Audit Results

https://www.nytimes.com/2026/03/23/opinion/denmark-mette-frederiksen-election.html

React

Lighthouse Lab Data

Measured in a simulated environment. Values may differ from real user experience.

Performance Score

Largest Contentful Paint

0.5 s

Cumulative Layout Shift

0.0362

Interaction to Next Paint

1702 ms

Good (90–100)Needs Improvement (50–89)Poor (0–49)

Field Data — Mobile (Real Users)

Core Web Vitals Poor

Chrome UX Report — p75 values from real mobile user experiences over the last 28 days

Largest Contentful Paint (LCP)

2.2 sNeeds Improvement

80.3%

11.1%

8.6%

Interaction to Next Paint (INP)

149 msNeeds Improvement

85.6%

11.4%

Cumulative Layout Shift (CLS)

0.99Needs Improvement

56.6%

39.5%

Time to First Byte (TTFB)

440 msGood

88.1%

9.1%

GoodNeeds ImprovementPoor

Summary

Critical issues detected

CrUX field CLS of 0.99 is catastrophically poor (threshold: 0.25) and is the most urgent issue — real users experience massive layout shifts from ads, font swaps, and dynamic content during scrolling that lab tests completely miss. LCP at 2174ms is in the needs-improvement zone, impacted by 2 GTM containers (305KB combined), 49 paused dead tags, 595 unused variables, and heavy ad-tech scripts competing with LCP resources. INP at 149ms is currently good but at risk given 37 long tasks and 6104ms main thread work.

Must do

1Reserve space for ad slots to eliminate catastrophic field CLS
2Add font-display to 5 web fonts missing it to prevent text swap shifts
3Delete 49 paused tags and duplicate trackers from GTM-P528B3
4Consolidate 2 GTM containers into 1 to eliminate duplicate payload

Can defer

—Preconnect to critical third-party origins used before LCP
—Reduce main JS bundle: story.js (631KB) and main.js (914KB) dominate the payload
—Defer ad-tech and measurement scripts that block main thread

Expected outcome

CLS: 0.99 → ~0.10–0.15, LCP: 2174ms → ~1700–1900ms, INP: 149ms → maintained at <150ms (protected from degradation)

Recommendations

1highclsBoth

Add font-display to 5 web fonts missing it to prevent text swap shifts

5 web fonts loaded from g1.nyt.com have no font-display property: franklin-normal-700, franklin-normal-500, cheltenham-normal-700, cheltenham-cond-normal-700, imperial-normal-400. Without font-display, browsers use their default behavior (Chrome: 3s invisible text block, then swap). When fonts finish loading at ~1700–2350ms, the swap from fallback to custom font causes layout shifts — especially with the large headline font (Cheltenham) which has very different metrics from system fallbacks. This is a significant contributor to field CLS = 0.99. The fonts are loaded via a CSS file at g1.nyt.com/fonts/css/web-fonts.css. Since NYT controls this domain:

1Add font-display: optional to the @font-face rules for body text fonts (imperial-normal-400, franklin-normal-500). font-display: optional gives the browser ~100ms to download the font; if it misses, it uses the system font with zero CLS — the custom font is cached for the next navigation.

2Add font-display: swap to headline fonts (cheltenham-normal-700, cheltenham-cond-normal-700, franklin-normal-700) since branding matters for headlines, but pair with size-adjust, ascent-override, descent-override on the fallback @font-face to minimize the shift.

3Add <link rel="preload" as="font" type="font/woff2" crossorigin href="...cheltenham-normal-700..."> for the headline font to reduce load time.

Before

@font-face {
  font-family: 'nyt-cheltenham';
  src: url('cheltenham-normal-700.woff2') format('woff2');
  font-weight: 700;
  font-style: normal;
}

After

@font-face {
  font-family: 'nyt-cheltenham';
  src: url('cheltenham-normal-700.woff2') format('woff2');
  font-weight: 700;
  font-style: normal;
  font-display: swap;
}

/* Fallback font-face with metric overrides to minimize CLS */
@font-face {
  font-family: 'nyt-cheltenham-fallback';
  src: local('Georgia');
  size-adjust: 105%;
  ascent-override: 95%;
  descent-override: 22%;
  line-gap-override: 0%;
}

@font-face {
  font-family: 'nyt-imperial';
  src: url('imperial-normal-400.woff2') format('woff2');
  font-weight: 400;
  font-style: normal;
  font-display: optional;
}

Learn more

Expected impact

CLS: reduce font-swap contribution by ~0.05–0.15. Combined with ad slot fixes, targets CLS p75 < 0.1.

2highclsMobile

Make the app download banner position:fixed to eliminate CLS

The screenshot shows a "Continue reading in the app — it's better" banner at the bottom of the viewport. This banner is injected dynamically after page load (likely by the unified-lire.bundle.js script loaded at 3629ms, 287KB). If this banner is rendered with position: relative or position: sticky with a top offset, it shifts page content when it appears. Fixed-positioned elements do not cause layout shifts because they're removed from the document flow.

1Change the banner's CSS positioning to position: fixed; bottom: 0; left: 0; right: 0; z-index: 1000.

2Add padding-bottom to <body> equal to the banner height (approximately 60–70px) to prevent the banner from covering article text — but add this padding from the start, not dynamically.

Before

.app-banner {
  position: sticky;
  bottom: 0;
  /* dynamically inserted, shifts content */
}

After

.app-banner {
  position: fixed;
  bottom: 0;
  left: 0;
  right: 0;
  z-index: 1000;
}

/* Reserve space in body from initial render */
body {
  padding-bottom: 68px;
}

Learn more

Expected impact

CLS: eliminate ~0.05–0.10 from banner insertion shift. This is a consistent shift affecting every mobile visitor.

3highlcpBoth

Consolidate 2 GTM containers into 1 to eliminate duplicate payload

The page loads 2 separate GTM containers: GTM-P528B3 (188.7KB, starts at 1060ms, takes 2663ms to fully load — 660ms Conn+TLS + 356ms TTFB + 1309ms download) and GTM-N5P6T9S (115.9KB, starts at 5007ms). That's ~305KB of GTM JavaScript — two separate network requests, two separate parse+execute cycles competing with LCP resources. GTM-N5P6T9S appears to be a subscription/conversion tracking container (all its variables reference subscription.* and purchaseFlow.*). On an article page, this container does almost nothing but still loads its full payload.

1Move the 25 enabled tags from GTM-N5P6T9S into GTM-P528B3 (the primary container with more active infrastructure).

2Add a trigger condition in GTM-P528B3 so the migrated subscription tags only fire on /subscription/ pages — they're irrelevant on article pages.

3Remove the GTM-N5P6T9S snippet from the page template. Search the source code (likely in the Express server templates or React <head> component) for GTM-N5P6T9S and delete the entire <script> block.

4The 24 Custom Image Pixel tags in GTM-N5P6T9S are likely conversion pixels — verify they fire correctly from the consolidated container before removing the old one.

Google Tag Manager · GTM-N5P6T9S3 tags

Custom Tag (__cvt_12450453_170)

cvt_12450453_170

MODIFY

Custom HTML

html

MODIFY

24 × Custom Image (Pixel) (unidentifiable)

img

MODIFY

Learn more

Expected impact

LCP: save ~116KB download + ~200–400ms main thread time on mobile. Field LCP 2174ms → ~1900ms.

4highlcpBoth

Delete 49 paused tags and duplicate trackers from GTM-P528B3

GTM-P528B3 contains 49 paused tags that still contribute to the container's JavaScript bundle size and browser parsing time. Key dead tags include paused ad-tech and social pixels from vendors likely no longer in use:

1Open GTM container GTM-P528B3 → Tags → filter Status: Paused.

2Delete the following identifiable paused tags: - Facebook Pixel SDK (loads connect.facebook.net/fbevents.js) — paused, adds ~60KB to container - Snapchat Pixel #1 and Snapchat Pixel #2 (both load sc-static.net/scevent.min.js) — duplicate paused pixels - TikTok Pixel #1 and TikTok Pixel #2 (both load analytics.tiktok.com/events.js) — duplicate paused pixels - Script: c.amazon-adsystem.com #1 through #6 (6 paused tags all loading c.amazon-adsystem.com/amzn.js) — massive duplication - Script: platform.iteratehq.com (loads platform.iteratehq.com/loader.js) — paused survey tool - Script: cdn.brandmetrics.com (loads cdn.brandmetrics.com/nyt.js) — paused brand measurement - Cookie Script (loads static.chartbeat.com/chartbeat_video.js) — paused Chartbeat - Custom HTML (dd.nytimes.com/tags.js) — paused, but note this script still loads inline (waterfall #39) - DataLayer Push #11 (loads apps.rokt-api.com) — paused Rokt integration - Google Ads Conversion (1008590664) — paused, Conv ID: 1008590664 - Google Ads Linker — paused

3Delete grouped paused tags: 8 × Floodlight Counter, 7 × Custom Image Pixel, 2 × Custom Tag __cvt_2703797_1426, 2 × Custom Tag __cvt_2703797_1514, 2 × Custom Tag __pntr. Find them via Tags → filter Status: Paused → look for those tag types.

4Also resolve the duplicate Google Ads issue: 4 Google Ads-related tags exist (Conversion __awct paused + Linker paused + AdSense enabled + Conversion __html paused). Keep only the enabled AdSense tag; delete the 3 paused ones.

5Publish the container after cleanup.

Google Tag Manager · GTM-P528B328 tags

Facebook Pixel SDK

paused· connect.facebook.net/fbevents.js

DELETE

Snapchat Pixel #1

paused· sc-static.net/scevent.min.js

DELETE

Snapchat Pixel #2

paused· sc-static.net/scevent.min.js

DELETE

TikTok Pixel #1

paused· analytics.tiktok.com/events.js

DELETE

TikTok Pixel #2

paused· analytics.tiktok.com/events.js

DELETE

Script: c.amazon-adsystem.com #1