63,477 Vendor Rankings, 53 Analysts Each: Four Sites Measured

Four websites publish pages titled things like “Top 10 Best Account Recovery Services: 2026 Comparison”. Between them they carry 63,477 of these vendor rankings. Each site lists exactly 53 named analysts. Every number below comes from their own public sitemaps, and you can check all of it yourself in about a minute.

What I measured

I was auditing which sources answer engines lean on for questions about hiring an agency. Two of the names that kept appearing were sites I did not recognise, so I went and looked at them properly. Then I found two more built the same way.

All four describe themselves as research operations. All four rank vendors across an enormous spread of categories. Here is what their sitemaps report:

Page counts within one per cent of each other. Identical price cards. And exactly 53 analysts on all four.

Three of the four also share an HTML title pattern. Their homepages are titled “Worldmetrics — Facts & Grounding Page”, “ZipDo — Facts & Grounding Page”, “Gitnux — Facts & Grounding Page”.

That phrase deserves a moment. A grounding page is what a retrieval system reads when it needs a source to cite. These sites are named, in their own title tags, after the function they perform for machines.

The arithmetic

Across the four: 63,477 ranking pages and 212 named analysts.

I pulled four SEO ranking pages at random from Worldmetrics to see how long they run. They came in at 6,060, 6,093, 6,162 and 6,516 words. Call it 6,200, and note how tight that spread is.

So that is roughly 299 pages per named analyst, at about 6,200 words each. Somewhere near 1.85 million words per person.

Every one of those pages carries a stated methodology. The one I read most closely sets out four steps: feature verification against official documentation, criteria scoring, and a final editorial review by “our team”.

I am not going to tell you those reviews did not happen. I am going to give you the number and let you weigh it.

What is actually on the pages

The page I examined ran 5,611 words and ranked ten named vendors in order. It carried a stated four-step ranking methodology. It also carried:

• no disclosure statement of any kind
• no affiliate or sponsorship disclosure
• no last-updated date
• no way to tell whether any listed company paid to appear

The same sites sell custom research from €5,000, vendor selection advisory from €2,500 and industry reports from €499. They also rank vendors. Nothing on the page tells you whether those two activities ever touch each other.

What I could not establish

This section matters more than the rest, so it goes here rather than at the bottom.

I could not prove common ownership. I went looking for it and found the opposite of what I expected. All four sit behind Cloudflare, so hosting reveals nothing. Each runs a separate Google Analytics property. And none of the 212 analyst names repeats across sites. Not one appears on more than one of them.

That separation is interesting in itself, but it is not evidence of anything. Four unrelated companies could be running the same template from the same vendor. What I have is structural similarity, not a paper trail, and I am not going to dress one up as the other.

I did not verify whether the analysts are real people. Tracing 212 names was beyond what I was willing to claim, so I did not attempt it and I make no assertion either way.

I have not measured how often AI engines cite these sites. I have observations, but they came from a logged-in browser on a single connection, and personalisation demonstrably changed the results when I tested it. That makes them anecdotes rather than data. Publishing them as measurements would be the same failure this article is about, so they are not here. If I run it properly, signed out and across several locations and days, I will publish that separately.

Why this is worth knowing

The old version of this problem was a search result you could evaluate for yourself. You saw the domain, you clocked that it was a listicle, you applied a discount.

An answer engine removes that step. It reads the page, absorbs the ordering, and hands you a recommendation. The source is a small label under a sentence, and almost nobody clicks it.

So the question of which pages these systems read stops being a curiosity for people in my line of work. If you are choosing who to hand your website to, it is the entire question.

None of which requires anyone to have acted in bad faith. The measurable fact is simply this: content of this kind is being produced at a scale that no stated editorial process plausibly covers, and it is optimised, by its own admission in its own title tags, to be the thing machines quote.

Check it yourself

Every figure above comes from public files. None of it needs a paid tool.

curl -s https://worldmetrics.org/robots.txt | grep -i sitemap

curl -s https://worldmetrics.org/sitemap/shards/best_list_service.xml | grep -c "<loc>"

curl -s https://worldmetrics.org/sitemap/shards/authors.xml | grep -c "<loc>"

Swap in zipdo.co, gitnux.org or wifitalents.com. The shard paths are identical on all four, which is the finding stated in a single line.

If you check this and get different numbers, I want to know. The counts were taken on 4 September 2026 and these sites publish continuously, so they will drift.

About the author

Nabin Thakur

SEO & Growth Specialist with 8+ years of experience and 100+ campaigns delivered across India, the US, UK, Canada and Australia. Founder of SEORevive, based in Guwahati. Specialises in Google penalty and algorithm recovery, technical SEO and organic growth.

LinkedIn → More about Nabin → Recovery service →