DEV Community: ArminNX

I Built a Website Auditor That Doesn't Just Tell You What's Wrong, It Fixes It

ArminNX — Thu, 13 Aug 2026 13:22:45 +0000

I've used a lot of website audit tools.

They all have the same basic workflow:

Scan website → generate 47 problems → hand the problems back to you → good luck.

And honestly, I got tired of it.

So I built something different.

Meet OptiQra.

The problem

A typical website audit might tell you:

Your title is too long
17 images are missing alt text
You have broken links
Your page isn't accessible enough
Your Core Web Vitals aren't great
Your content isn't optimized for search

Cool.

Now what?

I'm a developer, so I can go into the code and fix these things.

But even then, there's a lot of annoying manual work involved.

And for someone who isn't a developer, the report basically becomes a giant TODO list.

I wanted the audit tool to actually do something about the problems it found.

So I built one

OptiQra is an open-source website optimization tool that:

Crawls your website
Finds problems
Explains what they mean
Fixes many of them
Lets you verify the results

It also looks beyond traditional SEO.

SEO

It checks things like:

Meta titles and descriptions
Canonicals
Open Graph
Structured data
Sitemaps
robots.txt
Internal links
Broken links
Duplicate content
Images
Page structure

Accessibility

It looks for things such as:

Missing alt text
Missing labels
ARIA problems
Semantic HTML issues
Keyboard accessibility problems
Contrast issues

Performance

It can analyze performance characteristics and optionally pull real PageSpeed Insights data.

And then there's the interesting part: AI visibility

Search isn't the only place people discover websites anymore.

People increasingly ask ChatGPT, Claude, Perplexity and other AI systems questions instead of searching manually.

So I added GEO and AEO audits.

OptiQra checks things like:

Whether AI crawlers are blocked
Whether important content is hidden behind client-side JavaScript
Whether pages contain citation-friendly content
Whether your site has useful entity signals
Whether content can be easily extracted into answers
Whether your content is structured around questions people actually ask

I'm still experimenting with exactly how useful all of these signals are, but I think this is going to become an increasingly important part of website optimization.

But finding problems wasn't enough

This was the part that annoyed me most about existing tools.

If a tool tells me:

"Your page is missing a meta description."

I don't really need a tool to tell me that.

I need it to fix it.

So OptiQra has several layers of fixing.

Deterministic fixes

For things that don't require AI, OptiQra fixes them deterministically.

For example:

Missing attribute
      ↓
Detect
      ↓
Generate deterministic change
      ↓
Apply
      ↓
Verify

No AI hallucination required.

AI-assisted fixes

For things that actually require writing, you can connect your own AI provider.

OptiQra supports providers including:

OpenAI
Anthropic
Google
Groq
OpenRouter
Mistral
DeepSeek
xAI

You bring your own API key.

OptiQra doesn't need its own AI subscription.

Full AI Fix

There's also a more aggressive mode.

Instead of asking an AI model to fix one tiny issue at a time, OptiQra can give it an entire file along with the detected problems and ask it to return the corrected file.

Obviously, handing an AI an entire source file and letting it rewrite things needs safeguards.

So I added integrity checks that verify the result before accepting the modification.

If the result violates structural invariants, the change gets discarded.

It can audit your codebase too

This is another feature I really wanted.

A live website crawler can only see the final website.

It can't see:

next.config.js
robots configuration
source files
framework configuration
build configuration

So OptiQra can also audit an actual project.

You can upload a project archive and have it analyze the codebase itself.

That means it can find problems that aren't necessarily visible from the rendered website and make framework-aware fixes.

It currently understands things like:

HTML
JSX / TSX
Vue
Nuxt
Angular
Svelte
Astro

I also went a little overboard

At some point I decided a normal crawl tree wasn't interesting enough.

So OptiQra has an interactive 3D / "galaxy" visualization of your website.

Instead of:

Home
├── About
├── Blog
│   ├── Post 1
│   └── Post 2
└── Contact

you can explore the site visually.

Is this necessary for SEO?

Absolutely not.

Is it fun?

Yes.

It runs locally too

OptiQra is a web app, but there's also a native desktop version built with Tauri.

The desktop version is useful for things like:

Offline project audits
Background scheduled scans
Native notifications
Running scheduled tasks after the window is closed

The audit engine is shared between the web and desktop versions.

Everything is open source

The whole thing is MIT licensed.

You can inspect the code, run it yourself, modify it, or contribute.

There are no accounts required for the basic web app.

And AI features use your own API keys rather than forcing you into a proprietary subscription.

So... is it actually useful?

That's the part I'm still figuring out.

I built OptiQra because I wanted this workflow:

      Website
         ↓
       Crawl
         ↓
       Audit
         ↓
    Find problems
         ↓
      Fix them
         ↓
      Verify
         ↓
    Scan again

Instead of:

Website
   ↓
Audit
   ↓
47 warnings
   ↓
¯\_(ツ)_/¯

I'm not claiming I've solved website optimization.

There are already extremely mature tools in this space.

I'm trying to build something that feels more like a developer tool than a report generator.

And honestly, I'd like to know whether that's actually useful to other developers.

Try it

If you want to play with it, I put a live version here:

https://optiqra.vercel.app/

No signup required.

The project is also on GitHub:

https://github.com/armin5872/OptiQra

If you try it, I'd genuinely like to know:

What would make you actually keep using a tool like this?

Because that's the part I'm working on next.

SEO Is Becoming More Than Just Search Engines

ArminNX — Tue, 11 Aug 2026 13:46:23 +0000

For a long time, SEO was mostly about one thing:

Getting your website to rank on Google.

You researched keywords, optimized your pages, fixed technical issues, built links, watched your rankings, and tried to get more clicks.

That model still matters.

But the way people discover information is changing.

Someone looking for an answer today might not open Google, scan ten blue links, and choose one.

They might ask ChatGPT.

They might use Google's AI features.

They might search through Perplexity.

They might ask an AI assistant to compare products, explain a topic, or recommend a service.

And that creates a slightly different question:

What happens when your website needs to be understood by machines, not just ranked by a search engine?

SEO was never really about rankings

It's easy to reduce SEO to:

“Get to position #1.”

But rankings were always a means to an end.

The actual goal was discovery.

You wanted someone searching for something you could provide to discover your website, understand what you offer, and eventually become a visitor or customer.

Search engines were simply the primary gateway.

Now there are more gateways.

That's where terms like AEO and GEO have started appearing.

SEO, AEO, GEO — what's the difference?

There isn't one universally agreed definition for every acronym, and the terminology is still evolving, but the general idea is useful.

SEO (Search Engine Optimization) focuses on improving a website's visibility in traditional search engines.

AEO (Answer Engine Optimization) focuses more on making information easy for answer-oriented systems to understand and use when producing direct answers.

GEO (Generative Engine Optimization) generally refers to optimizing content for AI-powered search and generative systems.

The important part isn't memorizing three acronyms.

It's understanding the underlying shift:

Your website may now have to be discoverable across several different information systems.

Search engines and AI don't necessarily consume your website in the same way

A traditional search engine might primarily care about things such as:

Crawlability
Indexability
Relevance
Links
Page experience
Content quality
Structured data

AI systems introduce another layer.

They need to determine things like:

What is this page actually about?
Who created this information?
Is this information trustworthy?
How does this page relate to other entities and concepts?
Can the information be extracted and understood correctly?
Does this source provide a useful answer to the question being asked?

This doesn't mean that you can forget about traditional SEO.

Quite the opposite.

A technically broken, inaccessible, poorly structured website isn't suddenly going to become great because you added some AI-related metadata.

The fundamentals still matter.

The biggest mistake: optimizing for keywords instead of understanding

One of the oldest ideas in SEO is keyword optimization.

And keywords still matter.

But imagine a page that contains the phrase:

“best running shoes”

50 times.

Does that necessarily make the page useful?

Of course not.

A system trying to understand the page needs much more context.

Is the page reviewing shoes?

Selling them?

Comparing them?

Talking about running injuries?

Discussing the history of running shoes?

The words alone aren't enough.

Context matters.

This is why things like semantic HTML, clear headings, structured data, internal linking, descriptive content, and well-organized information are becoming increasingly important.

The goal isn't to make your website “look optimized.”

The goal is to make its information easy to understand.

But there's a problem with AI visibility

Traditional SEO gives us relatively familiar metrics.

You can look at:

Search impressions
Click-through rate
Rankings
Organic traffic

AI visibility is much harder to measure.

Suppose you run a company that sells developer tools.

You ask several AI systems:

“What's the best tool for X?”

One model mentions your company.

Another doesn't.

You change the wording.

The result changes again.

You ask tomorrow.

It changes again.

Now you have a measurement problem.

How do you reliably test whether an AI system understands and surfaces your website?

And even more importantly:

How do you know what to change if it doesn't?

Finding problems isn't the same as fixing them

This is where a lot of website optimization tooling starts to show its limits.

A typical audit might tell you:

Missing meta description.

Okay.

Then:

Images aren't optimized.

Okay.

Then:

Duplicate content detected.

Okay.

Then another 47 warnings appear.

Eventually, you have a giant checklist of things you're supposed to fix manually.

The difficult part isn't always finding the problem.

It's deciding:

What should I change?

How should I change it?

Will the change actually help?

Did the change break something else?

That last part is particularly important.

Automation without verification can easily turn into automated damage.

The future probably isn't “AI replaces SEO”

I don't think SEO is disappearing.

And I don't think every website suddenly needs to become an “AI-optimized website.”

The fundamentals of building a good website haven't changed:

Useful information.
Good structure.
Good performance.
Accessibility.
Technical correctness.
Trustworthy content.

What's changing is the number of systems that sit between your content and the person trying to find it.

Google is one.

AI search is another.

Answer engines are another.

And there will probably be more.

So instead of thinking:

SEO vs AEO vs GEO

I think it's more useful to think:

One website → many discovery systems.

Where this is heading

This shift creates a new kind of problem that traditional SEO tools were never really designed to solve.

Not just:

“Is this page indexed?”
“What position does it rank for?”

But more like:

“Can different systems consistently understand this page?”
“What parts of the content are being misinterpreted or ignored?”
“If something changes, what actually improved or broke?”

We’re moving from a world of static checklists to something much more dynamic and uncertain.

And that uncertainty is where most of the interesting work is happening right now.

Closing thought

If machines are becoming part of how people discover information, then websites need to get better at communicating with machines—without becoming worse for humans.

That balance is going to define the next version of SEO.

And it’s probably going to be a lot more interesting than just trying to rank one position higher.