A landing page that does not rank usually fails at one of three stages: a crawler cannot read it, a crawler reads it and finds nothing to index, or it gets indexed and loses to a better page.
Most advice skips to the third stage. We measured the first two.
On 25 July 2026 we crawled 800 company sites from Y Combinator's public directory, filtered to active B2B and SaaS companies, with a published seed (20260725) so the sample can be rebuilt. 675 produced a page we could score. The 125 that did not are the most useful part of this article.
Stage one: 83 sites showed a crawler nothing
Of the 800 sites we attempted, 83 served no meaningful HTML until a JavaScript bundle executed.
That single failure outnumbered every other cause combined. Full ledger:
| Reason | Count |
|---|---|
| No HTML without JavaScript | 83 |
| Network error | 13 |
| HTTP 403 | 8 |
| Body too small to score | 7 |
| HTTP 429 | 5 |
| HTTP 404 | 4 |
| Robots disallowed | 2 |
| 503, 526, 402 | 3 |
Googlebot renders JavaScript, though on a deferred queue rather than at first crawl. Other crawlers vary, and AI retrieval systems make no guarantee at all. If your marketing page is an application shell that fills in after hydration, you are betting discoverability on someone else's rendering budget.
Test it in one minute. Disable JavaScript in your browser and reload your landing page. A blank screen or a spinner is what part of the web sees.
This is the cheapest large win in landing page SEO, and almost nobody writes about it, because the companies publishing SEO checklists host their own pages and never see the failure.
Stage two: the page renders, and there is nothing to index
Among the 675 pages we scored:
| Measurement | Result | n |
|---|---|---|
No <h1> element at all | 8.1% | 675 |
| Headline over ten words | 7.1% | 620 |
| Median images with no alt text | 6 (p75: 34.5) | 675 |
| Median word count | 759 (IQR 458–1,218) | 675 |
| Median reading grade | 10.4 | 595 |
The H1 number surprises people. These pages have a visible headline, styled to look like a heading, marked up as a div. To a crawler the page has no main heading, and the strongest on-page signal is gone.
Alt text is the same shape of problem. The WebAIM Million report for 2026, covering one million home pages, found 53.1% missing alternative text for images, alongside 83.9% with low contrast text and 51% missing form input labels. Their headline finding is worse than ours: 95.9% of home pages had detected WCAG 2 failures, and errors rose 10.1% year on year to 56.1 per page, reversing six years of improvement.
Alt text is an accessibility obligation first. It also happens to be page context a crawler can read.
Stage three: ranking, and the part where evidence runs out
Once you are indexed, the advice market gets loud and the evidence gets thin.
That is worth stating plainly because of what happened to the AI-search version of this question. Olivier Martinez surveyed 45 studies on generative engine optimization published between November 2023 and July 2026 and concluded that no reviewed technique shows a stable, longitudinal, cross-platform causal effect on organic discoverability.
His breakdown is the useful part. Evidence is strong that a document already in an engine's context can change how it gets cited. Evidence is far weaker that any technique makes a page get retrieved at all. Evidence is close to absent that any of it produces a durable effect on clicks or conversions. Some GEO rewrite methods cut a page's AI retrieval by 16%.
So treat "optimise for AI search" claims the way you would treat any unproven tactic, and spend the effort on stage one instead, where the failure is measurable and the fix is certain.
What to do, in order
- Disable JavaScript and reload. Fix it if the page disappears. This outranks everything below.
- Confirm an
<h1>exists in the markup, not a styled div. Ten seconds in developer tools. - Write alt text. Median 6 missing per page in our sample, 53.1% of a million home pages in WebAIM's.
- Give the page one clear topic. 30.7% of the pages we scored had no call to action in the body, which tells a crawler as little as it tells a visitor.
- Match the query's format. A query whose top ten is galleries wants a gallery. Checking the SERP costs nothing and saves you writing the wrong article, which is a mistake we made on this very site and wrote up rather than hid.
What we did not measure
Page speed and Core Web Vitals. Our crawler fetches HTML and scores it; it does not measure render timing, and we will not quote numbers we did not produce.
Whether the CTA sits below the fold, and whether a page breaks at 390px. Both need a rendered viewport at a stated screen size. Any source quoting fold statistics without naming the viewport is guessing.
Method
Y Combinator public company directory via the yc-oss mirror, active companies classified B2B or SaaS. Seeded shuffle, seed 20260725. 800 attempted, 675 scored, collected 25 July 2026. Robots.txt honoured, identifying user agent, bounded concurrency. Failures are disclosed rather than scored as zeros, because a scored empty page drags every median toward zero.
Sources
- WebAIM, The WebAIM Million: The 2026 report on the accessibility of the top 1,000,000 home pages, February 2026.
- Martinez, O. (2026). Optimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization (2023-2026). arXiv:2607.14035.
- LeadzLander, The State of SaaS Landing Pages (2026). 800 attempted, 675 scored, collected 2026-07-25, seed 20260725.