Skip to content

How we optimized noktah.ai for SEO, AIO and speed

The AI audit we ran on our own site, what it did to the numbers, and the almost-free setup we use to keep tracking them.

This site is the whole public face of noktah. If you're reading this, you found it somehow: a link, a search, or maybe an AI assistant sent you here. We want more of the last two, so we pointed an AI agent at our own site and had it answer three questions.

  1. 01
    : can search engines find the site, read it, and tell what it's about?
  2. 02
    : can AI assistants like ChatGPT, Perplexity and Google's read it, and describe us correctly?
  3. 03
    Performance: do the pages load fast, on every kind of device?

This post covers how we audited the site, what changed, and the almost-free setup we use to keep track of it all. You can copy every piece of it.

The setup

We're open-seo, an open-source alternative to Semrush and Ahrefs, and we've connected it to , and . connects to it through , so we can easily query it for analytics.

We host it on under the free plan. The only metered part is DataForSEO, which charges per request with no subscription. Our instance is locked behind OAuth and MFA, so no, you can't look. Tough luck.

Fortunately, every piece of it is open source, so you can build your own.

We use for local testing, and open-seo runs it too, on mobile and desktop, as part of its site audits.

  1. Measure

    Google Analytics

    Every visit, and the LCP, INP and CLS it reports

    Search Console

    What people searched for, and what they clicked

    DataForSEO

    Rankings, keyword data and Lighthouse runs

  2. Collect

    open-seo

    Self-hosted on Cloudflare. Site audits, and one place to read the rest

  3. Act

    Claude Code

    Reads all of it through MCP, and makes the fix

The audit

We ran the SEO, AIO and performance audits with a Claude Code agent, and measured with Lighthouse. The orchestrator plans the work, hands the pieces to , and puts together what they bring back.

We gave it seven , all open source:

Before and after

We won't bore you with every finding and fix. We'll get right to the juicy numbers.

Performance

96

Median of nine pages on mobile. It was 81 on 22 September.

Largest Contentful Paint

2.4s

▼ 43%vs 22 Sep

Median of the same nine pages, down from 4.2 seconds.

Accessibility

100

On eight of the nine pages, up from 92–96.

Lighthouse performance on mobile

Scores out of 100, before the audit and after it.

  • 22 Sep
  • 24 Sep
Home
Work
Newsroom
About
Contact
Garmin article
Welcome
Privacy
Terms
Source: Lighthouse 13.5 on a laptop against noktah.ai (22 Sep), and open-seo's site audit, which runs Lighthouse through DataForSEO (24 Sep).

Although there's a significant change, SEO is something that takes some time to see real improvements.

On desktop, every page now scores 100 for performance, with the largest paint landing in 0.5–0.6 seconds.

›Accessibility and LCP, page by page
Accessibility22 Sep24 Sep
Home96100
Work96100
Newsroom96100
About92100
Contact96100
Garmin article9393
Welcome96100
Privacy96100
Terms96100
LCP (seconds)22 Sep24 Sep
Home5.45.1
Work3.82.4
Newsroom5.72.5
About5.52.6
Contact4.02.3
Garmin article5.82.3
Welcome4.22.4
Privacy4.02.3
Terms3.92.3
Both measured on mobile. Source: Lighthouse 13.5 on a laptop (22 Sep), then open-seo's site audit (24 Sep).