Advice has an expiry date, and it is shorter than you think
Google confirmed six updates between January and the end of August 2026: a Discover core update in February, a spam update and a core update in March, another core update in May, and spam updates in June and August. The year before that had three core updates and one spam update. Google also says its automated spam systems run continuously between the announced ones.
Advice ages in two different ways, and they need separating.
It becomes risky. Some tactics that were common recommendations in 2021 to 2023 now sit inside Google's written spam policies. Publishing a page for every long-tail wording variation can meet the definition of scaled content abuse or doorway abuse. Buying an expired domain for its history and pointing it at an unrelated commercial niche is expired-domain abuse. Renting a section of a trusted publisher's site to rank commercial content is site reputation abuse. Building links through payment, exchange, optimised guest-post anchors or template footers is link spam.
Worth being precise here, because the internet gets this one wrong constantly: Google's policy is not "AI content is banned". It is against producing many pages primarily to rank rather than to help anyone, regardless of whether a human, a scraper or a model wrote them. The tool is not the problem. The purpose is.
It stops paying. This is the subtler kind. The tactic still works mechanically, it just no longer produces the result it used to. Ahrefs compared Search Console data across 300,000 keywords in February 2026 and found position-one click-through rate on queries with an AI Overview was 1.6%, against a projected 3.7% without one, roughly 58% lower (Ahrefs).
Treat that number with care rather than repeating it as a law. Semrush ran a different design, tracking the same keywords before and after an AI Overview appeared, and found zero-click rate actually fell slightly, from 33.75% to 31.53%. Google's own line is that clicks from AI features tend to be higher quality. The honest conclusion is narrower than the headline: ranking and clicks have come apart, and the relationship now varies by query type rather than holding steady across your whole site.
So when a 2024 article tells you to build a large library of top-of-funnel question pages, the advice is not fraudulent. The economics behind it changed, and the article cannot tell you that.
The advice was written for a different kind of website
This is the one that costs people the most time. Two sites in the same country, both technically fine, both with a competent person working on them, need genuinely different work, because the job organic search is doing for each business is different.
| Site type | What actually moves the needle | What commonly misfires |
|---|---|---|
| E-commerce | Category architecture, product data, selective indexing of filters that have real demand and real stock | Indexing every filter and sort combination, or blocking all of them without checking which ones people search |
| Lead generation / services | A small number of strong service pages, proof of competence, reviews, enquiry tracking, local relevance where it is genuine | Hundreds of city pages for one office, or a blog built before the service pages are any good |
| SaaS | Use case, integration, alternatives and comparison pages, plus content that maps to a buyer's actual objections | Optimising a short feature page for a broad category term where the results page is full of comparison articles |
| Local business | Accurate profile data, opening hours, service area clarity, reviews, fast paths to call or book | Treating it as national content marketing and writing explainers instead of fixing the entity data |
| Publisher | Original reporting, testing, consolidation, internal linking, direct audience | Publishing more of the same thing, which splits topical signals across overlapping pages |
Faceted navigation is the clearest example of a tactic that is genuinely excellent in one context and actively harmful in another. On a retailer with thousands of products, controlled filter pages are a legitimate discovery system. On a fifteen-page plumbing site, the same setup manufactures duplicate URLs and crawl waste for demand that does not exist. Google's Gary Illyes said faceted navigation accounted for around 50% of the crawling problems reported to Google in its 2025 year-end review, with another 25% coming from action parameters.
The same applies to publishing frequency. "Publish three posts a week" is a quota, not a strategy. It ignores how much demand your niche actually holds, whether you have anything first-hand to add, and whether you already have four pages competing for the same thing. A newsroom can justify that pace because new facts keep arriving. A specialist consultancy usually cannot, and the cost is not neutral: overlapping pages compete with each other and dilute the signal that you are the specialist.
Your site does not start where the case study started
"We grew organic traffic 500%" is a result. It is rarely a method you can copy, because the preconditions that made it work are the part left out.
Some numbers that put a floor under this. Ahrefs tracked a million URLs and found only 1.74% of newly published pages reached the top ten within a year. Around 72.9% of pages currently in the top ten were more than three years old, and the average number-one page was about five years old (Ahrefs). Backlinko's April 2025 analysis of 11.8 million results found the average number-one result had 3.8 times more backlinks than positions two to ten, while keyword-optimised titles and keyword-rich H1s showed essentially no correlation with moving higher within page one (Backlinko).
Read those two findings together and a lot of generic on-page advice makes more sense. Fixing your title tag is worth doing. It was almost certainly not the thing holding you back.
One clarification that matters, because it drives bad decisions. Google does not use Moz Domain Authority or Ahrefs Domain Rating. Those are third-party estimates built by tool vendors, and Moz says so plainly on its own page about the metric. Ahrefs puts the correlation between DR and rankings at 0.131. Established sites do tend to have the underlying assets those scores partly reflect (relevant links, recognised topical coverage, internal linking, an audience that already searches for them by name), but chasing the score itself is chasing a proxy.
Before you copy a case study, ask what it does not tell you: was growth from 100 visits or 100,000, how many people and how much budget were involved, over what timeframe, did every page succeed or did a handful carry the result, and did any of it turn into revenue. The more of that is missing, the less safely you can copy it.
Why the small shop cannot outrank the category giant
This one comes up constantly, and the frustration is completely reasonable. You stock six products. A national retailer stocks four hundred. You optimise your collection page better than theirs, and you still lose.
The reason is structural, not a quality failure on your side. A broad commercial search implies choice. Someone typing "trail running shoes" has not decided anything yet, and the giant can offer brands, price bands, sizes, stock, reviews, comparisons and subcategories immediately. Your perfectly optimised six-product page is a worse answer to that particular question, no matter how good the page is.
The answer is not to fake catalogue depth with filter pages and thin variants. It is to compete on a narrower territory where your six products actually are the better answer, then cover that territory properly. Ahrefs' example is a useful one: specialist e-bike retailer Bicycle Motor Works, with a DR of 15, outranks Amazon at DR 96 for competitive e-bike terms, because the coverage is deeper and the relevance is tighter (Ahrefs).
There is also encouraging data here. Semrush and Kevin Indig analysed 1,094 topical categories across ChatGPT in July 2026, covering more than 50,000 brands. Only 15.2% of those categories had a clear owner. Another 53.7% were unsettled, with no brand appearing consistently. Topic winners had higher organic traffic in only 48.4% of head-to-head pairs. Most niches are not spoken for, and total site strength is not what decides who owns them.
Practically, that means picking something like "folding e-bikes for commuters" or "lightweight two-person backpacking tents" rather than "e-bikes" or "tents", then building the product pages, comparisons, buying help and ownership content around it until you are visibly the specialist. It is a smaller market. It is a market you can actually win. My piece on increasing traffic for small websites goes further into how that plays out day to day.
Two SEO managers, two different plans, similar results
Here is something rarely said out loud, because it sounds like an excuse: two competent SEO managers can look at the same site, choose substantially different priorities, sequence the work differently, and arrive at broadly similar results after 12 to 24 months.
Take a hundred-page B2B services site. One manager starts with technical eligibility and the commercial pages: fixing canonicals and indexing, rebuilding the service page templates, setting up proper enquiry tracking, then adding case studies and a small content cluster later. The other starts with the content and the expertise: mapping the customer problems, building an expert-led hub, capturing what the team actually knows, then rebuilding the service pages around what that reveals, fixing technical issues as they surface.
Different weeks, different reports, different-looking work. If neither made a sequencing error and both eventually removed the same real constraints, the end state can be very close: pages that match demand, proof a buyer believes, a site Google can crawl and understand, and enquiries you can trace.
What actually separates good from bad here is not tactic choice. It is three things.
Order. Some work only pays once something else is in place. Building links to a page that does not match what the results page rewards wastes the links. Rolling out a page template across two hundred URLs before proving it on ten creates two hundred problems. Adding structured data to a page that has nothing worth marking up changes nothing.
Pace. There is no confirmed Google threshold for publishing speed or link velocity, and anyone quoting one is repeating folklore. Retailers add thousands of legitimate product pages. Publishers cover live events at volume. The real limit is different: release work as fast as you can genuinely check, maintain and measure it. Beyond that point you are just generating things nobody is looking after.
What was declined. Which brings us to the part most people never account for.
What you avoid matters more than what you add
Good SEO work adds probability. A clearer page might help, might do nothing, might help on a subset of queries. Harm is not symmetrical. It creates a drag that sits across the site for months.
Google says that after a spam update, a site that fixes its problems may need to be seen complying for a period of months before automated systems reflect it. Link spam carries a sharper edge: when Google removes the effect of manipulative links, correcting things afterwards does not restore what those links were supplying, because that benefit is gone rather than paused.
Put that next to the upside of a shortcut and the maths is bleak. A useful page might lift one query set. A spam demotion can suppress the site for months, and some of it never comes back.
So restraint is a strategy, not passivity. In practice it looks like this:
- Not creating a page because it would duplicate one you already have.
- Not indexing a filter combination that has no stock and no demand behind it.
- Not chasing a head term where you have no credible route to being the better answer.
- Not buying links to satisfy a monthly report.
- Not publishing AI drafts without someone who knows the subject checking them.
- Not changing URLs, templates and internal links all at once during a core update rollout, when you will have no idea what caused what.
None of that shows up well in a status meeting. It is often the most valuable thing that happened that month. There is more on this in the mistakes that hurt organic traffic.
"High quality" is decided by someone who knows the topic
Quality is the most subjective word in this industry, and the disagreement is real. A skilled writer produces something clean, well structured, keyword-aware and complete-looking. A subject-matter expert reads the same piece and finds it thin. Both are being honest.
The difference is usually not length or polish. It is decision-useful specificity.
| Looks like good content | Is actually good content |
|---|---|
| Covers the same subheadings as the top-ranking pages | Identifies the decision the reader is actually trying to make |
| Well written and keyword-aware | Accurate, current, and caveated where the answer genuinely varies |
| Generic examples | First-hand evidence, original testing, real client or product detail |
| "Best practice recommends..." | What applies, what is disputed, and the conditions under which it fails |
| Has an author name on it | Gives the reader a reason to trust that particular person on that particular subject |
| Long enough to seem thorough | As long as the problem needs, with the trade-offs left in |
Google's September 2025 update to its Search Quality Rater Guidelines pushed in the same direction: pages showing little effort, originality or added value should receive the lowest rating, and that explicitly includes copied, paraphrased or automated content. Google is equally clear that using AI does not by itself decide quality. Effort, originality and purpose do.
This is also why an experienced SEO and a good writer can disagree about the same draft. The writer is assessing the text. The SEO is assessing whether the page has any chance against what is already ranking, and whether the site is even the sort of source Google would want to surface for that query.
Why I will not recommend one SEO course
I get asked this fairly often. Someone wants a course to put their team through so they can handle their own site. I understand the appeal, and I still cannot name one, for four reasons.
A course teaches a generic site. It has to. It will cover e-commerce, local, lead generation and publishing in the same syllabus, and you need the part that applies to yours, taught with your constraints. Nobody can build that in advance for a stranger.
A course captures a moment. It reflects whatever was working when it was recorded. Some of that ages gracefully. Some of it becomes exactly the advice that gets a site into trouble, and the video will not tell you which is which.
Someone showing what they did on their own site is a report, not a recipe. It happened on their domain, with their authority, their niche and their timing. Doing the same things on a new site and seeing nothing happen is the normal outcome, not a sign you executed badly.
Third-party tools do not show you the real situation. Search volumes, CPC figures, difficulty scores and authority metrics are estimates built from samples and models. They are useful for comparison and direction. They are not measurements. Reading the numbers is easy. Knowing which ones to ignore, and what a figure means for a site in your position, is the actual skill, and it is the part courses struggle to teach.
What I would send a team to learn instead: how to read their own Search Console properly, how to look at a live results page and work out what Google is currently rewarding for that query, how to run a small test and judge the outcome, and how to tell a real constraint from a tidy audit flag. That transfers. A tactic list does not. If you want the wider picture of what the role actually involves, I wrote about what an SEO manager does.
A note about the older articles on this site
Since this whole piece is about advice ageing, it would be dishonest not to apply it here.
If I go back and read things I published on this site in 2024, I see plenty I would not recommend today. Some of it is framing that made sense then and does not now. Some of it is thinner than what I would write for that topic in 2026, or it answers the search intent as it looked at the time rather than as it looks today.
I am mostly not updating them, and I know exactly what a strict reading of current advice says about that.
The reason is straightforward. This site was never built to rank for a target keyword list. Most of those articles were written for specific people. Somebody asked me a question, several companies had the same question, and it was more useful to write it once properly and send a link than to write the same explanation four times. They did their job on the day they were published.
I mention it because it is a real example of the point. Whether that content is a problem depends entirely on what the site is for. If seoman.uk were a commercial site competing for those terms, leaving it alone would be a mistake worth fixing. It is not, so it is not the priority. Same content, same signals, different verdict, because the context is different. That is the entire argument of this article in one paragraph.
Five questions to run any SEO advice through
Before you act on a recommendation from an article, a course, an agency pitch or an AI tool, work through these. If you cannot answer three of them, it is not a plan yet, it is an idea.
- Business fit. Does this connect to how the business actually makes money: a sale, a qualified enquiry, a booking, a trial, a subscription? If the only stated outcome is "more traffic", stop there.
- Intent fit. Look at the live results page for the query. What is Google currently rewarding: product pages, a map pack, comparison articles, a forum thread, an AI answer? If the page type you are planning is not in there, you are not competing, you are hoping.
- Site fit. Do you have what the tactic needs: the catalogue, the locations, the data, the expertise, the developer time? A tactic built on assets you do not have is not a shortcut, it is a way of manufacturing thin pages.
- Competitive fit. Can you state a concrete reason your page could be better than what currently ranks? First-hand testing, real product depth, local presence, original data, genuine specialism. "Better keyword placement" is not a reason.
- Measurement fit. What will you check, when will you check it, and what will you do if it has not worked? Ahrefs' data suggests that if a page has not reached the top ten in around six months, it usually will not without something changing. Decide in advance what "changing" means.
The short version: stop asking "is this best practice?" and start asking "what constraint does this remove on my site, for which searches, and how will I know it worked?" Most SEO advice survives that question in a much narrower form than it was written in — that narrower version is the part that was ever going to work for you.