Learn

Reference pages, written to be read rather than skimmed for keywords.

Googlebot's 2MB cutoff

It reads the first 2MB and indexes that as the whole page. We measured well-known homepages and found five already past it.

How to tell whether an audit tool is lying to you

Four questions to ask of any SEO finding, each of them learned here by getting it wrong first.

Log file analysis: what Googlebot actually fetched

A user-agent proves nothing, which is why Google publishes 1,641 crawler IP prefixes. Reading a log honestly, and where the dedicated tool wins.

Brand consistency: the question no crawler asks

Of 11 company sites linking social profiles, 9 declared none of them in schema. What the brand lane checks, and where design tools beat it.

The contact address that cannot receive mail

An address on a domain with no MX record bounces to the sender and never reaches you, and no tool asks whether yours works.

Which pages an AI answer replaces

Ranking and not being visited. Measured on two live sites — this one at 5% fully substitutable, a delicatessen at 0% — and three ways we measured it wrong first.

AI search visibility

The three gates a model has to clear before it can cite you — access, rendering and entity clarity — with measured data on who is blocking what.

What an SEO audit covers

Every area, in the order they should be worked, and the three tests a report has to pass to be worth acting on.

JavaScript rendering

What a crawler that does not run JavaScript misses — measured on a page that serves 0 characters of text and renders 2,068.

sameAs and entity signals

The cheapest entity signal there is, and the share of major sites that skip it — measured, with the dataset attached.

Internal link equity

The ranking signal your pages pass to each other, measured on our own site — where the download page held a fifth of what an average page did.

What Docket checks

All the checks, by area, with what each one actually looks at.