How Adnave measures websites

How is the audit carried out?

Our crawler, AdnaveBot, reads the page you enter, then follows its internal links to up to 12 pages (60 with a free account) and checks further links for errors. It follows the site's robots.txt rules and any requested pause between visits.

Pages are read as a search engine first receives them. If a page arrives nearly empty and builds itself with scripts, we open it in a real browser, read the result, and tell you, because not every search engine or AI assistant will do the same.

How is the organic health score calculated?

Each check has a weight reflecting how much it matters, from 1 (minor) to 12 (a page hidden from search). A pass earns the full weight, a partial result half, and a fail nothing. The score is the weight earned divided by the weight available, as a number out of 100.

Checks about advertising tags are scored separately as ads readiness, because they say nothing about being found in search.

The weights are our judgment, informed by what search engines publish about their own systems. They are not derived from ranking data, and we say so.

Does a high score mean a site will rank?

No. We test this on our own data and publish the result in our study of the top sites: among the world's most popular websites, audit score and authority are barely related. A good score means known obstacles are removed. Ranking also depends on what other sites link to you and how useful your content is compared with the alternatives.

Where does the authority score come from?

From open data. The Common Crawl project publishes a graph of links between domains and ranks every domain by how well-linked it is. We keep the best-linked 20 million. The score is 100 for the top-ranked domain and falls by about 12 points for each tenfold drop in rank.

We check the score against an outside source. In November 2026 we compared our ranking with the Majestic Million, a list built from a commercial company's own separate crawl of the web. For the 10,000 best-linked domains the two orders agree with a correlation of 0.67, where 1 would be identical and 0 unrelated; across the top 100,000 it is 0.56. Further down, between ranks 100,000 and 300,000, agreement falls to 0.23. In plain words: the score is a sound guide to the big picture and to large differences between sites, and a rough one for telling apart two small sites with similar scores. That test also made us change the formula: the web graph offers two rankings, and we now use the one that agreed better.

Popularity rank comes from the Tranco list of the top million sites. Wikipedia citations are counted through Wikipedia's own interface, up to 1,000. Registration year comes from the public domain registry. None of these is a figure Google uses or publishes.

Where do the backlink lists come from?

From the same Common Crawl release as the authority score. Common Crawl publishes which domains link to which, about two billion links in each release. For each of the 20 million best-linked domains we count the different domains linking to it and keep the 100 most authoritative of them.

We checked the counts against an outside source. Across 93,000 of the most-linked domains, the number of linking domains we count agrees with the number of linking networks in the Majestic Million, a commercial index built from a separate crawl, with a correlation of 0.71 (1 would be identical order). The two count slightly different things, so the figures themselves differ; the ordering of sites is what agrees.

The lists work at the level of whole websites. They do not show the page a link is on or its wording, and because Common Crawl reads a large sample of the web over three months rather than all of it continuously, they are less complete than a paid tool's own index, most of all for small sites.

How is speed measured?

With the open-source Lighthouse engine, in a real browser on our own server, simulating a phone on a slow connection. This is a single lab run: results vary between runs. It is not the real-visitor data Google collects, which needs a site's actual traffic.

How are keywords found?

By reading the pages: words and phrases are weighted by where they appear (titles most, then headings, descriptions and body text) and by how many pages use them. This shows what a site emphasises. It does not show how often people search for a term; that figure exists only inside search engines, so we never estimate it.

Where do the topic interest figures come from?

From Wikipedia, which publishes how many people read each article every month. For each term we find the article about it and report its average monthly readers over the last three full months, the last twelve months as a line, and the change on the same months a year earlier. Because Wikipedia's total readership has been falling, that change is measured against Wikipedia as a whole, so a topic is shown as falling only if it fell faster than the site did.

This is not search volume. It counts people reading about a subject, not people searching to buy something, so commercial phrases often have low figures or no topic at all. We tested it on fifteen pairs of topics where one is plainly more popular than the other, and it ordered fourteen correctly. It is reliable for comparing subjects and spotting trends and seasons, and should not be read as a number of Google searches.

How are the studies produced?

We audit the most popular websites, taken from the Tranco list, about once a month, and count the results. Sites that refuse automated visits or have no web page are left out, and each study says how many sites it covers. Adult and gambling sites are excluded.

What can Adnave not tell you?

Search rankings, search volumes, the specific pages and wording of the links to a site, traffic figures, and anything inside another company's account. Where a competitor's tool shows such a number, it comes from that tool's own estimates. We would rather show a measured figure with its source than an estimate presented as fact.