PageSpeed and Pharos: two readings of the same site

Run PageSpeed Insights on your site and the score is green: it loads fast, it works well on a phone, the Core Web Vitals are fine. You relax, rightly so. Then a client tells you they asked ChatGPT for a supplier like you, and your name did not come up. Two opposite verdicts on the same site, and neither is wrong: they are measuring two different things.

PageSpeed and Pharos read the same address to answer different questions. One looks at what the site is like for people. The other looks at whether an AI can read it, reach it and cite you. They serve different questions, and today you need both.

What PageSpeed measures

PageSpeed Insights, Google's free tool, measures the experience of someone opening the site in a browser. How long it takes to load, how stable it stays while the page assembles, how responsive it is to the first tap: these are the Core Web Vitals, plus a list of technical tips to go faster. It is a snapshot of how fast and comfortable the site is for a person, and it is something that matters and will keep mattering.

What Pharos measures

Pharos looks at your site's other audience: crawlers and answer engines (ChatGPT, Perplexity, Google's AI summaries). They do not open a browser and do not wait for the JavaScript: they read the HTML the server sends, as it is. Pharos imitates them, does not run the code, and shows you what a machine finds in your place.

• Crawler access: whether robots.txt and the server actually let GPTBot, ClaudeBot, PerplexityBot and the others through.
• Server-side readability: whether the content is already in the HTML or only appears after the JavaScript, invisible to anything that does not run it.
• Structured data: title, meta, JSON-LD, Open Graph, the labels that tell a machine who you are.
• Files for agents: sitemap, llms.txt, the emerging signals under .well-known, and how fresh what you publish is.
• Off-site presence: whether you show up in the sources many AIs draw from, like Common Crawl.

The report is free, repeatable and honest: a score, the two or three things that weigh most, the concrete actions with the evidence underneath. When a check does not apply to your site it marks it as such, without counting it as a failure (N/A is not a failure). How it works in detail is in the Pharos documentation.

PageSpeed is starting to look at AIs too

Until recently they were two separate worlds: PageSpeed for speed, other tools for AIs. Now Google is moving on that side too, but it is early days. In Search Console a report has appeared that counts how many times your pages showed up inside Google's AI answers, AI Overviews and AI Mode: for now it measures impressions only, not clicks, and it is rolling out to a subset of sites at a time.

And Lighthouse, the engine behind PageSpeed, has added an experimental category, 'Agentic Browsing', with checks meant for agents: a valid llms.txt at the root, a clean accessibility tree, a stable layout, WebMCP signals. It does not give a 0 to 100 score yet: it just shows how many checks you pass, because the standards for the agentic web are still moving.

These are essential steps, and they look at a different slice: more whether you appear than whether you are readable. But the direction is clear. If even Google is starting to ask whether an AI understands you, the question has become everyone's. Pharos was built to answer exactly that, and in full.

PageSpeed's little brother

Pharos does not hide where it took the model from: you paste an address, an analysis runs in the background, you come back when the report is ready, just like PageSpeed. It is younger and does one thing, but it does it thoroughly. The same spirit as PageSpeed, applied to another question.

That is why they complete each other. PageSpeed tells you whether a person opens the site without getting impatient. Pharos tells you whether a machine reads it well enough to cite you. Two readings of the same site, for two audiences that both count today.

How to use them together, in practice

You do not have to choose. A healthy round is short:

• Run PageSpeed for speed and experience: start from the Core Web Vitals and the tips at the top.
• Run Pharos for AI access and readability: look first at whether you are blocking crawlers in robots.txt, often it is one line to remove.
• Keep both as a repeatable check: run them again after every significant change to the site, not just once.
• Share the two reports with whoever works on the site: they are already ordered by impact.

Neither replaces the other: PageSpeed will never tell you whether an AI reads you, and Pharos does not measure how fast you are. Together they cover the site as it is seen today, by people and by machines. Try Pharos on your address, the report is open and email-free; and if you first want to understand why a machine reads your site differently from you, we told that story in If an AI can't read you, it won't cite you.