Why Google Wanted Better Content, Not More Content

Most content strategies still answer a question nobody at the company actually asked. Twelve posts a month. Four pillar pages a quarter. A calendar built backwards from a keyword export, filled in by whoever has capacity. Ask why any single piece exists and the answer is usually that the slot was there.

That is the habit Google spent the last four years dismantling, and the reason the fix is not a better calendar.

What Google changed, and what it quietly stopped doing

Background rather than the argument, so: Google announced its “helpful content” system in 2022 as a separate, site-wide signal. It no longer exists in that form. Google’s own documentation now lists it as retired, absorbed into core ranking in March 2024. Alongside it came a spam policy aimed at mass-produced pages made to game rankings rather than help anyone — and that policy is explicit that it does not matter whether a human, a machine or both produced them.

Two consequences follow, and most coverage only gets the first. The obvious one: there is no discrete update to recover from, because the assessment now runs continuously. The less obvious one: Google stopped treating unhelpful content as a separate problem to be diagnosed and started treating it as part of what a site is. You cannot quarantine it in an underperforming folder any more.

Volume is not the diagnosis

Here is where most of the commentary goes wrong, and where we want to be careful, because the fair version of the counterargument lives in this distinction.

Nothing in Google’s documentation penalises publishing a lot. There is no recommended cadence, and its guidance explicitly denies having a preferred word count — the official answer to that question is a parenthetical “(No, we don’t.)” What the guidance actually flags is intent, in a form that is uncomfortably easy to recognise: are you producing lots of content on many different topics hoping some of it performs?

That is not a question about volume. It is a question about whether a decision was made. Ten thousand pages of product data, each answering a real question about a real thing you sell, is high volume with a reason. Forty blog posts a year on topics adjacent to your category, commissioned because a tool said the keywords had traffic, is low value at any quantity. Conflating the two lets people draw the wrong conclusion — that the answer is to publish less, when the answer is to publish only what the business has standing to say.

A dermatology practice we work with publishes a seasonal piece on summer skincare. It is competent and reasonable: drink water, wear sunscreen, use a lighter moisturiser, exfoliate, eat well. Eight headings, not one of which could not appear on fifty other sites, and no named author.

What makes it worth examining is not that it is bad. It is that the practice employs people who look at real skin all day, in one climate, across one season — and none of that is in the article. The slot got filled with what anybody could write, while the only thing nobody else could write went unwritten. No ranking report surfaces that, because the page is not failing. It is merely substitutable.

High-volume publishing is still right for some businesses

For some businesses, publishing constantly is not a symptom. It is the product.

News organisations, marketplaces, job boards, price-comparison sites, documentation teams and support functions all legitimately produce large quantities of pages, and would be worse if they did not. A support team answering three hundred distinct customer questions in three hundred articles is doing exactly what the guidance asks: writing for people who would find it useful arriving directly. Cadence there is a symptom of demand, not a substitute for it.

So the honest position is not “publish less”. It is that volume should be a consequence of having that much to say, rather than a target set in advance. The test is directional: does the quantity fall out of the work, or does the work get invented to fill the quantity? A support library grows because customers asked. A thought-leadership calendar grows because the calendar exists. Both can be large. Only one survives the question.

Our own policy, since we are arguing for one

It would be poor form to argue this without saying what we do. Our Insights section is written by the people on the project and published when there is something to say. No cadence, no listicles. That sentence is on the page, and it is a constraint rather than a slogan.

Behind it sits a set of tests we try to hold every piece to: a stated thesis, an identified gap in what is already published, sourced evidence, an answered counterargument, and a decision the reader can act on afterwards. We would rather describe that as the standard we aim at than as a gate nothing gets past — the point of writing it down is to make the misses visible. We also work to a no-fabrication rule: no invented client, metric, quote, result or project. That is why our articles carry fewer case studies than most agency writing and more reasoning.

The test that costs us most is the kill criterion. If the gap closes during research — if it turns out three people have already made the argument well — the piece gets dropped. Work already done, nothing published. That is the part that makes the rest mean anything, and the part hardest to keep to when a quarter is busy. A standard that never rejects anything is a calendar with extra steps.

Pruning is not auditing

The standard remedy on offer is a content audit that ends in deletion. Cut the thin pages, merge the duplicates, watch the site recover.

Google’s own guidance is more sceptical than the advice that invokes it. It asks whether you are adding or removing content mainly to make the site seem fresh, and answers directly: that will not help. Deleting pages is not a ranking tactic. If a page is genuinely serving someone, removing it costs you that.

The deeper problem is that a prune-led audit asks the wrong question. It asks which pages are underperforming. The useful question is which pages should never have been commissioned — and those are different sets. A page can rank adequately, hold traffic, and still have had no reason to exist.

An audit that asks whether the page deserved to exist

Take any twenty pages from the last two years of output. For each, answer these in order, and stop at the first “no”.

1. Who decided this should exist, and what did they know? If the answer is a keyword tool, that is the finding. If the answer is a named person with direct experience of the subject, continue.
2. What does it claim that we are in a position to claim? Point to the decision, the project, the data or the source behind the central assertion. If nothing supports it, the page is a summary of other people’s work.
3. Would the intended reader find this useful arriving directly, with no search involved? Google’s phrasing, and a good test regardless of Google.
4. What is the reader supposed to do differently? If there is no answer, the page has an audience but no purpose.
5. Is anything here unavailable elsewhere? Our colleagues set out the test for this in what AI cannot easily summarise; apply it here and stop.

Score the twenty. The number that matters is not how many failed — it is where they failed. Failures at question one are a commissioning problem and will recur. Failures at two are an evidence problem, usually fixable by involving the person who did the work. Failures at four and five mean the piece was not worth writing.

Then fix the commissioning process before the pages. Rewriting the output of a broken brief just produces a better-written page that still had no reason to exist. That is the work we do inside growth and performance engagements, and it starts before anything is written.