Research · September 2026
The Funded Site Audit
We loaded the homepage of 1,348 companies that closed a funding round in June and July 2026 and recorded what a machine actually receives. One in twelve serves almost nothing without JavaScript — and the companies most likely to be invisible are the ones running the most modern technology.
Method: headless Chrome plus a parallel no-JavaScript fetch, 1 September 2026. Aggregate results only; no company is identified.
1,348
homepages analysed, of 1,403 attempted
117
serve a shell — 8.7% of the sample
1.2%
median text surviving without JavaScript
28
of those are buying Google Ads
What one homepage load can and cannot prove
Every site was visited twice within the same second: once in a real headless browser that executed JavaScript, and once with a plain HTTP fetch that did not. The comparison between those two documents is the entire basis of the central finding.
What we state as fact
- Whether a tag is present and its ID — taken from the network log, not a database lookup.
- The rendered-versus-no-JavaScript text gap. Both documents were captured; the ratio is arithmetic.
- Title, H1 and meta description as served.
- Whether a consent platform loaded before anything else could.
What we deliberately do not claim
- That any specific crawler saw this. The fetch used a normal browser user-agent, not GPTBot’s. What we measured is that the content is unavailable without JavaScript execution — the condition GPTBot, ClaudeBot, PerplexityBot and Google’s AdsBot operate under.
- That a missing tag is missing from the site. Only from the homepage, on one load. A conversion tag fires on a thank-you page; a Meta pixel fires Lead on submit.
- That any company is doing something wrong. Every figure counting an absence excludes the cases where we know we could not have seen it.
Definition. A homepage counts as a shell when it renders more than 600 characters of text with JavaScript and retains under 25% of that without it. Deliberately strict: substantial when rendered, nearly empty when not.
One site in twelve is effectively blank to anything that does not run JavaScript
117 of 1,348 homepages met the definition. The severity is the part worth sitting with: on a typical one, the rendered page carried 5,358 characters of text and the same URL fetched without JavaScript carried 55. Ninety of the 117 retained under 5%.
Median shell — rendered in a browser
5,358
characters of text
The same page, no JavaScript
55
characters — 1.2% survives
The more modern the stack, the more likely the site is invisible
Splitting the shells by platform inverts the expectation. The hosted builders developers dismiss produced no shells at all. Modern JavaScript frameworks produced the most.
Platforms with at least 20 sites in the sample. Higher is worse.
A further 500 sites could not be attributed to a platform; 14.0% of those are shells, the highest rate in the study. Next.js can render on the server — these 21 sites were configured not to.
A company that raised millions and hired engineers is more likely to be invisible to an assistant than one paying $16 a month for a website builder.
More money does not fix it
The obvious objection is that shells belong to the smallest companies and that scale solves it. Splitting the same 1,348 sites by round size and headcount produces a nearly flat line.
| Round size | Shells | Rate | Headcount | Shells | Rate |
|---|---|---|---|---|---|
| Under $1M | 29 / 309 | 9.4% | 1–10 | 55 / 626 | 8.8% |
| $1M – $5M | 27 / 293 | 9.2% | 11–50 | 30 / 421 | 7.1% |
| $5M – $20M | 31 / 331 | 9.4% | 51–100 | 8 / 104 | 7.7% |
| $20M and above | 30 / 415 | 7.2% | 101–250 | 6 / 91 | 6.6% |
Twenty-eight are buying clicks into a page the ad platform cannot read
Of the 117 shells, 28 carry Google Ads tags and 38 carry some paid pixel. Google evaluates landing pages to assess quality, and it evaluates them the same way every other crawler does. A page that returns 55 characters is being scored on those 55 characters while the advertiser pays for every click delivered to it. Across the whole sample, 327 companies are running Google Ads.
The unglamorous gaps are larger than the glamorous one
| Observation | Sites | Share |
|---|---|---|
| No meta description on the homepage | 369 | 27.4% |
| More than one H1 on the homepage | 294 | 21.8% |
| No H1 at all | 237 | 17.6% |
| No analytics detected on the homepage | 274 | 20.3% |
| No <title> element | 7 | 0.5% |
The analytics figure carries the heaviest caveat and has been reduced twice because of it. 327 sites showed no analytics signal of any kind. Removing those where a consent platform held every tag pending a click nobody made brings it to 304. Removing those on platforms that ship their own built-in measurement brings it to 274. That is the number in the table. The raw 327 is the one most studies would have published.
Three findings we wrote, then deleted
Each survived a first pass and was removed after audit. They are published because a study that reports only what confirmed its thesis is not worth reading.
“Their Meta pixel fires PageView and nothing else.”
A pixel fires Lead when a form is submitted. A crawler that loads one page and submits nothing can never observe it. Would have been false on 98 companies.
“No conversion tracking is wired to the ad account.”
Google Ads conversion tags fire on the thank-you page, not the homepage. Unobservable by this method.
“Nothing is measuring this site.”
Wrong three times over: consent banners hold tags; several analytics vendors were missing from our detection patterns entirely; and hosted builders ship their own measurement. Each correction cut the count — 327 to 304 to 274.
What follows from this
The check takes one minute and almost nobody runs it
Fetch your own homepage without executing JavaScript and read what comes back. If it is an empty container, every assistant, every ad-quality crawler and every link preview is working from that. We built a free tool that does it for you.
Framework choice is now a visibility decision
It used to be a developer-experience decision. On this evidence it determines whether a company can be read at all, and the teams best equipped to get it right are getting it wrong most often.
Nobody can promise you a ranking in an answer engine
Ask an assistant the same question twice and the business it names changes far more often than it stays the same. What can be controlled is eligibility: whether there is something to read, whether it is structured, whether the claims are verifiable. Treat anyone selling placement with suspicion.
About this study. Conducted by Revio Consulting on 1 September 2026. 1,403 domains attempted, 1,348 analysed, 16.5 minutes at concurrency 10, using headless Chrome and a parallel no-JavaScript fetch. Company-level results are not published and no company is identified. Figures are as observed on the day and websites change.