One Rival Had 2,486 Pages, Another Had 2

Someone has quoted you for twenty pages a month and you are trying to work out whether that is ambition or waste. The quote is persuasive because it is countable. You can see twenty. You cannot see whether twenty works.

On 3 October 2026 we measured ten Singapore companies competing on the same four buyer searches in one industry and read each one's sitemap. The smallest site listed 2 URLs. The largest listed 2,486. That is a 1,243-fold spread between businesses selling the same service to the same city, and they were sitting on the same results page.

Page count did not predict who answered the buyer's question. At the big end, 1 of the 3 largest sites put the price and its minimum together where a buyer could see both. At the small end, 1 of the 3 smallest did the same. Same proportion, 800 times the page count.

Where the page-count pitch comes from

It comes from a true statement used dishonestly. More pages can mean more search coverage, because a page can only answer the questions it contains. That part is real.

The leap is assuming the relationship keeps going. It does not, for a simple structural reason: the questions that lead to money are a short list. A service buyer wants to know what it costs, what is included, how fast you come, whether you are licensed and whether anyone has regretted hiring you. There are maybe eight of those. Page nine onwards is answering something nobody is buying on.

An agency's incentives make this worse. Twenty pages a month is easy to quote, easy to invoice and easy to show on a report. Four pages that answer the buying questions looks like a light month until the enquiries arrive, and a quarter can pass before anyone can tell. We sell content by the month, so we are describing our own industry's temptation here, not someone else's.

Google's spam policies put the far end of this plainly: scaled content abuse is "when many pages are generated for the primary purpose of manipulating search rankings and not helping users", and the policy names "using generative AI tools or other similar tools to generate many pages without adding value for users" among its examples. The method of writing is not the test. Whether the page is useful is the test.

What 10 competitors in one industry actually run

Here are the sitemap sizes we read on 3 October 2026, sorted. All ten businesses appeared for the same set of buyer searches in the same Singapore industry. One site did not return a readable sitemap, so nine are shown.

Sitemap URLs listed	How many of the sites
2	1
32 to 48	2
128 to 199	3
259	1
943	1
2,486	1

The median was 135. Four of the nine ran under 50 pages. One ran more pages than most Singapore SMEs will publish in a decade.

The pages themselves varied as widely. The shortest page we measured ran 387 words. The longest ran 5,066. The 387-word page was a pricing page that showed its rates. The 5,066-word page showed its rates, its minimum session, what it excluded and roughly eighteen question headings. Length tracked thoroughness on those two. It did not track it across the set.

Page count predicted nothing

We checked the same ten pages against one question a buyer actually asks: does this page give me a price and tell me the minimum I have to buy?

Among the three largest sites by page count, one did both in the same place, one had both but split across the page and its FAQ, and one showed a price with no minimum anywhere.

Among the three smallest, one did both in the same place, and two showed a price with no minimum.

One of 3 at each end. The 2-page site showed a price for a small office, listed its add-ons with prices beside them and ran six question headings. The full measurements are in the study we published alongside this.

A null result is less fun to publish than a correlation, so it rarely gets published. It is also the more useful finding here, because it tells you the lever you were sold is not connected to the thing you wanted. If 2 pages and 2,486 pages produce the same odds of answering the buying question, then answering the buying question is a separate decision you have to make on purpose.

Our own numbers, which do not flatter us

We will not ask you to take this on reasoning alone, so here is our own site.

Our content API returned 139 blog entries for Singapore on 3 October 2026, including drafts and a couple of old test pages. In the 28 days to that date, our top 60 queries by impressions in Search Console, each with at least 20 impressions, carried 21,688 impressions between them and produced 2 clicks. Both clicks sat on the same kind of query, an agency search, and 58 of the 60 produced none at all.

That is what a large content library looks like before it earns position. Average positions on those 60 queries ran from the twenties to the eighties. Volume of pages was not our constraint; trust and position were, and we have written separately about impressions without clicks rather than repeat it here.

The second number is about reading your own progress. We ran a robust trend test over 90 days of our own daily query data to 26 September 2026: Theil-Sen slope, Mann-Kendall significance and Benjamini-Hochberg false-discovery control. 97 queries had enough data to test. 13 survived. So 84 of 97 apparent movements, 87% of them, were not distinguishable from noise once you correct for testing many things at once.

Hold those two facts together and the page-count pitch gets harder to defend. A library of 139 pages produced 2 clicks on its best 60 queries in a month, and 87% of the movement in that data was noise. Nobody can honestly tell you, a month into a 20-page-a-month plan, whether it is working. Our note on reading a trend properly covers the statistics.

The rule I use instead

Three questions, in order. Stop at the first "no".

1. Is there a question buyers actually type that none of my pages answers? Not a topic. A phrase. If you cannot name the phrase, the page is speculative, and speculative pages are where libraries of 2,486 come from.
2. Can my page be the best answer to it, today, with what I know? If the honest answer is that three existing pages already answer it better and you have nothing to add, write something else. Information gain is the whole job.
3. If someone reads it and believes me, is there a next step that makes money? A page with no commercial exit is a nice thing to have done. It is not a reason to pay monthly.

Three yeses, write the page. Any no, you have found your answer: the next page is not the problem. Usually what is left is an editing job on pages you already own, which costs nothing and nobody can invoice you much for.

The test is not how many pages. It is whether the next page has a question waiting for it.

When more pages is the right answer

We should be fair to the other side, because sometimes it is.

If you sell many genuinely distinct things, page count follows the catalogue, not a strategy. A supplier with 400 real products needs 400 real product pages, and the 2,486-page site in our sample is a cleaning company that has built out a deep library of service and location variants over years. We measured its sitemap, not its results, and we are not saying it was wrong to build.

If you are in a category where buyers ask dozens of distinct questions, the list in rule one is genuinely long. Regulated and technical trades are like this: every rule, every certification, every edge case is a real page with real demand behind it.

And if you have already answered the buying questions well, the next tranche of pages is how you reach the next layer of buyers. That is the legitimate version of the pitch. The question to ask whoever is quoting you is simply which layer they think you are on, and what evidence puts you there.

What it costs

Going slower costs you coverage. A site publishing four well-aimed pages a month will, after a year, be absent from searches that a site publishing twenty will appear for. Some of those searches will have been worth something. You are trading breadth for a better hit rate, and that is a real trade, not a free lunch.

It also costs you the comfort of a visible plan. Twenty items on a content calendar is reassuring in a way that "we are not writing anything this month because nothing passed the test" is not. Some owners genuinely prefer activity, and there is no shame in that as long as you know which one you bought.

The point against our own side is unavoidable. We sell 10 or 20 articles a month. A rule that frequently returns "write nothing this month" is not in our commercial interest, and we are publishing it anyway, because the alternative is taking money for pages aimed at demand nobody could name. If you run the three questions and come back with four pages, hire somebody for four pages, and it does not have to be us. We have also written about the times SEO is the wrong thing to buy altogether.

One next step

Send us your sitemap and the three phrases you most want to win. We will tell you which of your existing pages already aims at them, which of them nothing on your site answers, and whether the gap is a writing job or an editing job. It starts as a free visibility check through our contact page.

If the answer turns out to be an editing job, we will say so, and you can do it yourself in an afternoon. If you decide you want the ongoing work, our SEO, AEO and GEO programme starts from S$500 a month on a three-month minimum. Our guide for cleaning companies is where today's measurements came from if you work in that trade.

Figures describe sitemaps read on 3 October 2026 and SingRank's own Search Console data for the 28 days to that date. This page names no business and rates no company.

FAQ

How many pages should a small business website have?

As many as there are buying questions you can answer better than the current results, which is usually between five and fifteen for a single-service firm. In our sample of ten Singapore competitors on 3 October 2026, sitemaps ran from 2 to 2,486 URLs with a median of 135, and size did not predict who answered the buyer's price question.

Is publishing more pages still good for SEO in 2026?

Only when each page answers something people search for. Google's spam policies define scaled content abuse as many pages generated primarily to manipulate rankings rather than help users, and list AI-generated pages that add no value as an example. The volume is not the offence; the emptiness is.

Should I buy a 20-article-a-month content package?

Ask the seller to name the specific search phrase each article targets and what it will say that page one does not. If they can do that twenty times, the package is real. If the answer is a list of topics, you are buying activity rather than coverage.

How long before I can tell whether content is working?

Longer than a month. On our own site, a robust trend test over 90 days found that only 13 of 97 testable queries showed movement distinguishable from noise once we corrected for multiple testing. Judging a content plan on four weeks of data is reading noise.

Does a bigger site rank better than a smaller one?

Not by size alone. Size often correlates with age, links and brand, which do matter, so big sites frequently do rank better for reasons that are not the page count. Adding pages does not import those reasons.

Can I cite these figures?

Yes. Please cite "SingRank, 3 October 2026" with a link to this page, and quote the sitemap range with its base of nine measured sites. The Search Console figures are our own domain's, not a client's. Corrections are welcome through our contact page.