Your docs are emptier than your homepage: 7 of 10 sites

FIELD TEST · 2026-08-15 · 10 sites · single snapshot

Sample and method: 10 well-known sites. For each we fetched the homepage and one documentation or help page with the OAI-SearchBot user agent on 15 August 2026, stripped scripts and tags, and compared readable characters.

Your docs are emptier than your homepage more often than not: on 7 of 10 sites, the documentation page delivered less readable text to a crawler than the marketing homepage. On linear.app the docs page returned 1,264 characters against a homepage of 9,526 — the page built to explain the product gave a crawler an eighth of what the page built to sell it did.

How we measured it

Two requests per site, no rendering, no retries. We took the homepage and the most obvious documentation or help entry point, removed <script> and <style>, stripped tags, collapsed whitespace and counted characters.

Docs entry points are not a perfect comparison — some are landing pages for a docs section rather than an article. That is part of the finding rather than a flaw in it, because a docs landing page that delivers nothing is still a docs landing page that delivers nothing.

Your docs are emptier than your homepage: the numbers

Sorted by ratio, best first. Anything under 1.0 means the documentation delivered less than the marketing page.

Why this is the wrong way round

Marketing homepages are the pages companies rebuild most often, with the most attention paid to how they load. Documentation is frequently an application: a sidebar, a search box, and content that arrives after the shell does.

But documentation is the content an answer engine most wants. Someone asking how to configure a thing is asking a documentation question. If the docs are a shell and the homepage is not, the crawler ends up with your positioning statement and none of your answers.

The one that got it right

stripe.com is the outlier in the useful direction: 21,369 characters on the docs side against 7,342 on the homepage, nearly three times more. That is what it looks like when documentation is served rather than assembled in the browser.

It is worth noting we are measuring delivery, not quality. A page can deliver 20,000 characters of navigation and still be useless. What this measurement establishes is the floor: below a few hundred characters there is nothing to be useless with.

The site that is empty in both directions

canva.com returned 342 characters for the homepage and 342 for the help page. Identical, which almost certainly means the same shell was served for both and neither contains content before scripts run.

We did not investigate further, and one request per URL is not enough to claim this is how it always behaves.

What this does not show

Ten sites, one request each, homepages and one docs entry point, from a single machine on one day. It does not show whether deeper article pages behave like the entry points, whether Google's delayed rendering pass eventually picks the content up, or whether any of this affected a citation anywhere.

It also cannot separate "docs are client-rendered" from "this particular URL is a landing page with little text by design". Both produce the same number, and we did not open each one to tell them apart. That ambiguity is the main weakness of this sample.

Common questions

How did you measure this?

One curl per URL with the OAI-SearchBot user agent, scripts and styles removed, tags stripped, characters counted. The step-by-step version, including how to run it against three of your own URLs at once, is in can AI crawl JavaScript.

Does this mean these sites have an SEO problem?

Not necessarily for Google, which renders JavaScript on a separate, later pass. It means the content is absent for clients that do not render, and most crawlers feeding AI answers are in that group.

Why compare against the homepage rather than an absolute threshold?

Because there is no meaningful absolute threshold — pages are supposed to be different lengths. The homepage is a within-site control: same stack, same company, same day.

The one number to take away

Seven of ten sites deliver less readable text on their documentation than on their marketing homepage. If you have only ever tested your homepage, you have tested the page least likely to have the problem. Fixing that, shipping it and pushing it for indexing is what changes the answer.