Google Updates Lighthouse Loading Tests

Google has updated the Lighthouse version used by PageSpeed Insights, changing how the tool simulates website loading times. As a result, some websites may see higher performance scores even if no changes have been made to their pages.

PageSpeed Insights now runs Lighthouse 13.6.0, released on 27 October 2026. The update includes changes to its loading simulation, which affects how First Contentful Paint (FCP) and Largest Contentful Paint (LCP) are estimated.

Barry Pollard, Chrome’s Web Performance Developer Advocate, confirmed that the updated version is now available in PageSpeed Insights. He explained that the revised measurements should be more accurate and will generally produce higher scores.

Websites that rely heavily on JavaScript may notice a greater difference following the update. This means SEO teams should be careful when interpreting improvements in performance reports, as some changes may be the result of the updated testing method rather than actual website optimisation.

JavaScript-Heavy Websites May See Bigger Improvements

Pollard highlighted reports from developers who had noticed improved Lighthouse results on single-page applications (SPAs), including react.dev.

He agreed that the results were expected but clarified that the update was not designed specifically for SPAs. Instead, websites that depend heavily on JavaScript are more likely to experience changes in their scores.

He also pointed out that these adjustments are separate from Chrome’s experimental soft navigation metrics, which measure user experience using real-world data from Chrome users through the Chrome User Experience Report (CrUX). Lighthouse, by comparison, relies on simulated testing.

What Has Changed in Lighthouse’s Loading Simulation?

Lighthouse normally loads a website at full speed before estimating how it would perform under slower network conditions. This simulated throttling method is also used by PageSpeed Insights.

The Lighthouse 13.6.0 release includes several adjustments to Lantern, the system responsible for these simulated loading estimates. The changes aim to make its performance calculations more accurate.

One important adjustment concerns how Lighthouse identifies resources that prevent a page from rendering. In version 13.5, the simulation used each request’s fetch priority to determine whether it blocked rendering.

Version 13.6 now uses Chrome’s own render-blocking label for each request. Fetch priority is only used as a fallback when that information is unavailable.

Another change affects pages where the Largest Contentful Paint element is text displayed in the same frame as the first visible content.

Previously, Lantern could calculate LCP using a wider range of requests than those used to estimate FCP. The updated version now uses the same requests and page tasks for both measurements in these situations.

Developer Ivan Akulov previously highlighted an issue with the older simulation using a test page built with Framer. The page’s main visible content was text, meaning the same element counted towards both FCP and LCP.

However, Lighthouse reported a much later LCP time. Even though preloading JavaScript modules improved the real page’s interactivity, the simulated test suggested that LCP had become worse.

Pollard credited Akulov for helping highlight some of the problems addressed by the update. The Lighthouse team subsequently closed the issue as completed through the Lantern improvements included in version 13.6.

Lighthouse Updates Its Agentic Resource Discovery Audit

The latest release also changes Lighthouse’s Agentic Resource Discovery (ARD) audit, which was introduced in version 13.5.

The earlier version continued looking for ARD manifests under the older ai-catalog names, despite the specification having moved to ard.json. Websites using the newer file name without an Agentmap entry in their robots.txt file could therefore receive a ‘Not Applicable’ result.

Lighthouse 13.6 now checks for several updated discovery methods before looking for the older names. These include:

An Agentmap entry in the robots.txt file.
A link tag using rel=”ard”.
A Link header.
A manifest located at /.well-known/ard.json.

If Lighthouse only finds the manifest under an older ai-catalog location, it will also explain that ARD consumers are not required to check those locations.

Understanding the Difference Between Lab Scores and Field Data

PageSpeed Insights provides performance information based on both simulated testing and real user experiences.

Its Lighthouse performance score comes from lab data, which is generated using a simulated page load on a single device under fixed network conditions. This allows website owners to test performance in a controlled environment.

However, a strong lab score does not necessarily mean visitors are experiencing the same results.

Field data provides a different picture. It is based on actual visits to a website over the previous 28 days and is collected through Google’s Chrome User Experience Report (CrUX).

PageSpeed Insights displays both types of information in its reports. It also identifies the Lighthouse version used for the test alongside the emulated device details.

Why the Update Matters for SEO Teams

Because PageSpeed Insights now uses a different loading simulation, performance scores generated before and after the Lighthouse 13.6 update are not directly comparable.

A website that receives a higher lab score may not have become faster. The improvement could simply reflect the changes Google has made to its testing process.

This is particularly important for SEO teams tracking performance improvements over time or reporting the results of technical optimisation work.

Before attributing a higher score to website changes, teams should:

Check which Lighthouse version was used in each report.
Make sure the results being compared were generated using the same version.
Review field data to see whether real visitors have experienced an improvement.
Consider the impact of the updated simulation when reviewing performance reports and historical trends.

Checking these details can help prevent teams from reporting improvements that are actually caused by changes to Google’s testing method.

What Happens Next?

Lighthouse’s release notes indicate that version 13.6 is expected to be included in Chrome DevTools with Chrome 159. At the time of the update, Chrome 159 was available through the Dev and Canary channels, with its stable release scheduled for 8 December.

The Lighthouse team is also investigating cases where Lantern’s simulated LCP results are significantly slower than measurements taken from real users.

An open GitHub issue highlights examples involving websites with large JavaScript files and single-page applications. Further work on these areas could lead to additional improvements in how Lighthouse estimates website performance.