Home › Guides

SEO That Feels Right but Isn't: Nine Good Instincts That Backfire

Published 17 August 2026

None of the mistakes in this guide come from laziness. Each one starts as a sensible instinct — hospitality, thrift, diligence — aimed at a machine that doesn't reward it. Here are the nine we meet most, what each instinct gets right, where the mechanism breaks, and the move that works instead.

Some SEO mistakes present no warning. In fact they feel right, like good housekeeping. We've never met anybody who redirects a hundred dead pages to their homepage in a spirit of recklessness. They do it because it looks like the careful thing to do (and possibly, neat), and because ten years ago some version of it even worked.

So this is a different list from the scams — the snake oil has its own guide. These are your own better judgement misfiring, each with the mechanism behind it, because a correction that arrives with its reasons is the kind that survives the next confident blog post telling you otherwise. Where Google's own documentation settles a point, we say so.

1 · Redirecting every dead page to the homepage

The instinct is hospitality. An error page is a door slammed in a visitor's face, so you make sure nobody ever meets one: every address that no longer exists gets pointed at the front page, and the site feels sealed.

Google reads it differently. When a redirect lands somewhere that has nothing to do with the old address, Google classifies the result as a "soft 404" — an error wearing a redirect — and treats it as if the page were missing anyway. Nothing is gained in the rankings, the visitor who wanted last year's price list is blinking at your homepage, and you're maintaining a pile of redirect rules for the privilege.

The fix keeps the judgement and drops the blanket. A page that moved, or a dead address that other sites still link to or visitors still reach, deserves a redirect to its real successor — one line of config, and value you had already earned starts arriving somewhere live. A retired page with no successor is allowed to die: Google's documentation is plain that a 404 there is normal and costs your live pages nothing. Sorting one kind from the other is the real work, and our site-health guide walks through how to read a wall of 404s without alarm.

2 · Working the keyword in one more time

If you want to be found for a phrase, saying the phrase more often looks like simple clarity — you're helping the machine along. And in the industry's childhood it worked, which is why the advice refuses to die.

Today Google's spam policies name the move — keyword stuffing — and its systems have matched pages by meaning rather than by repetition for over a decade. A page can rank for "boiler repair" without those words in every heading, and a page that does force them into every heading reads like a till receipt, which costs you the customer the ranking was meant to bring.

Put the phrase where it earns its keep — the page title, the opening, perhaps one heading — and write the rest for the person who typed it.

3 · Polishing meta descriptions to climb

Every audit tool flags them — missing, too short, too long — and a flag implies a cost, so afternoons get spent rewriting descriptions in the belief that rankings follow. The tools aren't wrong about the gaps; what's inflated is the price of a gap.

Google has said for many years that the meta description is not a ranking input, and it rewrites the ones you provide more often than not. The real reason to write them well sits in plain view: the description is often the first thing a customer sees from you, right in the Google listings, before your site has shown them anything at all. That makes it a good, obvious piece of text for communicating an idea and converting a visitor — worth writing with care on the pages that already show up for things, and not worth an ounce of guilt anywhere else.

4 · A page for every phrase

More pages, more chances — the raffle-ticket theory of content. It produces sites with a near-identical page for every town in the county and every synonym of the service, and the thoroughness would be admirable if the arithmetic weren't backwards.

Google's guidance calls these doorway pages, and its systems are built to collapse them. The copies also compete with each other, so the strength one substantial page would have gathered is split ten thin ways and none of them ranks. A single page that treats the subject properly beats the fleet.

Separate pages are right when there is separate substance — a service you deliver differently in each place, with something true to say about each. When the only word that changes is the town, it's one page.

5 · Publishing every week because Google rewards fresh

The content calendar carries a moral force: activity should be rewarded, and a moving blog looks like a living business. Google, though, holds no reward for frequency itself. Freshness matters where the right answer changes — news, prices, schedules — and hardly at all where it doesn't; a five-year-old page that still answers its question best still wins.

The same hours spent returning to pages that already earn — correcting the facts, widening the answer, mending what's dated — reliably do more than a new post written to fill the slot. Publish when you have something to say. Readers can tell that cadence from the other kind, and so, in effect, can the index.

6 · Hitting the word count

Word-count targets feel like rigour: the top results are long, the tool suggests 1,500 words, so the draft gets padded until the bar turns green. Google has stated flatly that word count is not a ranking factor. Long pages rank when their subject needed the length — the length was a symptom of a full answer, and copying the symptom gets you nothing.

Padding does carry a price, though. The person who arrived for a straight answer scrolls, gives up and returns to the results, and that is the one behaviour you least want attached to your page. Make the page the length of the answer; three hundred words that settle a question beat fifteen hundred that circle it more often than the tools let on.

7 · Buying links to lift your authority score

The logic seems sound: links are votes, the authority number should rise with more of them, and fifty pounds buys a hundred. Two things break it. The score is a vendor's estimate — Moz, Ahrefs and the rest each mint their own — and Google keeps no such number, so a package that inflates the estimate hasn't touched the thing being estimated. And the links in those packages are the directory-and-bot chaff that every domain collects on its own for free. We read backlink profiles daily and the pattern repeats: dozens of sites "link" to a domain, a handful carry any weight, and the handful were all earned.

If you want to see where a domain actually stands before deciding it needs help, our free domain authority checker will read one for you without an account. Paid links are also, in plain terms, against Google's spam policies — the cheap kind get ignored, and the industrial kind invite a manual penalty. The slower advice stands because it's the one that works: a few links from places whose opinion counts, earned by giving them something worth pointing at. Our backlinks guide covers what the scores measure and how the earning goes.

8 · Rewriting everything for fear of the duplicate-content penalty

This one is caution rather than tidiness. Somewhere along the line "duplicate content" entered the folklore as a punishable offence, and owners rewrite product variants, printer pages and their own quoted paragraphs to stay ahead of a penalty they have never once seen levied.

For an ordinary business site there is no such penalty. Google meets duplication constantly — it is most of the web — and its response is to group the copies and choose one to show. The real, smaller risk is that it chooses a version you'd rather it hadn't, and that is solved by declaring a canonical address on each page: a line of markup, no rewriting required. What does get punished is scraping other people's content at scale, which is a different activity from having your terms and conditions on two URLs.

9 · Driving the audit score to 100

A score invites management, and watching yours climb from 61 toward 100 feels like progress you can report. We build one of these scores ourselves, so we say this with some self-knowledge: a health score is a percentage of whatever that particular tool checked, and the first sites ours examined got perfect marks because our early checks were narrow. The score's ceiling was our eyesight's.

The useful reading sits under the number: which findings block machines outright, which leak value you'd already earned, and which are polish that can wait. Fix in that order, then spend what remains of the effort on the pages themselves — a tidy site does not outrank a useful one, and usefulness is the part no score measures.

Each of the nine manages something visible — the count, the score, the calendar — while Google prices what the visitor got.

That's the pattern to carry out of this guide, and it's also the test for the next piece of advice that reaches you: does the move improve what a reader gets from the page, or does it decorate a number near the page? The instincts themselves are sound — pointed at the reader instead of the crawler, they count double.