PageSpeed Insights: how do you read Google’s report?

What is PageSpeed Insights?

PageSpeed Insights (PSI) is an online tool from Google that measures the experience of a page and suggests ways to improve it. It is one of the basic tools of an SEO audit.

According to Google’s documentation (updated on October 21, 2024), PSI provides lab data, collected in a controlled environment for debugging, and field data, which comes from real users. The lab covers four Lighthouse categories: Performance, Accessibility, Best Practices and SEO.

Since November 2021, the tool has lived at pagespeed.web.dev: on October 19, 2026, the old developers.google.com address redirected there with a 301.

Tool sheet: PageSpeed Insights

An online tool from Google, with nothing to install, to check the performance of a public URL.

• Publisher: Google
• Price: no published price from Google, API included (a key is recommended for automated use)
• Test engine: Lighthouse 13, according to Google’s release notes
• Strengths: field and lab data on the same page, API; limits detailed below
• Access: Open PageSpeed Insights

Information checked on October 19, 2026 in Google’s documentation. Elev8 Lab has no commercial relationship with Google.

How do you run a PageSpeed Insights test?

The PSI release notes (updated on October 27, 2025) describe two useful behaviors: the tool follows 3XX redirects before running Lighthouse, and each analysis produces a link you can share for 30 days.

1. Open the tool: pagespeed.web.dev.
2. Paste the full URL: with https://; if there is a redirect, PSI analyzes the destination page.
3. Read the mobile results first: it is the most demanding simulation.
4. Wait for the simulation to finish: the score and the recommendations come from Lighthouse.
5. Copy the report link: valid for 30 days, it can be sent to a developer.

How do you read the report: field or lab data?

The top part answers “what do my visitors experience?”, the bottom part “why is the page slow?”.

The top part: the Core Web Vitals assessment

PSI shows the CrUX data for the tested URL or, if it lacks data (a recent page or one with few visits), the data for the origin.

Each metric shows a distribution in three bars: green, amber and red. Google’s example: 11% in the amber LCP bar means that 11% of the observed LCP values fall between 2,500 and 4,000 ms. The value shown above the bars is the 75th percentile.

The assessment passes if the 75th percentiles of LCP, INP and CLS are good. Without enough INP data, LCP and CLS are enough. Without enough LCP or CLS data, the page cannot be assessed. The Core Web Vitals thresholds and their weight in rankings are covered in our dedicated guide.

The bottom part: the Lighthouse score out of 100

The Performance score ranges from 0 to 100: 90 or above is good, 50 to 89 needs improvement, below 50 is poor. According to the Lighthouse scoring documentation (Chrome for Developers), it is a weighted average of five metrics.

The recommendations do not change the score: they tell you what to fix. Each value is converted into a score by a curve drawn from real HTTP Archive data; an LCP of about 1,220 ms, for example, scores 99. Since Lighthouse 6, desktop has its own curves.

Insights and diagnostics: what changed with Lighthouse 13

Since October 20, 2025, PSI and its API have used Lighthouse 13, according to Google’s release notes. The post “What’s new in Lighthouse 13” (October 10, 2025) explains that this version replaces a series of performance audits with insights, shared with the Performance panel in Chrome. The audits that did not change stay in the diagnostics.

Other audits disappear with no replacement, including offscreen-images (offscreen images), no-document-write, third-party-facades and font-size (font size, SEO category). For the latter, the Lighthouse team writes that there are no signals that it is still an SEO concern.

Why does PageSpeed Insights show no field data?

Missing field data almost always has a cause documented in the CrUX report methodology (Chrome for Developers, updated on June 20, 2024).

• Recent page or one with few visits: PSI falls back to the origin level. If the origin also lacks data, the top part stays empty.
• Unpublished threshold: Google does not disclose the number of visits required, which is the same for a page and an origin, and you cannot submit a page manually.
• Page that cannot be indexed: a page served with a status other than 200 (after redirects) or marked noindex, by meta tag or X-Robots-Tag header, never enters CrUX. Indexing rules therefore also decide whether you get field data.
• Staging site: password-protected or set to noindex, it meets these exclusion criteria.

CrUX also strips URL parameters (?utm_medium=email) and fragments: all visits to a page are grouped under a single URL.

Why does the score change from one test to the next?

The score varies because conditions change from one test to the next: ads or A/B tests, Internet routing, device. The Lighthouse documentation on variability also lists browser extensions and antivirus software in a local test.

The simulation conditions

PSI emulates a mid-range phone on a mobile network and a desktop computer on a wired connection. The test runs from a Google data center located in North America, Europe or Asia.

Lighthouse default values, read in its code and its throttling documentation on October 19, 2026 (development branch: the version PSI runs may differ):

This network roughly matches the 85th percentile of mobile connections, so it is slower than most of them. On December 5, 2024, Google adjusted the PSI CPU throttling: lab TBT went up on mobile.

Why field data contradicts lab data

A score of 98 with a failed assessment is nothing unusual. The web.dev article “Why lab and field data can be different” (updated on July 18, 2022) gives the causes:

• LCP element: it changes with the screen, personalization, A/B tests or fonts.
• Cache: the lab loads the page with an empty cache and ignores the bfcache (back button), which real visitors benefit from.
• CLS: the lab only counts layout shifts during loading in the visible area; the field counts the whole life of the page.
• INP: it requires a real interaction. TBT serves as an indirect indicator by measuring how much the main thread is blocked during loading, a point covered in our JavaScript SEO guide.

Does the PageSpeed Insights score matter for SEO?

The Lighthouse score is a diagnostic tool. The signal Google describes is the Core Web Vitals: its page “Understanding page experience” (Google Search Central, updated on September 22, 2026) states that its ranking systems use them.

The same page sets two limits: good results in a tool do not guarantee the top spot, and Google shows the most relevant content even when the page experience is poor. Chasing a perfect score “for SEO reasons” is probably not, according to Google, the best use of your time.

PageSpeed Insights, Search Console or Lighthouse: which tool for what?

PSI answers “how good is this URL?”. For everything else, Google offers other tools.

According to the Search Console help, this report relies on the same real-user data, and Google points to PSI for testing a single URL. More options are listed in our selection of free SEO tools.

Testing several pages with the PageSpeed Insights API

The PageSpeed Insights API v5 returns the CrUX and Lighthouse data for a URL in JSON. The guide “Get Started with the PageSpeed Insights API” (updated on August 28, 2025) and its reference list the useful parameters: url (the only required one), strategy (MOBILE or DESKTOP, desktop by default), category (Performance only if nothing is specified) and locale.

curl "https://www.googleapis.com/pagespeedonline/v5/runPagespeed?url=https://www.example.com/&strategy=MOBILE&category=PERFORMANCE&key=YOUR_KEY"

The API works without a key, but a key is recommended for automated requests. On October 19, 2026, two calls without a key received a 429 error: the shared daily quota for keyless calls was used up.

The same guide announces that Google plans to remove real-user CrUX data from the API, and recommends the CrUX API or the CrUX History API for that data.

For a whole site, a crawler such as Screaming Frog queries the API for each crawled URL (licensed version). Recommended method: test a sample of URLs per template.

What are the limits of PageSpeed Insights?

PageSpeed Insights tests one URL at a time and has several blind spots.

• Partial field data: only Chrome users on desktop or Android who have turned on usage statistics and history sync are counted. Chrome on iOS, WebViews and other Chromium browsers, such as Edge, are excluded, in a proportion Google does not publish.
• Fixed test location: you cannot choose the data center.
• Accessibility and SEO: these categories are partial automated checks, not a technical SEO audit.

Performance is only one part of technical SEO, alongside crawling and indexing, which PageSpeed Insights does not measure.