Depending on who’s counting, Google uses somewhere north of two hundred ranking factors. About six of them will decide whether your site works.

The published lists are the problem. They’re flat, so keyword placement in an H1 sits beside whether anyone in your industry has ever linked to you, as though those are comparable. They’re also mostly reverse-engineered from correlations, which is how the SEO industry ended up spending a decade on things that were symptoms rather than causes.

What follows is the order we work in, and what we’ve seen it do. It isn’t a study, and I’ll say so properly at the end.

Whether a page exists that deserves the query

Most of the time, “we can’t rank for that” means no page on the site is about that.

The homepage isn’t about it. A blog post that mentions it in the fourth paragraph isn’t about it. A service page covering nine things isn’t about any of them. This is the least interesting finding in SEO and comfortably the most common one, and it’s usually worth more than everything else on this list combined.

Pace is the clean version of this. A brand-new domain, a Domain Rating of 12, and 116 Google impressions in the whole of February. What we did was build pages where nothing existed: pacing guides for each platform, comparison pages for the tools buyers already knew, free calculators, and one piece of original research. Nothing exotic. The research page now pulls 5,757 impressions in 28 days at an average position of 6.8, the CPA calculator sits at position 6 for its head term, and the alternatives pages hold page one against companies with a decade of content and a funding round behind them.

The order matters here too. Build the page, then worry about the page’s authority. Doing it the other way round is how agencies end up buying links to a homepage nobody can navigate.

Internal links, the lever nobody pulls

Free, entirely under your control, routinely ignored.

If your most commercially important page is four clicks from the homepage and linked from nothing, you’ve told Google what you think it’s worth. And the anchor text on your own internal links is the only description of your page that you fully control, which makes it strange that most sites’ internal links say “click here” or nothing at all.

Every site we take on gets this audited early, because it’s the shortest distance between work and movement. It’s also the fix that survives algorithm updates, since it isn’t a trick.

The technical floor

Crawling and indexation are not a growth lever. They’re a floor you fall through.

Broken canonicals, a noindex left on a template after a staging push, content that only renders after a slow JavaScript call, redirect chains, a robots.txt rule blocking a folder somebody added two years ago. None of this makes a good page rank higher. All of it can stop a good page ranking at all, which is a different and much worse problem.

We run the same technical checklist on every new site in the same order, because the failures cluster and finding them is a process rather than an instinct.

Links from sites with something to lose

Still real, still hard, still the reason authority compounds.

The distinction that matters is whether the linking site would be embarrassed to be caught selling links. Editorial mentions in trade press, references from suppliers and partners, coverage of something you actually did. The rest is decoration at best, and the market for it is enormous precisely because it’s easy to sell.

Being a known entity

Search and AI answers both reward brands that are described consistently everywhere they appear. Same name, same category, same claims, across your site, directories, trade press and anywhere else a machine reads.

Branded demand is part of this and it’s underrated as a ranking asset. Pace now ranks first for its own name with a 68% click-through rate. In February, nobody was typing it. Creating the demand for your own brand name is a ranking factor by a longer route, and it’s the one competitors can’t copy.

What looks busy

Keyword density, and any tool that scores a page out of 100. The score is the tool’s opinion, not Google’s.

Domain Rating or Domain Authority as a goal. They’re third-party estimates of link strength. Nobody at Google can see yours, and chasing the number is not the same as earning the links behind it.

Publishing volume with no plan behind it. Twelve thin posts a month will not out-rank one page that answers the question properly, and most of what people believe about this is a decade out of date.

Rewriting title tags every fortnight. Make a change, then give it six weeks. Changing three things at once and reading the result a week later isn’t testing, it’s noise generation.

Reflexive disavowing. Google is reasonably good at ignoring rubbish links. Disavowing a domain that was actually helping you is a real risk and an unforced one.

Core Web Vitals pursued as a ranking project. Fix a genuinely slow site, because slow sites lose people before Google is even involved. Spending a quarter moving your largest contentful paint from 2.6 seconds to 2.4 and expecting positions to follow is a budget decision I’d argue against.

Schema markup applied everywhere for its own sake. It makes you eligible for certain features. It doesn’t move position.

The evidence, and its limits

AGL’s programme with us produced a 58% increase in average search position, a 22% increase in clicks, a 40% increase in keywords sitting in the top three, and 14.5k features in Google’s AI Overviews. Pace went from 116 impressions in February to 43,641 in July, a cumulative 118,507 across Google and Bing, with no link spend at all.

Two accounts is not a study. Both are ours, both had someone competent on them continuously, and neither ran a control, because clients don’t fund controls. What I’d claim is narrower than the numbers look: in the work we can see, the sites that moved were the ones where somebody built the missing pages and connected them properly. The sites that stalled were usually busy.

What we’d do first on your site

Export your Search Console queries and find everything sitting between positions five and twenty. Look at which URL is ranking for each. It’s frequently the wrong one, and pointing the right page at the right query is a week’s work with a fortnight’s payoff.

List your five most commercial pages and count the internal links pointing at each. If the answer is under five, you’ve found your afternoon.

Crawl the site once, properly, and fix whatever’s broken in the index.

Then, and only then, think about links. If you want the longer version of the on-page half, how to improve your SEO covers it, and if you’d rather someone else did the work, that’s what our SEO team does all week.

What actually moves rankings, and what just looks busy