PageSpeed Intelligence Complete Web Performance Analysis w/AI avatar

PageSpeed Intelligence Complete Web Performance Analysis w/AI

Pricing

from $0.01 / 1,000 results

Go to Apify Store
PageSpeed Intelligence Complete Web Performance Analysis w/AI

PageSpeed Intelligence Complete Web Performance Analysis w/AI

PageSpeed Intelligence combines: PageSpeed Insights: real-user CrUX data and lab metrics Lighthouse audits: performance, accessibility and SEO Tech stack detection: CMS, frameworks, analytics, CDN, hosting Recommendations for your stack: rule-based, or by Claude AI with your own key

Pricing

from $0.01 / 1,000 results

Rating

0.0

(0)

Developer

The Howlers

The Howlers

Maintained by Community

Actor stats

1

Bookmarked

30

Total users

0

Monthly active users

7 days ago

Last modified

Features

Limited permissions

Share

PageSpeed Intelligence

Comprehensive website performance intelligence suite combining Google PageSpeed Insights (real-user CrUX data), Lighthouse audits, Tech Stack detection, and optimization recommendations: rule-based by default, or written by Claude with your own Anthropic API key. Get actionable insights tailored to your specific tech stack.

For real results you need your own free Google PageSpeed Insights API key. The Anthropic API key is optional.


Before You Start

This actor runs in Demo Mode by default, so a default run returns labeled sample data (no key needed) to show you the output format. Set demoMode to false to analyze real sites.

Google PageSpeed Insights key (needed for real results): the key is free. Get one at https://developers.google.com/speed/docs/insights/v5/get-started (click Get a Key) and paste it into pageSpeedApiKey. The actor has no shared key set up yet, and Google currently answers every PageSpeed Insights call without a key with a 429 quota exceeded error. So until a shared key is set up, a run without your own key gets an error row for each URL and device instead of scores. Those rows are not charged. Without a key you can still get tech stack detection and rule-based recommendations: turn off includePageSpeedInsights and includeLighthouse.

Anthropic key (optional): with your own Anthropic API key in anthropicApiKey, Claude writes the recommendations (AI Recommendations (Claude), $0.05 per URL). Without it you get rule-based recommendations (Recommendations, $0.02 per row). Recommendations are charged as one or the other, never both.


Quick Start

Test with Demo Mode (no API key needed)

{
"demoMode": true,
"urls": [
"https://example.com"
],
"url": "https://example.com"
}

Run with real data

{
"demoMode": false,
"urls": [
"https://example.com"
],
"url": "https://example.com",
"strategy": "both",
"includePageSpeedInsights": true,
"includeLighthouse": true,
"includeTechStack": true,
"includeAIRecommendations": true,
"pageSpeedApiKey": "YOUR_API_KEY_HERE",
"anthropicApiKey": "YOUR_API_KEY_HERE"
}

Input Parameters

ParameterTypeDefaultRequiredDescription
demoModebooleantrueNoRun with labeled sample data (no API calls or browser automation). No task events are charged; the actor start fee still applies.
urlsarray-NoList of URLs to analyze comprehensively
urlstring-NoSingle URL to analyze (use this OR urls array)
strategystring"both"NoDevice strategy for analysis
includePageSpeedInsightsbooleantrueNoGet real-user metrics from Chrome UX Report (CrUX)
includeLighthousebooleantrueNoRun detailed Lighthouse performance audits
includeTechStackbooleantrueNoDetect CMS, frameworks, analytics, and hosting
includeAIRecommendationsbooleantrueNoGenerate optimization recommendations for each URL: written by Claude with your Anthropic key, rule-based otherwise (rule-based ones need tech stack detection)
pageSpeedApiKeystring-Yes, for real resultsYour free Google PageSpeed Insights API key. Without it, Google currently answers every call with 429 quota exceeded, until a shared key is set up
anthropicApiKeystring-NoYour own Anthropic API key. With it, Claude writes the recommendations; without it, rule-based recommendations are used
webhookUrlstring-NoURL to POST results when analysis completes (Zapier, Make, n8n, custom endpoint)

Recommendations

  • With your Anthropic API key: for each URL, the actor sends one short request to Claude (claude-sonnet-5) with that URL's audit findings: scores, real-user and lab Core Web Vitals, opportunities, diagnostics and the detected tech stack, for every device analyzed. Claude returns a prioritized list. Each row gets the recommendations for its device in aiRecommendations (text) and aiRecommendationDetails (priority, title, action, reason, device), and aiRecommendationsSource is claude. The tokens are billed to your Anthropic account, and the actor charges AI Recommendations (Claude) once per URL. Your key is used only for these requests and is never written to the dataset or the webhook.
  • Without a key: the actor uses its rule-based recommendations (no AI), which need tech stack detection, and aiRecommendationsSource is rule-based. Each row that gets rule-based recommendations is charged one Recommendations event ($0.02). A row with no recommendation that applies is not charged.
  • When the Anthropic API returns an error (for example a rejected key, no credit, or no access to the model): the row still gets the rule-based recommendations when tech stack detection ran, charged as Recommendations ($0.02 per row) instead of AI Recommendations (Claude), aiRecommendationsError shows the error with its HTTP status and Anthropic's message, and aiRecommendationsNote explains what happened. The run's status message says how many URLs were affected. If the error would repeat for every URL (a rejected key, no credit, or no model access), the actor stops asking Claude for the rest of the run.

Pricing

This actor uses pay-per-event billing:

EventWhen it is chargedPrice
Actor startOnce when the run starts, one event per GB of memory (minimum one). At the default 4 GB of memory a run starts at $0.40$0.10 per GB
URL AnalysisOnce per URL with at least one successful analysis (includes overall scores)$0.05
Strategy AnalysisEach device strategy (mobile or desktop) analyzed per URL$0.02
PageSpeed Insights FetchEach successful call to the Google PageSpeed Insights API (one per URL and strategy)$0.02
Field Data ExtractionEach analysis that returns Chrome UX Report (CrUX) real-user Core Web Vitals, when Include PageSpeed Insights is on$0.01
Lab Data ExtractionEach analysis that returns Lighthouse lab metrics, when Include Lighthouse Audits is on$0.01
Opportunities AnalysisEach analysis that lists at least one optimization opportunity, when Include Lighthouse Audits is on$0.01
Diagnostics AnalysisEach analysis that lists diagnostic audit details, when Include Lighthouse Audits is on$0.01
Tech Stack DetectionEach analysis that ran tech stack detection on the crawled page (CMS, framework, analytics, CDN, hosting)$0.03
Page CrawlBrowser-based page crawl using Playwright for tech detection$0.02
RecommendationsEach row (one URL on one device) that gets rule-based recommendations (no AI), including rows that fall back to them after a Claude error. Not charged for rows Claude wrote$0.02
AI Recommendations (Claude)Once per URL whose recommendations Claude wrote with your own Anthropic API key. Those rows are not also charged Recommendations. The tokens are billed to your Anthropic account$0.05
Webhook DeliveryOnce per run when your webhook URL accepts the results$0.01
ResultEach item written to the default dataset$0.00001

Demo mode charges no task events for sample data. The actor start fee still applies to every run, including demo runs ($0.40 at the default 4 GB of memory).

Failed analyses are not charged. A full analysis of one URL on both mobile and desktop with every option on costs up to $0.35 with rule-based recommendations, or $0.36 with Claude recommendations, plus the actor start.

Maximum total charge: the run delivers only what it can pay for. Set the maximum charge for the run high enough for every URL and device you want; if it is too low, you get fewer results. Before each analysis the actor checks that the rest of the budget covers everything that analysis may deliver, including its recommendations, and analyses it cannot pay for are not started. The last dataset row lists them: skippedUrls has each URL with at least one device not analyzed, and skippedAnalyses has each URL and device. Failed analyses are not in that list; they have their own rows with an error field. A row whose recommendations the run can no longer pay for gets none, and its aiRecommendationsNote says so.


Troubleshooting

"PageSpeed Insights quota exceeded (429)"

Without your own key, Google currently answers every PageSpeed Insights call with this error, so the analysis fails. Those rows are not charged. Add your free key in pageSpeedApiKey (https://developers.google.com/speed/docs/insights/v5/get-started) and run again. With your own key, this error means the key used up its free daily quota.

"Claude recommendations were not available" or aiRecommendationsError

The row has rule-based recommendations instead, charged as Recommendations, and AI Recommendations (Claude) was not charged for that URL. aiRecommendationsError gives the HTTP status and Anthropic's message: 401 means the key was rejected, 400 often means a low credit balance, 404 means the account has no access to the model, and 429 or 5xx are temporary. Check the key and the credit balance in your Anthropic Console.

"API error 403" or "Unauthorized"

Your API key is invalid, expired, or doesn't have access to this specific API endpoint. Double-check your key and account permissions.

"API error 429" or "Rate limit"

Too many requests. Add your own PageSpeed Insights key, wait a minute and try again, or reduce the number of URLs per run.

No results or empty dataset

Check the run log for error messages. Common causes:

  • Invalid input format (check the examples above)
  • API key without proper permissions
  • The target data doesn't exist or is too small to track

How do I test without an API key?

Enable Demo Mode in the input. This returns realistic sample rows, each marked demo: true, so you can verify the output format works for your workflow. No task events are charged; the actor start fee still applies.


Built by John Rippy | Actor Arsenal