How I accidentally built an AI blog writer

A personal story about desperation, duct tape solutions, and discovering that the best products come from solving your own problems

Updated October 2026. I wrote this in October 2025, before Draftly was commercially available. The story is the same. The product at the end of it has changed a lot, so I updated that part.

I never intended to build a SaaS product. I’m not a developer by trade. I’m a marketer who got thrown into the deep end when my company’s organic traffic fell off a cliff, and I had to figure out how to save it.

This is the story of how a domain migration disaster led to late-night coding sessions, a janky Google Apps Script, and eventually, Draftly.

The crisis that started everything

In 2024, I joined a data company that had enjoyed decades of brand recognition. They were the kind of company that didn’t need to work hard at SEO because everyone already knew who they were. Type their name into Google, and boom, top result.

Then they migrated domains.

The traffic collapsed. We’re talking exponential losses. The kind of losses that make executives panic and marketing teams scramble.

I was brought on to fix it. My boss and I did the initial audit, made the technical improvements, optimized what we could, and then… we waited. We waited for Google to “figure it out” and restore our rankings.

It didn’t happen.

Our brand wasn’t going to bounce back on its own

After months of waiting and hoping, the truth became clear: our brand wasn’t just going to magically bounce back. The rules had changed. Brand recognition alone wasn’t enough anymore. We needed content, lots of it, and we needed it fast.

But here’s the problem: we couldn’t sacrifice quality for speed. We were a leader in our space, and our reputation depended on maintaining that authority. Every blog post needed to sound like us, reflect our expertise, and position us as thought leaders.

My boss and I had a frank conversation. We needed to scale our blog content quickly without destroying the brand we’d spent decades building.

The problem was simple. The solution was not.

The duct tape solution

I’m not a developer, but I know how to make APIs talk to each other. So one weekend, fueled by coffee and frustration, I built something ridiculous.

It was a Google Apps Script that:

• Had RSS feeds hard-coded into it (yes, really)
• Connected to OpenAI, Claude, and Gemini APIs
• Included massive context blocks about our brand, our voice, our products
• Generated blog posts that actually sounded like us

It was ugly. It was fragile. It probably violated several best practices.

But it worked.

Within weeks, we were publishing blog posts that our executives couldn’t distinguish from content our team had written manually. The topics were relevant. The voice was on-brand. The quality was there.

More importantly, our traffic started recovering.

What I learned from building the wrong thing

That first version taught me something crucial: AI-generated content doesn’t have to be soulless and generic.

Most AI content fails because of a simple principle: garbage in, garbage out.

If you just ask ChatGPT to “write a blog post about SEO,” you’ll get exactly what you’d expect: bland, generic, forgettable content that sounds like every other AI-generated post on the internet.

But if you give the AI your opinions, your expertise, your brand’s personality, and context about what makes your company unique, something changes. The content goes from generic to genuinely useful.

Here’s why: Large Language Models are statistical word prediction machines. They’re trained on massive datasets and have gotten incredibly good at predicting what word should come next in a sentence. When you inject your brand’s unique perspective into that process, you’re giving it better data to work with.

Suddenly, instead of predicting “what would a generic blog post say,” it’s predicting “what would YOUR brand say about this topic.”

From hack to product

I didn’t set out to build a product. I just wanted to solve my own problem. But as I refined the system, added more features, and watched it consistently produce quality content, I realized something:

If I’m struggling with this, so is every other agency and content team out there.

Agencies spend countless hours on the basics: meta titles, descriptions, keyword research. Clients care most about content that’s high quality, relevant to their industry, and positions them as experts.

They especially care that it sounds like them. And that’s where most of the editing time goes. That’s the time-suck that bleeds agencies dry.

So the first version of Draftly I offered to other people was built for agencies.

Then I narrowed it to small businesses

That agency version taught me a second lesson. In December 2025, after testing it with a few inside users, I realized it was overcomplicating SEO work instead of simplifying it. Too many clients, too many settings, too many decisions.

So I cut Draftly down to one business and one website per account. I wrote about that change in why Draftly went from agency tool to self-serve blog writer.

The philosophy behind Draftly

Draftly emerged from a simple belief: AI should amplify your expertise.

Most people aren’t great writers. I’m not. I rely heavily on AI to help me communicate ideas clearly. But here’s the thing. Even if AI only writes at a “C” grade level (it can do better than that), that’s still better than failing. And lots of businesses are failing at writing.

They’re engineers, developers, entrepreneurs, and innovators. They’re brilliant at what they do, but writing isn’t their strength. AI can help them communicate their brilliance, and that’s incredibly valuable.

But AI needs your input. It needs your perspective, your expertise, your brand’s unique voice. Draftly is built to take that input and turn it into posts you’d be proud to publish.

What Draftly does today

You paste your website URL. Draftly reads your site to learn what you sell and how you talk, picks a timely topic from your industry’s news, and writes a post in your brand voice.

• It checks every draft against writing rules in code before you see it, and runs a fix pass when something fails. The rules are on our SEO copywriting page.
• It writes with GPT 6 Luna by default, picked after testing models against those rules. Paid plans can choose from 8 models. See how 14 models compared in our AI writing benchmark.
• It links only to real pages on your site, and to outside sources that actually load.
• It publishes to WordPress, Shopify, Webflow, Ghost, Squarespace and HubSpot.
• It works inside Claude and ChatGPT on paid plans, through a connector with read-only SEO tools. More on the MCP page.

The goal is to cut the parts of content creation that drain your time and energy so you can focus on what matters: strategy, creativity, and building relationships with your customers.

What this means for you

If you’re a small business owner who knows you need content but can’t afford a full-time writer, I built this for you.

If you’re a marketer like me who’s tired of choosing between quality and quantity, I definitely built this for you.

The future of content combines human creativity with AI efficiency to make something better than either could alone.

Try it on your own site

Draftly is live now in the United States. Your first post is free, with no credit card. Paid plans start at $29 a month, and the details are on the pricing page.

If you want the full walkthrough first, the AI copywriter page shows how Draftly goes from your URL to a published post.

What’s the part of blogging you’d hand off first if you could?