Pillar one of six
Engineered for Lighthouse 100
Lighthouse is Google's public examiner for websites. It scores every page out of 100 for Performance, Accessibility, Best Practices, and SEO. AutoHive engineers every page to score 100 in all four, on a phone, before launch.
Think of it as the building inspector for your digital home. Most builds hope to pass. An agentic build is constructed to the inspection, then inspected again for the rest of its life.
Watch the inspection climb. Red is danger, orange is almost there, green from 90 is the standard, and 100 is where we stop.
Digitally Driven Growth, Naturally
The standard
Why does 100 matter when 90 already counts as good?
Because the last ten points are where the discipline lives. Lighthouse marks 90 and above as good, and every point past 90 costs more engineering than the point before it. A page that scores 100 was built with intent. A page that scores 90 got lucky somewhere.
The score is a lab measurement, a controlled test on a simulated phone and network. Real visitors then confirm it in the field through Core Web Vitals: the biggest element paints within 2.5 seconds, the page answers a tap within 200 milliseconds, and nothing jumps by more than 0.1 while loading.1 Google's own tooling has also started examining how readable a site is to AI agents, which is the agentic web arriving inside the inspector itself.3
See it, do not take our word
How is the Performance score actually calculated?
Five timed measurements, each with a fixed weight. Move the sliders and watch the score respond the way Google's own configuration weighs it.2
The score studio
Run your own audit free at pagespeed.web.dev, then set each slider to the scores in your report. The diagnosis below reads the pattern and names what is holding you back.
Performance 100
Honesty note: real Lighthouse converts raw milliseconds through statistical curves before weighting. The studio shows the weighting itself, which is the part that decides where engineering effort goes. Notice that script blocking time carries the single heaviest weight, which is why lean, deferred code wins.
The race every visitor runs
Two pages, one connection. Choose the network your customer is on, then start the race. The figures are illustrative and deliberately kind to the heavy page.
Agentic page, about 0.3 MB
Typical heavy builder page, about 3 MB, illustrative
The lean page is not faster because of a trick. It is faster because it ships less: fewer scripts, smaller images, no third party baggage. Every megabyte is a toll your visitor pays on their own data.
Feel a layout shift
You have lived this: you go to tap a link and the page jumps under your finger. That jump is Cumulative Layout Shift. Toggle the two builds below.
You are reading the paragraph you came for. It holds still when the page reserves space for everything that arrives late, and it leaps away when the page does not. A shift below 0.1 is the standard.1 An agentic build declares the size of every image and every late element in advance, so the page you see first is the page you keep.
What lands in your hands
What does the certificate actually look like?
Three pages, issued at handover: the certificate face below, an AI readiness register showing the evidence for each optimisation layer, and a verification register listing every one of the 49 automated checks with its pass. This sample is a real issue with the client's identity removed under POPIA.
This certifies that the website of
SAMPLE COMPANY (PTY) LTD
Somerset West, Western Cape, South Africa
PASSED ALL 49 OF 49 STATIC COMPLIANCE VERIFICATIONS
Scores reflect the engineered build in which every statically verifiable Lighthouse audit passes. Live scores are produced by Google against the deployed URL and depend on hosting response times, confirmed after go live at pagespeed.web.dev, with Performance certified to the 95 to 100 band.
AI READINESS, VERIFIED IMPLEMENTATION
Pages two and three carry the AI readiness register and the full 49 check verification register, each line marked PASS with its evidence.
Every term, explained
What do all these terms mean?
Plain definitions, no jargon left standing. Open any term.
Lighthouse
Google's open source auditing tool. It loads your page in a controlled test and scores it out of 100 in four categories: Performance, Accessibility, Best Practices, and SEO. Anyone can run it, which is why it works as a public standard.2
Performance
How fast the page loads and settles. Scored from five timed measurements with fixed weights: First Contentful Paint 10, Speed Index 10, Largest Contentful Paint 25, Total Blocking Time 30, Cumulative Layout Shift 25.2
Accessibility
Whether everyone can use the page, including people with screen readers, keyboards only, or low vision. Contrast, labels, focus, and structure are all examined. Accessibility is also machine readability, which is why it matters to AI agents too.
Best Practices
The engineering hygiene category: secure connections, no browser errors, correct image ratios, and safe third party behaviour. Each audit carries roughly equal weight.2
SEO, the Lighthouse category
The technical floor for being found: crawlable links, valid metadata, legible font sizes, and a page search engines can parse. It is the entry ticket, not the whole game. The wider game is the second pillar, answer engine readiness.
Core Web Vitals
Google's three experience measurements taken from real visitors: Largest Contentful Paint within 2.5 seconds, Interaction to Next Paint within 200 milliseconds, and Cumulative Layout Shift below 0.1.1
Largest Contentful Paint, LCP
The moment the biggest visible thing, usually your hero image or headline, finishes rendering. Good is within 2.5 seconds.1 It is the metric your visitor experiences as the site being there.
Interaction to Next Paint, INP
How quickly the screen responds after a tap or click, measured across the whole visit. Good is within 200 milliseconds.1 INP is measured in the field from real visitors, which is why the lab Performance score weights it at zero and uses Total Blocking Time as its lab stand in.2
Cumulative Layout Shift, CLS
A score for how much the page jumps while loading. Good is below 0.1.1 Shifts happen when images, adverts, or fonts arrive without reserved space. The cure is declaring sizes in advance.
First Contentful Paint, FCP
The time until the first text or image appears. It is the visitor's first proof that something is happening.
Total Blocking Time, TBT
How long scripts freeze the page between first paint and full interactivity. It carries the heaviest weight in the lab score, 30 of 100, because a frozen page feels broken even when it looks loaded.2
Speed Index
How quickly the visible part of the page fills in. Two pages can finish at the same time, yet the one that paints progressively feels faster, and Speed Index rewards it.
Lab data and field data
Lab data comes from a controlled test, which is what Lighthouse produces and what we certify before launch. Field data comes from your real visitors on real networks, which is what Google collects afterwards. An agentic build is engineered in the lab and then confirmed in the field.
Agent accessibility
The newest frontier: whether AI agents can read and act on your site. Google's Lighthouse source now carries an agent accessibility group, including a check for an llms.txt file.3 The inspector itself has started asking the agentic question.
Sources, triple verified
Where does this page get its facts?
Each source passed three checks on 21 July 2026: it comes from the supplier, it was retrieved and read in full, and it is current.
- Google Chrome team. web-vitals, the official Core Web Vitals measurement library. Thresholds LCP 2.5 s, INP 200 ms, CLS 0.1 read from the repository at github.com/GoogleChrome/web-vitals. Latest release 28 May 2026.
- Google Chrome team. Lighthouse source configuration, core/config/default-config.js and docs/scoring.md at github.com/GoogleChrome/lighthouse, repository updated 20 July 2026. Performance weights read directly from source: FCP 10, SI 10, LCP 25, TBT 30, CLS 25, INP weighted 0 in the lab score.
- Google Chrome team. Lighthouse agent accessibility audit group including the llms-txt check, read from core/config/default-config.js at github.com/GoogleChrome/lighthouse, repository updated 20 July 2026.
The illustrative page sizes in the race carry no citation because they are illustrations, and the page says so where they appear. Field data references from the Chrome UX Report will be added under the same standard before publication.
The next step is small
Where does your site stand today?
Choose two answers. The audit shapes itself around you, and the first finding usually arrives in the first conversation.
Your site today
What matters most
The audit is free, and the findings are yours to keep, whoever you build with next. This page sets no cookies, loads no third party scripts, and tracks nothing. That is the standard, demonstrated.