopt it.

A useful test is to ask what would change the answer. If nothing would — if the rule holds regardless of your platform, audience or content type — it is probably a policy or a legal requirement, not an optimisation. Those are the only absolute rules in this field, and there are very few of them.

The second useful test is recency. Google's systems change continuously, and a claim that was verifiable in 2019 may now be simply false. Where a guide here makes a claim about current behaviour, it says when it was checked.

Corrections

If one of these guides is wrong about something, that is worth an email — see the contact page. Technical SEO advice ages badly, and several widely repeated claims were accurate five years ago and are not now. Being told which ones I have repeated is genuinely useful.