Compare
Aiscovery vs Lighthouse and SEO audits
Lighthouse, PageSpeed Insights and SEO crawlers check how pages are built. We check whether agents finish the task.
The short answer
Lighthouse, PageSpeed Insights and SEO crawlers such as Screaming Frog check how a page is built: speed, accessibility, SEO issues and, in Lighthouse's experimental Agentic Browsing category, signals such as WebMCP forms, llms.txt and the accessibility tree. Aiscovery has named AI agents try real shopping tasks on your store and checks whether they got the right answer. Audits find build issues page by page; agent tests show whether a task works end to end. They work well together.
What each one is for
In each product's own words, paraphrased. Follow the source link to read more.
- AI agent testing for online storesAiscoveryAn automatic site check of your public pages, then named AI agents (OpenAI GPT-6 Luna and Google Gemini) trying key shopping tasks in a read-only browser. Literal checks decide each result, and you get step screenshots, a ranked fix list and retests. AI visibility tracking is coming later.How Aiscovery works
- Open-source page auditsLighthouseGoogle's open-source, automated tool for improving page quality, with Performance, Accessibility, Best Practices and SEO categories, run from Chrome DevTools, the command line, Node or PageSpeed Insights. Its Agentic Browsing category, labelled experimental and based on proposed standards, runs deterministic audits for WebMCP tools and forms, llms.txt, accessibility for agents and layout stability, and reports how many checks pass.Source: developer.chrome.com
- Page experience reportPageSpeed InsightsReports a page's user experience on mobile and desktop: field data from real Chrome users over the previous 28 days, Lighthouse lab data and a Core Web Vitals assessment, with suggestions to improve.Source: developers.google.com
- Technical SEO crawlerScreaming Frog SEO SpiderA desktop website crawler for technical SEO audits: broken links, redirects, titles and meta descriptions, structured data validation and JavaScript rendering. It can crawl as AI crawler user agents and run AI prompts during a crawl.Source: screamingfrog.co.uk
Side by side
Not their focus means their site doesn't describe it, not that they can't do it.
| Capability | Aiscovery | Lighthouse | PageSpeed Insights | Screaming Frog |
|---|---|---|---|---|
| Performance and Core Web Vitals | Partly: Layout shift per agent step only | Yes: Lab data | Yes: Field and lab data | Partly: Through the PageSpeed Insights API |
| Agent-facing signals: WebMCP, llms.txt | Yes: llms.txt and declared WebMCP forms | Yes: Agentic Browsing, experimental | Not listed | Not their focus |
| Technical SEO checks | Partly: Reachability, redirects, page basics | Partly: SEO category, per page | Partly: SEO category, per page | Yes: Site-wide crawl |
| AI crawler rules in robots.txt | Yes: Declared rules per AI company's crawler | Not their focus | Not their focus | Yes: Crawls as any AI user agent and tests robots.txt for it |
| Named AI agents try shopping tasks | Yes: Available, in a read-only browser | Not their focus: Deterministic checks, no agent run | Not their focus | Not their focus |
| What you get | Yes: Verdict, ranked fixes, retests | Yes: Category scores; agentic checks passed | Yes: Scores and Core Web Vitals assessment | Yes: Issues found in the crawl |
- Not their focus
- Their site doesn't describe it.
- Not listed
- Their docs don't list it.
- Coming later
- Ours, not running yet.
How they relate
Where they overlap
- Lighthouse's Agentic Browsing category and our site check both look at llms.txt and declared WebMCP forms. Lighthouse also checks the accessibility tree agents rely on and layout stability.
- Our site check covers reachability, redirects, page basics and structured data syntax, which SEO crawlers cover in more depth across a whole site.
- Layout shift matters to both: Lighthouse measures it in a lab page load, and our test runs measure it during each agent step.
Where Aiscovery is different
- Audits are deterministic checks on a page. They tell you a signal is present or a metric is in range. Lighthouse's own docs describe the agentic category as checks, not an AI agent trying a task.
- Aiscovery runs the task: a named AI agent reads your product pages, follows links and uses search and filters, and literal checks decide whether it found the right price, stock or delivery facts.
- A failed task is attributed to a website, agent or platform cause and ranked in a fix list, with a retest to confirm the fix. We give a plain verdict, not a single score, and we don't measure performance scores.
When to use both
- Run Lighthouse in your build pipeline to catch regressions in speed, accessibility and agent-facing signals on each page.
- Use an SEO crawler for site-wide technical SEO such as broken links and redirect chains.
- Use Aiscovery to check that AI agents can complete the shopping tasks those pages exist for, and to confirm fixes with a retest.
Sources
Each vendor's own pages, read on October 21, 2026. Products change; if something here is out of date, tell us at support@aiscovery.io.
- Introduction to LighthouseChrome for Developers
- Lighthouse agentic browsing scoringChrome for Developers
- Accessibility for agentsChrome for Developers
- Layout stabilityChrome for Developers
- About PageSpeed InsightsGoogle for Developers
- Screaming Frog SEO Spider Website CrawlerScreaming Frog
Related
- CompareAI agent testing vs AI visibility toolsPeec AI, Profound, Otterly.AI and Semrush track what AI answers say about you. We test what agents can do on your store.
- CompareAiscovery vs Playwright and QA automationPlaywright and QA Wolf test your own release. We show what outside AI agents experience on your live store.
- Agent-facing technologyWhat the site check detects, such as llms.txt, structured data and WebMCP.
- Our checker, AiscoveryAuditWhat it reads, how fast it goes and how to allow or block it.
- How a test run worksA named AI agent tries a real task, with step screenshots and checked outcomes.
Questions
Lighthouse has an Agentic Browsing category. Isn't that agent testing?
It is a set of deterministic checks for signals agents use, such as WebMCP forms, llms.txt, the accessibility tree and layout stability. Google labels it experimental. It reports how many checks pass. It doesn't have an AI agent try a task on your store, which is what Aiscovery does.
If our Lighthouse scores are good, will agents succeed?
Not necessarily. A page can pass every audit and still hide a price behind a control agents can't open, or show delivery facts only after a step agents don't take. Agent tests catch that because they check the answer the agent found.
Does Aiscovery check performance?
Not as a score. Our test runs record step timings and layout shifts during each agent step, because shifting content makes agents miss buttons. For speed and Core Web Vitals, use Lighthouse and PageSpeed Insights.
How does your checker read our site?
As one identified visitor, AiscoveryAudit, at most one page per second, following robots.txt. Our checker page explains how to allow or block it.
See what AI agents do in your store.
A short call: we run key shopping tasks with named AI agents on your public pages and walk you through the fixes.
Have a real workflow to test now? Apply as a launch partner