Four Months of Building Top Title

Hey everyone,

It's been just over four months since we launched Top Title on July 4th. This has been a journey of constant iteration, learning from your feedback, and building something better every day.

TLDR

AI Model

Migrated from Sonnet 3.7 → GPT-5 (tested and rejected for robotic output) → Sonnet 4.5 (current)
Higher throughput, natural output quality, better brand voice

Interface & Experience

Redesigned UI - cleaner, faster, more intuitive
Sorting, pagination, and arrow key navigation
In-app QA workflow - edit titles, save changes, export to CSV

Performance

50-60% lower credit costs (tier-dependent)
Faster, more reliable processing
Significantly more page content analyzed per URL

New Features

Configurable character limits
Optional explanations for faster processing
Smarter company research
Two-stage keyword research system

Stats

180+ jobs processed
18,000+ URLs optimized

Coming Soon

Top Title 2.0 - new architecture, more intelligent, agentic decision-making

The AI Model Journey

One of the biggest challenges we faced was finding the right AI model for title optimization. Here's how that journey went:

Sonnet 3.7: Great Quality, Limited Scale

We launched with Claude Sonnet 3.7, and honestly, we loved the output. The titles were natural, creative, and hit the right balance between SEO and readability.

But there was a problem: cost and throughput.

Processing was expensive, and we could only handle so many URLs at a time. Sonnet 3.7 became the primary bottleneck in our processing pipeline. We needed something that could scale.

GPT-5: The Promise of Scale

When GPT-5 came out, it looked like the answer. Lower costs, much higher throughput, and as a larger model than Sonnet 3.7, we expected better intelligence and task performance.

We tested it thoroughly - minimal reasoning, medium, and high reasoning levels.

The result? Consistently robotic, unnatural, and keyword-stuffed output. It had an overly processed feel that just wasn't good marketing. The intelligence was there, but the writing quality wasn't meeting our standards. It wasn't appealing to readers, didn't capture brand voice, and wasn't the quality we needed.

Sonnet 4.5: The Sweet Spot

So we went back to Anthropic and moved to Claude Sonnet 4.5.

The difference was immediate. We got much higher throughput than 3.7 - solving our scaling problem - while maintaining the natural, appropriate output quality we needed. The titles feel human, they maintain brand voice, and they actually work in the real world.

Yes, it's still expensive as far as models go. But our goal here is to produce the best possible quality and consistency we can. And that's what we're getting from Sonnet 4.5.

That's where we are today, and we think it's a great mix of throughput and quality.

Interface & Experience Improvements

We've completely revamped how you interact with Top Title:

Navigation & Workflow

Redesigned interface - cleaner, more intuitive, faster
Sorting and pagination for easy URL organization
Arrow key navigation (left/right) for rapid review
Built-in QA workflow - edit titles directly, save changes, export to CSV. Many of you are using Top Title to QA right inside the app instead of jumping between tools.

Customization

Character limits now configurable per your needs
Explanations are optional - turn them off for significantly faster processing
Smarter company research with better, more relevant context

Performance & Cost Improvements

50-60% lower credit costs - depending on your tier
Faster processing - jobs complete quicker and more reliably
More page content - significantly more tokens analyzed per page for better AI context

Smarter Keyword Research

We've introduced a two-stage keyword research system that fundamentally changes how we identify target keywords for your pages. Instead of a single-pass approach, we now:

Analyze your page content deeply to understand what it's actually about
Cross-reference against search data to find the best keyword opportunities

The result? Better target keywords that actually match your content and search intent.

Learning From Real Users

One of the most valuable discoveries: working with real users is completely different from development testing.

When you're building in a development environment, you test against a limited set of scenarios. But when real users started using Top Title across diverse domains and page types, we found areas where the system wasn't performing as well as it should.

That's been invaluable.

Real-world scenarios exposed nuances we couldn't have anticipated. E-commerce product pages behave differently than blog posts. Local business pages have different optimization needs than SaaS landing pages. News articles, educational content, service pages - each has its own patterns and requirements.

Working with this diversity has made Top Title significantly better.

The Complexity of Our Craft

As someone who's spent years in SEO, I knew this would be hard. But building Top Title has reinforced just how nuanced our work really is.

The fine details we navigate every day:

Knowing how and when to use keywords - not just stuffing them in, but understanding where they add value
Staying natural - titles must work for humans first, search engines second
Balancing conflicting best practices - SEO rules sometimes contradict good UX, requiring judgment calls
Understanding the main content amid navigation, footers, and sidebars
Capturing company context - market position, offerings, and brand voice

Even with state-of-the-art AI, replicating what we do as SEO professionals requires constant testing and refinement. It's not a solved problem - it's an ongoing pursuit of excellence.

The Numbers

Over these first four months:

180+ jobs across diverse websites and industries
18,000+ URLs optimized - product pages, blog posts, service pages, and more

Every URL has taught us something.

Bug Fixes & Stability

Huge thank you to everyone who's helped identify and fix bugs. We've squashed dozens of issues, from small UI glitches to complex edge cases in the optimization engine. Your feedback has been invaluable.

Balancing Quality, Speed, and Cost

Our vision from the beginning: focus on quality, speed, and cost.

Time is money. But quality is trust - and that's the most important thing of all.

Our goal from day one was to build a tool that does professional-grade SEO optimization at scale. Not just faster or cheaper - actually better than doing it manually. But that vision sometimes conflicts with speed and cost. You can't always max out all three at once.

Different use cases require different priorities:

Large-scale sites - marketplaces with tens of thousands of pages, e-commerce with thousands of products. For these, throughput is critical. We need massive scale processed quickly and cost-effectively.

Smaller batches - agencies or in-house teams optimizing 20, 50, or 100 pages where quality is paramount.

Making Top Title work for both - processing thousands of long-tail pages efficiently while also enabling careful, high-quality optimization for critical pages - has been a core focus these four months.

What's Next: Top Title 2.0

Here's something exciting: we're already testing Top Title 2.0.

Completely new architecture. Built from the ground up. Early results are promising:

Better performance than the current version
More intelligent - thinks about your content in a more sophisticated way
Significantly more context - way more tokens to understand your pages
Agentic - smarter decision-making about how to optimize

We're getting ready for wider testing soon.

Want to be an early tester? Reach out - I'd love to work with you and get your feedback.

Thank You

To our agency users and direct users: thank you. Seeing Top Title in the wild and getting your real-world feedback has been invaluable. You've shaped our roadmap for 2026.

One thing is clear: we're staying laser-focused on our mission - making Top Title the first real and reliable "at scale" agent for professional-grade meta optimization.

That's what we do. That's what we're going to keep doing better.

Got ideas, issues, or want to chat about SEO? I'm here.

Here's to better titles and better rankings.

– Dan