PageSpeed Insights now runs Lighthouse 13.6.0, and scores can change without any site change
Search Engine Journal reported on November 2, 2026 that PageSpeed Insights now runs Lighthouse 13.6.0, a version released on October 27. The update changes how Lighthouse simulates slow-connection load times, so a page's lab score can move even though the page itself did not.
PageSpeed Insights switched to a new Lighthouse version
According to Search Engine Journal, PageSpeed Insights (PSI) now runs Lighthouse 13.6.0. The Lighthouse GitHub release page dates that version October 27, 2026. Lighthouse normally loads a page at full speed and then estimates how it would behave on a slower connection; this estimation layer is called Lantern. The 13.6 release notes list several Lantern changes meant to improve accuracy, and SEJ highlights two of them.
Barry Pollard, a Web Performance Developer Advocate on the Chrome team, said the simulated First Contentful Paint (FCP) and Largest Contentful Paint (LCP) are now more accurate and generally produce higher scores. SEJ's practical reading is that a page's score can change while the page stays exactly as it was.
Two notable changes inside the simulation
The first concerns render-blocking requests. Version 13.5 judged them by each request's fetch priority. Version 13.6 uses the render-blocking label Chrome itself attaches to a request and falls back to priority only when that label is absent. The release notes list the matching item as taking the render-blocking status from the protocol.
The second concerns text-based LCP. When the largest element is a piece of text painted in the very frame where the page first shows content, Lantern now derives the LCP estimate from the same requests and tasks it uses for FCP, instead of a wider set that could include every request made before LCP. The notes also list a fix to the simulated LCP breakdown and a change that lets the LCP network record include media requests. Ivan Akulov had shown the old behavior in a GitHub issue, in which a Framer test page with text-only above-the-fold content showed LCP far later than FCP. Pollard thanked Akulov for starting some of these improvements, and Lighthouse maintainers marked that issue done as part of the 13.6 Lantern upgrade. Pollard said JavaScript-heavy sites are more likely to be affected and that the effect is not limited to single-page apps.
A better score is not always better work
For anyone who reports PageSpeed numbers to a client, the consequence is direct: part of a score increase may come from the new Lighthouse version rather than from optimization. Lab scores taken before and after 13.6 were not produced by the same simulation, so a chart of lab scores may show a step at the switch.
SEJ's advice is to compare two reports only when both name the same Lighthouse version, and to look at field data from the Chrome User Experience Report (CrUX), which covers real visits over 28 days. SEJ quotes Google's documentation: "having good lab data does not necessarily mean real-user experiences will also be good". An open Lighthouse issue is also examining cases in which the simulated LCP is much higher than the measured one, with large JavaScript and SPA sites as its examples. Separately, the Agentic Resource Discovery audit now checks newer naming first: an Agentmap line in robots.txt, a Link header or rel="ard" link tag, and the /.well-known/ard.json path, before the older ai-catalog names. Lighthouse 13.6 is expected in Chrome 159 DevTools, and Chrome's schedule lists the stable release of Chrome 159 for December 8.
The sources say nothing about Türkiye, so this section is commentary
Neither the Lighthouse release page nor the SEJ article mentions Türkiye. What follows is therefore not a fact from the sources but UNALSOFT commentary. We have no Türkiye-specific data on how many local sites saw score changes, and this piece does not claim any.
Our reading is practical: if you have been sending a client a PageSpeed screenshot every month, add the Lighthouse version to the report. When a jump appears around this date, check the version before crediting the work, and lean on CrUX field data for the decision.
The UNALSOFT view
We treat a lab score as a diagnostic, not a result. When we report on a site build, the Lighthouse version and the field data sit next to the number, and before/after comparisons are only made on the same version. That keeps an honest improvement separate from a measurement shift. See how we approach this in web design.
Is your speed report comparing like with like?
A short conversation is enough to set up measurement that separates real improvements from version shifts.