Most checks in the rubric read the homepage. D7 is one of the few that reads all six crawled pages together and asks whether they behave like one site. It awards four points for three properties: every crawled page returns 200, no page takes more than four seconds to arrive, and the navigation is present on all of them or on none. A site fails it not because any single page is bad, but because its pages disagree with each other.
The premise is that an agent moving through a site is building a model of it as it goes. A page that 404s, a page that stalls, or a page with no navigation where every other page had one, is a page the agent cannot place, and a site it has less reason to trust on the next fetch.
Which pages the scanner reads
The free scan fetches the homepage and up to five more pages, chosen from the sitemap and the homepage's links. The choice is not random. Same-origin URLs are ranked with pricing and plans first, then products, shop and collections, then docs, then about and company, then contact, then cart and checkout, with anything else after. Duplicate paths are removed and the first five are fetched. The rules for our crawler, including how to block it, are on the bot page.
Two things follow. The pages D7 examines are the commercially important ones, because those are the pages an agent is most likely to be sent to. And they are on one origin: a docs site on its own subdomain is never in the set, which matters for the drift examples below.
Every page returns 200
This is worth two of the four points, and it is the strictest of the three. The crawler follows redirects, so a page that redirects to a 200 counts as a 200, and the report records the final URL it landed on. A page that redirects to a 404 is a 404. A page that does not answer within ten seconds is recorded with a status of 0 and an error, and it counts as a failure just as a 500 would.
Any page not returning 200 fails the whole two points. The evidence block lists each one with its URL and status under nonOk, so the report tells you which link on your homepage, or which entry in your sitemap, is pointing somewhere it should not.
The usual culprits are sitemap entries for pages that were removed without being removed from the sitemap, links in the footer to a page that was renamed, and a pricing or docs route that moved behind a login. None of these is visible from the homepage, which is why a homepage-only check would never find them.
No page over four seconds
One point. The crawler starts a timer when it sends the request and stops it when it has read the response body in full, and the report calls that number ttfbMs. In practice it is closer to a total response time than a strict time to first byte, so a page that starts responding quickly but streams a large body slowly is measured on the whole of it. Any page over 4,000 milliseconds fails the point, and the evidence records how many pages did.
Four seconds is deliberately loose. The bot probes in A5 hold the homepage to a much tighter bar, and Time to first byte for bots explains why 1.5 seconds is the line there. D7's threshold is not about performance; it is about pages that are pathologically slow relative to the rest of the site, typically because they run an uncached query, call a third-party API on every request, or render a large collection with no pagination. A site whose homepage is fast and whose products page takes six seconds has a stability problem, not a speed problem.
The same navigation throughout
One point, and the one that needs explaining, because it does not measure what people first assume.
For each page that returned 200 with a non-empty body, the check parses the raw HTML and counts nav elements. The site is consistent if there is only one page to compare, or if every page has at least one nav, or if every page has none. Anything else, some pages with a nav and some without, fails the point, and the fix text says navigation is missing on some pages.
Three consequences are worth spelling out. First, the check reads raw HTML, not a rendered page, so a navigation that is injected by JavaScript after load does not exist as far as D7 is concerned. Second, it counts elements, not links, so it does not care whether the links inside the nav are the same on each page, only that a nav is there. Third, a site with no nav element anywhere passes this point, because it is consistent. Whether it should have landmarks at all is a different check; the landmarks post covers B3, which is where a missing nav costs you.
What D7 is measuring, then, is whether the six pages were built from the same template. The nav element is a proxy for that. It is a good proxy because it is the thing most likely to differ when a page was built somewhere else.
What template drift looks like
Drift is rarely a decision. It is what happens when a site is built by more than one process over more than one year.
The landing page with its own header. Marketing builds a campaign page in a page builder, or from a separate template that strips the global navigation to keep visitors on the page. It has a logo and a form and nothing else. It is linked from the homepage, so the crawler picks it up, and it is the one page in six with no nav. The point is lost, and more to the point, an agent that arrives there has no route to anything else on the site.
The docs subdomain with its own navigation. Documentation is often a separate product on a separate host with its own sidebar, search and header. The free scan does not cross origins, so this one does not show up in D7 directly. It shows up when docs live under a path on the main host, and the docs tool renders a completely different shell from the marketing pages. An agent following a link from pricing to docs then walks from one site into what looks like another.
The legacy page on an old template. The contact page, the terms page, or the one product category that was never migrated when the site was rebuilt. It still has a navigation, but the navigation is a div with a list of links, because the old template predates the team's use of landmarks. Every other page has a nav; this one does not. It is the hardest drift to see in a browser, because it looks identical to a visitor, and the easiest to see in the evidence block.
The common fix for all three is the same: one layout, applied to every route, with the navigation as a nav element in the server-rendered HTML. Where a page genuinely needs a stripped-down header, keep the nav and reduce its contents rather than removing the element.
Reading the evidence
The D7 block on a report records pagesCrawled, the nonOk list with URL and status, slowPages as a count, and consistentTemplate as true or false. If the score is less than four, the fix text names which of the three properties failed and, for status failures, exactly which pages. Because the check fails on any one page, a single stale sitemap entry can cost two points; it is worth fixing that before anything more structural.
Check your own site
Run the free scan and read the D7 evidence, particularly the nonOk list. The definition of the check, and the rest of the trust pillar, is on the methodology page.