Building a Blog with AI: From Zero to 50 Articles
Updated · May 14, 2026
We gave ourselves one constraint: build a real blog from a blank domain to 50 published articles using AI tools only. No freelancers, no agency, no human writers on retainer. Just us, a keyword list, and roughly $200 a month in software. The niche was remote work productivity — competitive enough to be meaningful, specific enough that we weren’t going head-to-head with Forbes on day one. Eight weeks later, here’s an honest account of what that actually looked like.
The rules we set for ourselves
Budget: $200/month total, including hosting. Articles: minimum 800 words each, written to rank. Timeline: 50 articles in eight weeks, which works out to about six or seven per week. Quality bar: would we publish this on a site we were proud of?
That last rule did the most damage to our early optimism. AI tools can produce volume. Producing volume that a human would actually want to read is a different problem, and it took us longer than we expected to solve it.
Keyword research almost broke the budget before we started
Our first instinct was Semrush at around $130/month — full-featured, the industry default — but that would’ve consumed 65% of our budget before we wrote a single word. We needed to be smarter about the stack.
We paired Ahrefs Lite (around $99/month) with Frase at its $45/month Pro tier. The logic: Ahrefs for raw keyword data and competitor gap analysis, Frase for building content briefs that writing tools could actually act on. Frase’s brief builder pulls the top SERP results for any keyword and produces a structured outline — suggested headers, target word count, NLP terms Google expects to see. That brief becomes the instruction set for the AI draft.
We generated 80 keyword targets in one afternoon. Pruned to 55, picked the 50 with the clearest search intent. Total time spent on strategy before writing a single article: about six hours.
What we rejected: Surfer SEO at $89/month is excellent, but we couldn’t justify running it alongside Ahrefs when Frase covered most of the same brief-building ground. If we were scaling past 200 articles, we’d reconsider. At 50, the overlap wasn’t worth the budget.
The writing tool decision took longer than expected
We tested three tools for actual drafting: Jasper, Claude, and ChatGPT. All three for a full week, using the same five Frase briefs, the same prompts.
Jasper at around $49/month on the Creator plan was the most structured experience. Its campaign and template system pushes you toward repeatable formats, which sounds ideal until you realize how much voice it flattens. For informational content — how-tos, listicles, step-by-step guides — it’s fast and serviceable. For anything that requires a real perspective, the output needed heavy rewriting.
Claude surprised us. At $20/month for Claude Pro, we got a longer context window than ChatGPT at the time, which meant we could paste a full Frase brief — headers, NLP terms, competitor notes — and receive a complete draft in one shot. On nuanced topics like remote work burnout or managing async teams, the writing quality was noticeably higher. The tradeoff: no native SEO workflow. We had to manually cross-check NLP terms from Frase against each draft before publishing.
ChatGPT at $20/month handled fast, high-volume content well — “best tools for remote standups,” “how to set up a home office on a budget.” On longer analytical pieces, the outputs were thinner. We ended up using it primarily for generating title variations and meta descriptions in bulk, where speed matters more than depth.
Our final workflow: Frase brief → Claude draft → 30-45 minutes of human editing → Frase NLP check → publish. For simpler informational targets, we swapped ChatGPT for Claude to save time. The split was roughly 60/40 by article count.
The thing nobody told us: editing consistently took longer than drafting. Budget at least 30 minutes of real human time per article if you want something that doesn’t read like a content farm output. For our 50-article run, that was roughly 30 hours of editorial work total — which is where the actual quality bottleneck lived, not in the tools.
Images, publishing, and the invisible layer nobody writes about
For featured images and in-article graphics, we went with Canva Pro at $15/month. Midjourney would have produced more striking visuals, but for a productivity blog, consistent branded templates beat artistic variety. We built one template in Canva, swapped the headline text per article, and exported. Four minutes per article once the template existed.
WordPress handled publishing. We tried one AI publishing plugin briefly — it added friction without removing any. Dropped it after three articles.
Automation: Zapier at around $20/month connected the workflow. When a draft was marked ready in our Notion editorial calendar, a Zap created a WordPress draft and flagged it for review. Small thing — saves about 10 minutes per article, which compounds to 8+ hours across 50 pieces. Notion at $10/month served as our single source of truth: keyword, target URL, brief link, status, publish date. Dead simple. Worked.
What we’d change next time
We over-invested in Ahrefs for month one. Starting from zero, we’d validate the niche first using Ahrefs Webmaster Tools (free) and Google Search Console before committing $99/month. Paid keyword tools earn their cost at scale, not at launch.
We underestimated the editing bottleneck. By week four, drafts were piling up faster than we could review them. The actual constraint wasn’t AI output speed — it was human review capacity. If we ran this again, we’d budget for a part-time editor rather than another AI tool. More tooling at that stage would have made the backlog worse, not better.
We also tested three “all-in-one AI SEO writing” platforms that promised to replace the Frase plus Claude workflow. None of them delivered. The SEO guidance was generic, the writing quality dropped, and each one added steps rather than removing them. Specialization consistently beat consolidation.
The final stack
- Ahrefs Lite — keyword research, competitor gaps — around $99/month
- Frase Pro — content briefs, NLP optimization — $45/month
- Claude Pro — primary drafting (long-form, nuanced topics) — $20/month
- ChatGPT Plus — fast drafts, meta descriptions, title testing — $20/month
- Canva Pro — featured images, blog graphics — $15/month
- Zapier Starter — workflow automation — around $20/month
- Notion Plus — editorial calendar — $10/month
- WordPress + hosting — publishing — around $15/month
Total: around $244/month. We came in slightly over our $200 target once hosting was counted. The first article published 11 days after domain registration. Average time per article from brief to live post: 1.8 hours. Fifty articles: done.
Frequently asked questions
Do you need to pay for all these tools to get started?
No. A reasonable entry point is Claude Pro or ChatGPT Plus for drafting ($20/month each) combined with the free tiers of Ahrefs Webmaster Tools and Google Search Console for keyword validation. Spend $40/month first and upgrade individual tools as your output volume justifies it.
Can AI-written content actually rank on Google?
Our articles are ranking, but it took six to eight weeks before positions moved meaningfully. Heavily edited AI content performs in line with human-written content in our testing — thin, unedited AI output does not rank. The quality bar is the same; the authorship is not what Google is measuring.
Which single tool would you keep if you had to cut everything else?
Frase. The content brief changes what you can ask any writing tool to produce. Without a real brief, you’re prompting in the dark. With one, you’re giving the AI a specific, rankable target — and the output quality difference is significant enough to feel it on the first article.
Fifty articles with a two-person team in eight weeks is genuinely achievable with this stack — but the hours don’t disappear, they shift from writing to editing and strategy. The tools solve the blank-page problem. Solving the quality problem is still on you.
This article contains affiliate links. If you subscribe through one, we may earn a commission at no extra cost to you. It never changes what we recommend — we only link to tools we actually use. Full disclosure.
