Core Web Vitals (LCP/INP/CLS)

Quick facts

What it is: Core Web Vitals are Google's real-user UX metrics: LCP, INP, and CLS. Chrome UX Report data informs their use as Google Search ranking signals.
How it affects GEO: The effect is indirect and limited. It appears when an AI feature reuses its host search engine's ranking and is nearly absent when it does not.
Effect by engine: The effect is direct for AIO through Google ranking, partial for Bing Copilot, and negligible for ChatGPT Search, Perplexity, and Claude, which have no documented CWV input.
Common misconception: Core Web Vitals measure real-user UX. AI crawler performance concerns TTFB, SSR, and fetch timeouts. The metrics and fixes are different.
SEO baseline: Continue to maintain Core Web Vitals while pursuing GEO, but do not treat them as a new optimization lever.

1. What Core Web Vitals are

Core Web Vitals (CWV) are Google’s metrics for real-user experience. Introduced in 2020 and incorporated into Google Search ranking in 2021, they measure loading with Largest Contentful Paint (LCP), responsiveness with Interaction to Next Paint (INP), and visual stability with Cumulative Layout Shift (CLS). Google defines them this way: “Core Web Vitals are the subset of Web Vitals that apply to all web pages, should be measured by all site owners, and will be surfaced across all Google tools” (Web Vitals overview).

These signals belong to Google, not to the AI ecosystem as a whole. Their field data comes from the Chrome UX Report (CrUX), which includes Chrome users who have opted in through sync. No AI engine has documented an equivalent system that independently weighs LCP or INP. CWV therefore affects visibility in AI answers only when an engine reuses a host search system that already uses these metrics.

Practitioners often conflate three related concepts:

Term	What it measures	Where the signal operates
Core Web Vitals	LCP, INP, and CLS for real Chrome users	Google ranking systems, as part of page experience
Page experience	A broader Google measure that includes CWV, HTTPS, mobile friendliness, and freedom from intrusive interstitials	Google ranking systems, where CWV is the subset with a demonstrated ranking effect
AI crawler performance	TTFB, render path, and fetch timeout for a bot request	Infrastructure for individual bots, separate from Google ranking and CWV

Improving CWV does not make a bot more likely to finish fetching a page within its time limit. That problem requires a different fix; see SSR for AI Crawlers. For AI visibility, CWV matters only through the engines that inherit it from an existing search ranking system.

2. The three current metrics: LCP, INP, and CLS

Metric	What it measures	”Good” threshold (75th percentile)
LCP (Largest Contentful Paint)	Time to render the largest visible element in the viewport; measures loading	≤ 2.5 s
INP (Interaction to Next Paint)	End-to-end latency from a click, tap, or keypress to the next presented frame; measures responsiveness	≤ 200 ms
CLS (Cumulative Layout Shift)	Cumulative score of unexpected viewport shifts over the page’s lifespan; measures visual stability	≤ 0.1

Under Google’s stable lifecycle, these are the current “Good” thresholds. The web.dev pages for each metric list the “Needs Improvement” and “Poor” bands, which Google occasionally revises. For ranking and AI Overviews eligibility, the practical question is whether the 75th-percentile field value falls within the Good band. No additional ranking weight has been documented for optimization beyond Good (§4).

Historical note: INP replaced FID as a Core Web Vital on March 12, 2024 (web.dev announcement, Jan 31, 2024; original Google Search blog, May 2023). FID (First Input Delay) thresholds in older audit tools and articles are historical; INP is the current metric. INP measures the full interaction-to-frame latency across all interactions, while FID measured only the input delay of the first interaction. INP is the stricter and more representative measure.

3. How Core Web Vitals affect each AI engine

CWV affects an AI engine when the engine reuses a host search system that already uses CWV. This produces three distinct levels of effect.

Engine	CWV effect on AI visibility	Why
Google AI Overviews	Direct (via Google ranking)	AIO draws from Google’s classic web index and ranking systems. CWV is a documented Google ranking signal (page experience), and Google explicitly describes optimization for AI features as “still SEO” (AI optimization guide).
Bing Copilot	Partial (via Bing ranking; less documented than AIO)	Bing names page load time among its ranking factors (Bing Webmaster Guidelines) but does not specifically name Core Web Vitals. Copilot’s use of ranked results for grounding is less formally documented than Google’s.
ChatGPT Search · Perplexity · Claude	Negligible / undocumented	These engines make their own retrieval and citation decisions, and no public documentation identifies CWV as an input. For them, the relevant performance variables are on the crawler side: TTFB, SSR readiness, and fetch timeouts (§5).

CWV affects AI visibility when an AI feature inherits a search ranking that already uses CWV. It is not a universal AI-engine signal. Optimizing CWV for AI therefore applies primarily to Google’s ecosystem.

4. How the indirect effect works

On Google, the full causal chain is:

CWV → Google Search ranking signal → page is indexed and ranked for the underlying query → page enters the candidate pool AI Overviews draws from → AIO selects a subset of candidates to ground and cite → your page is (or is not) one of the supporting links.

CWV is one Google ranking signal among many. It is weighted alongside content quality, E-E-A-T, link signals, and topical relevance. Google states: “Beyond Core Web Vitals, other page experience aspects don’t directly help your website rank higher in search results” (page experience). In other words, CWV contributes to ranking, while the other page-experience attributes do not receive additional ranking weight. The signal is real, but its role is limited.

Even on Google, CWV does not directly inform the AIO selection step. Selection from the candidate pool is governed by the quality systems that rank classic results, together with AI-specific factors that make content easier to use in an answer. Once a page is in that pool, citability, including clear sections and quotable claims, E-E-A-T, entity recognition, and Knowledge Graph presence influence whether AIO chooses it. GEO has not increased CWV’s importance at this stage.

That chain does not apply to the other engines. Perplexity, ChatGPT Search, and Claude maintain their own retrieval indexes and choose their own citations. None has documented an equivalent process in which Bing’s page-load-time signal becomes a Perplexity citation factor.

5. Core Web Vitals are not AI crawler performance

CWV measures real users in Chrome through CrUX field data. AI crawlers are not part of that dataset because CrUX eligibility requires a signed-in Chrome user with sync history enabled (CrUX methodology).

Layer	Measures	GEO failure mode	Where to start
Core Web Vitals	LCP, INP, and CLS for real Chrome users in the field	Lower Google rankings can weaken eligibility for the AIO candidate pool, an indirect effect.	Standard web-performance tools such as web.dev and PageSpeed Insights
AI crawler performance	A bot’s TTFB, first-byte HTML completeness, and fetch timeout	If the bot cannot obtain parseable content within its budget, the page cannot be cited on any engine.	SSR for AI Crawlers, plus AI Crawlers for fetch behavior

For AI performance, start with one question: can the bot fetch and parse the HTML within its time limit? The answer is independent of CWV. A site can pass every Core Web Vital yet return a client-side-rendered shell that Perplexity’s fetcher cannot read. Another site can fail INP for human users while returning a fully readable SSR HTML payload in the first response. These are different problems, and their fixes are not interchangeable.

When a performance question arises during a GEO audit, first determine whether it affects human users or crawler fetches. CWV problems require web-performance work; bot-fetch problems require rendering and infrastructure work. A site may have either problem, both, or neither.

6. Three common mistakes

Each mistake starts with a reasonable observation but reaches the wrong conclusion. The first overstates the value of CWV for AI, the second dismisses it, and the third measures the wrong outcome.

Anti-pattern	Why it sounds right	Why it actually fails
”Improve CWV further to increase AI citation rate” (e.g., push LCP from 2.0 s to 1.0 s “for AI”)	CWV is a Google signal, and Google powers AIO, so better CWV appears to imply more AIO visibility.	Once a page meets the Good threshold, further CWV gains receive no additional ranking weight (page experience). AIO selection also depends on factors such as citability and E-E-A-T, which CWV does not affect. The expected AI benefit from improving an already-passing score is close to zero.
”AI engines don’t use user-experience metrics, so CWV doesn’t matter”	ChatGPT Search, Perplexity, and Claude have no documented CWV input.	This ignores the share of AI exposure that comes through Google AI Overviews and Bing Copilot, where CWV can still affect entry into the candidate pool through search ranking. It also overlooks the correlation between CWV and the human outcomes that motivated the work.
”Track CWV as a GEO KPI”	CWV is measurable, easy to audit, and tied to a Google signal.	GEO KPIs measure AI outcomes such as Citation Rate, Citation Share, and Average Position; see GEO Metrics. CWV is a web-performance KPI that contributes to a Google ranking input and, in turn, to one engine’s candidate pool. Treating it as a GEO outcome confuses these distinct stages.

CWV is not a GEO signal. It is a Google ranking signal whose AI relevance is limited to engines that reuse Google’s ranking.

7. What to do

Your intent	First stop
Measure my current CWV	Google PageSpeed Insights, the Search Console Core Web Vitals report, and CrUX field data
Fix LCP / INP / CLS	The standard web-performance resources at web.dev/vitals and the individual metric pages
Fix what the AI bot sees	SSR for AI Crawlers, which addresses the separate crawler-performance problem
Add performance checks to a GEO audit	GEO Audit, with separate checks for CWV and bot-fetch readiness
Understand how AIO inherits CWV	Google AI Overviews §quality-systems
Know which signals actually move AI citation	Citability, E-E-A-T, Entity recognition, and Knowledge Graph presence
Track AI-side outcomes (not CWV) as KPIs	GEO Metrics
Understand the shared baseline for SEO and GEO	SEO vs GEO, which presents CWV as a baseline requirement rather than a new GEO lever

Continue to maintain CWV as a GEO baseline, but do not treat it as a separate optimization lever. It still matters wherever Google ranking matters, and the goal is to meet the published Good threshold rather than pursue aggressive over-optimization. Once a site meets that threshold, additional CWV work is not the best use of GEO effort.

References

Primary: Google web performance and AI documentation

Google web.dev: Web Vitals (overview) · INP: Interaction to Next Paint
Google Search Central: Understanding page experience in Google Search results · AI features and your website · Optimizing for generative AI features
Google: INP becomes a Core Web Vital, March 12, 2024 (web.dev) · Introducing INP to Core Web Vitals (Search Central, May 2023)

Primary: other engines and data sources

Microsoft Bing: Bing Webmaster Guidelines (page load time is named among its ranking factors; CWV is not specifically named)
Google Chrome for Developers: Chrome UX Report (CrUX) methodology (the data includes only real, opted-in Chrome users; it does not include bots)

Secondary: independent measurement and coverage

HTTP Archive: Web Almanac 2024: Performance chapter (industry CWV pass rates)
Google web.dev: Rakuten 24 case study: Core Web Vitals and business KPIs (one published example of the correlation between CWV and business KPIs)
Search Engine Land: Google Core Web Vitals to add Interaction to Next Paint on March 12 (independent coverage of the transition from FID to INP)

Frequently asked questions

Do AI engines use Core Web Vitals to decide what to cite?

Only indirectly, and only when an AI feature reuses its host search engine's ranking. Google AI Overviews inherits CWV's ranking weight because it draws from the same index and quality systems as classic Google Search (AI optimization guide). Bing Copilot partly inherits the effect through Bing's ranking, which names page load time as a factor. ChatGPT Search, Perplexity, and Claude use their own retrieval and citation logic, with no publicly documented CWV input. For those engines, a server's ability to complete the bot's fetch in time and return parseable HTML matters more than its LCP score.

If I improve LCP from 2.0 s to 1.0 s, will my AI citation rate go up?

Almost certainly not as a direct effect. Once LCP meets Google's 'Good' threshold of ≤ 2.5 s, further optimization has diminishing returns for Google ranking and no documented marginal effect on AI Overviews selection. Google states that 'beyond Core Web Vitals, other page experience aspects don't directly help your website rank higher in search results' (page experience). For AI citations, citability, genuine E-E-A-T, and clear entities are more productive investments than fine-tuning performance metrics that already pass.

Should I stop caring about page speed for AI search?

No. Dismissing page speed is as mistaken as over-optimizing it for AI. CWV remains a Google ranking signal, so it matters for the share of AI exposure that comes through Google AI Overviews. It also correlates with human outcomes such as conversion and retention, regardless of AI. What it does not do is earn extra weight as an AI-engine signal in citation decisions.

Are AI crawler fetch timeouts the same as INP or LCP?

No. CWV measures real Chrome users through the Chrome UX Report (CrUX methodology). Bots are excluded because eligibility requires a signed-in Chrome user with sync history enabled. AI crawlers have their own fetch budgets, typically a few seconds, and depend on TTFB and server-rendered HTML that they can parse from the first response. Improving CWV does not automatically solve this bot-side problem. SSR readiness and TTFB require separate work; see SSR for AI Crawlers.

Did INP replace FID?

Yes. INP replaced FID on March 12, 2024. Google announced the transition in January 2024 (web.dev announcement) after a roughly ten-month preview (Search Central, May 2023). INP has been the third Core Web Vital ever since. FID thresholds in older audit tools or articles are historical. The current metric is INP (Interaction to Next Paint), with a Good threshold of ≤ 200 ms.

Sources

Primary

Web Vitals (overview, definition, current metrics) · Google (web.dev)
Interaction to Next Paint (INP) · Google (web.dev)
Understanding page experience in Google Search results · Google Search Central
AI features and your website · Google Search Central
Optimizing for generative AI features in Google Search · Google Search Central · 2026-05-15
Advancing Interaction to Next Paint (INP becomes a Core Web Vital, March 12, 2024) · Google (web.dev) · 2024-01-31
Introducing INP to Core Web Vitals · Google Search Central · 2023-05-10
Bing Webmaster Guidelines · Microsoft Bing
Chrome UX Report (CrUX) — methodology · Google (Chrome for Developers)

Secondary

Web Almanac 2024 — Performance chapter · HTTP Archive
Rakuten 24 case study — Core Web Vitals and business KPIs · Google (web.dev)
Google Core Web Vitals to add Interaction to Next Paint on March 12 · Search Engine Land