Until March 27, 2026, Google evaluated Core Web Vitals page by page. A slow article from 2019 on a legacy Meredith template hurt that page's ranking. Nothing else.
That changed. The March 2026 core update shifted CWV to site-wide aggregate scoring. Google now rolls up performance data across your entire domain. If more than 25% of your URLs fall into "Poor" or "Needs Improvement" on any single metric, every page on that domain is suppressed.
For a publisher with three pages, this is a non-event. For People Inc — 40+ brands, millions of legacy URLs, templates spanning three CMS generations — it is the most consequential technical SEO change since the Page Experience update.
The Old Math vs. The New Math
The March update also equalized the three metrics. Previously LCP dominated the ranking signal. Now LCP, INP, and CLS are weighted equally. And LCP's "good" threshold tightened from 2.5 seconds to 2.0 seconds.
Early data from tracking tools shows affected sites lost 20–35% of traffic, with some domains losing over 50% on their worst-hit sections. Sites with 500+ pages and inconsistent templates face the largest recalibration.
People Inc has tens of thousands of pages across inconsistent templates. This is their problem at industrial scale.
Where the Weight Lives
An ad-heavy publisher page loads three categories of performance-taxing scripts:
| Script Layer | What It Does | INP Impact |
|---|---|---|
| CMP / Consent Banner | GDPR/CCPA compliance. Fires on every page load. Triggers third-party callback cascade on consent. | Major. Unoptimized CMPs push INP past 400ms. PubTech documented 470ms→230ms after optimization. |
| Contextual Intelligence | D/Cipher: real-time content analysis, intent classification, embedding generation. Runs on every page. | Moderate-to-high. Client-side inference adds main-thread work. No public performance data. |
| Ad Delivery & Measurement | Header bidding, viewability tracking, conversion pixels, analytics. Multiple vendors per page. | Cumulative. Each script small; together they saturate the main thread. Only 37% of mobile sites with analytics + ad scripts achieve "good" INP. |
| Legacy Template Code | Old Meredith templates. jQuery patterns, synchronous DOM manipulation, unminified CSS. | Structural. LCP and CLS failures baked into layout. Cannot be fixed by script optimization alone. |
The first three layers compound. A consent interaction fires the CMP, which unlocks D/Cipher and ad scripts, which collectively block the main thread for the duration of the user's first real interaction. Under per-URL scoring, this only mattered on high-traffic pages. Under site-wide scoring, it matters on every page that CrUX measures.
As of the May 2026 CrUX release, 55.9% of all tracked origins pass all three Core Web Vitals. INP pass rate is 86.6% — the highest of the three. But that aggregate masks the publisher reality: only 37% of mobile sites running analytics and ad scripts achieve "good" INP. The scripts that make the money break the metric that controls visibility.
The People Inc Problem, Specifically
People Inc is not a generic publisher with a CWV problem. People Inc is a company that:
1. Operates 40+ brands across at least three template generations (legacy Meredith, transitional Dotdash, current unified)
2. Runs D/Cipher's contextual intelligence scripts on every page — the differentiator that generates 126% higher conversion rates and powers half of direct digital revenue
3. Has a dedicated performance engineering team — Russell Brown presented Dotdash Meredith's INP approach at the NY Web Performance Meetup in 2024
4. Is experiencing a 17% session decline while D/Cipher+ expands off-platform to 770+ domains — the same scripts causing on-site INP pressure are being deployed more widely
The performance engineering team exists. They know the problem. But under per-URL scoring, they could triage — fix the top revenue pages, defer the long tail. Under site-wide scoring, the long tail IS the problem. Ten thousand legacy URLs that nobody visits individually can collectively suppress a domain that millions visit daily.
The D/Cipher Paradox
This is the tension at the center of the problem. D/Cipher is People Inc's most valuable technology. It is also client-side weight on every page.
Every page that D/Cipher analyzes contributes to the site-wide CWV aggregate. The contextual intelligence that drives 126% conversion lift and 50% of direct revenue also adds main-thread processing time to every URL that CrUX samples. If legacy archives already push the 25% failure threshold, D/Cipher scripts compound the problem on pages that were marginal.
This is not a D/Cipher bug. It is a structural tension between two performance definitions: ad performance (which D/Cipher maximizes) and page performance (which Google now evaluates domain-wide). The March 2026 update made these two definitions directly competitive.
The resolution is not to remove D/Cipher. The resolution is to ensure that the denominator — total page performance across the domain — is healthy enough that D/Cipher's contribution does not push the aggregate past the threshold.
What the Engineering Org Should Do
The Crawler Problem, Continued
This connects directly to The Crawler Problem. That post was about structured data — making content machine-readable so AI systems cite it. This post is about the prerequisite: making sure Google's crawlers and performance evaluators even rank the content high enough for AI systems to find it.
Site-wide CWV scoring and AI Overviews are two expressions of the same shift. Google is raising the floor on what counts as a quality publisher. Structured data without performance is metadata on a page nobody sees. Performance without structured data is a fast page that AI systems cannot cite. People Inc needs both.
The performance engineering team has the skills. They presented on INP optimization two years ago. The question is whether, under the new site-wide scoring regime, they have the mandate and budget to remediate at the scale the problem now demands — not fifty pages, but fifty thousand.