What is the single most common SEO mistake on a small site?

A page that cannot be indexed, usually a noindex tag or a robots.txt rule left over from staging. It costs nothing to check and it stops everything else from working.

How do I know if two of my pages are competing?

Open the Performance report, filter to the query, then switch to the Pages tab. If two of your URLs appear for it and both sit mid-page, you have split one topic across two pages.

Are keyword density and meta keywords still mistakes?

They are not mistakes to avoid so much as work not to do. Google has said keyword density is not a thing it measures and the meta keywords tag is ignored. Write for the reader and use the words they search.

Should I delete old pages that get no traffic?

Only after checking links and past clicks. A page with links or a purpose is refreshed or merged. A page with no clicks, no links and no purpose for a year can go, with a 410 or a redirect to the nearest equivalent.

How often should I run through this list?

The first group when you deploy something structural, the rest once a quarter. Most of the items stay fixed once you fix them.

The SEO mistakes small product sites actually make

Most small sites are not held back by anything clever. They are blocked from the index, split across two pages that fight, or writing for searches nobody makes. Here are twenty mistakes, with the check for each and the fix.

By Théophile Louvart, founder of Porteur · Updated 15 September 2026 · Markdown

How to use this list

Work down it in order. The first group stops pages existing in Google at all, so nothing below it matters until those are clear.

Each row gives the check and the fix. The checks use Search Console, your browser and the page source, because those are free and they are what Google actually sees.

Group one: blocked from the index

Mistake	The check	The fix
A noindex tag shipped from staging	View source on the home page and a deep page, search for meta name="robots"	Remove it from the production template, then inspect the URL again
Disallow: / left in robots.txt	Open yourproduct.com/robots.txt	Remove the rule. A disallow stops the crawl, so Google cannot even read a noindex behind it
An X-Robots-Tag header nobody knows about	Browser dev tools, Network tab, read the response headers	Remove it at the server or the edge for public paths
A canonical pointing at another host	URL Inspection: compare the user-declared canonical with the Google-selected one	Point the canonical at the page itself, one per page
The site is new and has no links at all	Search site:yourproduct.com and check the Links report	Add one real link from a profile you control, then submit the sitemap

Group two: pages that fight each other

Mistake	The check	The fix
One page per keyword variant	Performance report, filter a query, look at the Pages tab	Merge them into one page and redirect the weaker URL
Two pages trading places on one search	Compare the pages tab for a query week to week	Decide which page owns the search, then rewrite the other for a different question
The same content on /feature and /features/feature	Crawl the site, or search site:yourproduct.com for the duplicate title	Pick one URL, redirect the other, fix the internal links
Tag or category archives with one item each	Look at what is indexed under /tags/ or /category/	Noindex the thin archives and keep them for people, or delete them

This group is worth an hour more than any technical fix. Two half-pages on one subject rank worse than one page that answers it.

Group three: titles and snippets that waste the impression

Mistake	The check	The fix
The title says the brand twice	Read your titles in the Performance report or a crawl	One brand mention, at the end, or none on deep pages
Every page shares one description	Crawl and sort by description	One description per page, 120 to 155 characters, a promise not a tease
Titles longer than the result shows	Aim for about 60 characters and check in a snippet preview	Put the search near the front so it survives the cut
The title and the h1 say different things	Compare them on ten pages	Say the same thing twice, in the page's own words

The pattern to look for is a page with many impressions and almost no clicks. Google shows it, people read the title, and they choose something else. That is a title problem, not a ranking problem.

Group four: content nobody searches for

Mistake	The check	The fix
A blog of company news	Filter the Performance report to /blog/ and read the impressions	Write pages that answer searches, and keep the news for the changelog
Pages written from ideas, not from results pages	Search the target query and compare what ranks with what you wrote	Read the results page first, then write for the intent it shows
A page family generated from a template	Ask whether each page holds data, a method or an answer of its own	Cut the ones that do not. Google's spam policies call the rest scaled content abuse
No page for the searches that convert	Look for 'alternatives', 'vs' and 'best' queries in your category	Write the comparison and alternatives pages before the tenth blog post

Group five: work that stops

Mistake	The check	The fix
Pages published and never touched again	Sort pages by clicks lost against the previous period	Refresh the facts and the date on the pages that are falling
A launch nobody heard about	Search your brand in quotes and count the mentions	List the product where your buyers already read, one place a week
Nothing measured, so nothing learned	Open Search Console. If you have not this month, that is the answer	Ten minutes a week: what fell, what rose, one page fixed

Two of these three are not technical at all, and they are the ones that decide whether a small site grows.

The mistakes that are not mistakes

Half of what founders worry about does not matter, and worrying about it costs the hour that would have fixed something real.

Keyword density and the meta keywords tag: Google has said neither counts.
A single 'domain authority' score: that is a tool's model, not a Google metric.
Word count as a target: Google has said there is no number that ranks.
Social shares as a direct ranking signal: not confirmed, and not a lever you can pull.
A Lighthouse score of 100: the field data at the Good thresholds is what page experience reads.
Every 404 on the site: a 404 for a page that should not exist is correct.

Questions

Check my site, free

Paste your URL and the free check reads your site the way a crawler does, in about thirty seconds, and names the three mistakes costing you most.

Free check, no card
Read-only, your own accounts
Readable by your agent