Verification · three numbers · line by line

Three numbers look too bad to be true. All three are real.

Before anyone asks "says who?" — the ad count, the load time, and the landing-page split all trace to public, checkable sources. Every receipt below has a link so you can run the count yourself in about two minutes. The same evidence the audit reads from, no engine in the way.

Domainpulmonx.com
Verified2026·06·11
SourcesGoogle · LinkedIn · Meta
Numbers reviewed3 of 3
01 · The ad count

Fifty-two active creatives, counted from the platforms' own libraries.

Every ad platform publishes a public library of the ads each account is running. Counts only come from accounts the platform itself ties to pulmonx.com — by domain key, by the verified page ID, or by the company page's own declared website. A lookalike never pollutes the number.

Google Ads Transparency
12 active
Domain-keyed lookup against pulmonx.com — Google's own listing, not an estimate. The strongest identity tier of the three.
Source Count them
adstransparency.google.com
LinkedIn Ad Library
24 active
linkedin.com/company/pulmonx — LinkedIn's record for that page declares pulmonx.com as its website, so the count is identity-verified.
Source Count them
linkedin.com/ad-library
Meta Ad Library
16 active
Meta page "Pulmonx" (page ID 250712428387404, IG handle zephyrbypulmonx). Verified by the page's own identity matching the company.
Source Count them
facebook.com/ads/library
All verified ad libraries combinedActive creatives, three platforms
52

Counts as of 11 June 2026. Ad libraries move daily, so the total drifts a few creatives either way. X ads do not appear here because X publishes no equivalent public library, so the audit refuses to guess a number for it.

02 · The load time

Google's own data says the corporate site is slow. Twice over.

Two different measurements, both from Google. The one that matters most is the field data: what real Chrome users actually experienced on pulmonx.com over the last 28 days. Lab numbers swing run to run — that's why any single score can look like an outlier. The field number doesn't move.

Real Chrome users · 28-day window
4.9s field LCP
Main content paint at the 75th percentile (CrUX). Google's "good" bar is 2.5s.
Google rating · slow
PageSpeed Insights · 2026·06·11
37 / 100
Mobile performance score Google produced when we measured fresh today. Three same-day scores — 35, 37, 49 — every one in Google's red band.
Lab main-content paint · three runs
39–46s
38.9s, 39.1s, 45.9s. Lab simulates a mid-range phone on throttled 4G, so it swings — every run still failed.

The honest takeaway is not any single lab number; it is that real users wait ~5 seconds for the main content, which Google formally categorises as slow — on the domain where the conversion tracking lives.

03 · The split

The Meta ads land on one domain. The tracking lives on the other.

Open the Meta Ad Library and click any creative — the library shows each ad's destination link. 13 of the 16 active Meta ads send patients to zephyrvalves.com/breathe-it-in, a separate campaign domain. The conversion stack verified firing in a real browser sits on pulmonx.com.

What fires where, exactly as measured:

A tag manager can inject tags after delivery, so the claim is held to exactly what was measured: eight ad-platform tags observed firing on pulmonx.com in a real browser; none of those conversion tags present in the campaign page's delivered HTML (Google Tag Manager and HubSpot only). The destination count comes from the ad library itself — click any creative to see its landing link.

04 · How the numbers stay honest

Four rules every claim has to pass before it ships.

When the audit shows a number, one of these four rules backs it. When a rule can't be satisfied, the number isn't shown.

Measured, or it doesn't ship

Every figure traces to an API response or a real-browser observation. A claim with no measurement behind it never reaches the page.

Withheld, never guessed

If a count can't be verified, it is left out entirely. A missing number is reported as not measured — not as zero, not as an estimate.

Identity-gated

Ad counts only come from accounts the platform itself ties to pulmonx.com — by domain key or by the page's own declared identity. Name lookalikes never count.

Time-bounded

Observations keep their limits. Lab speed runs are reported as the runs they were, with the swing shown. A one-off reading is never inflated into a permanent fact.