How I Used AI to Automate UniApplyForMe’s Bursary Content Workflow

The problem with bursary content at scale

UniApplyForMe lists hundreds of bursaries. Every year they change — amounts, deadlines, requirements, and eligibility criteria. Keeping that content accurate, structured, and SEO-ready is a real problem when you’re a small team.

The old workflow was manual: someone finds a bursary, pastes the raw text, tries to format it consistently, adds the right headings, writes a meta description, and picks a keyphrase. Multiplied by hundreds of listings, it was eating hours we didn’t have.

What I built

I built a React-based tool I’m calling the Bursary Content Formatter. You paste in raw bursary text – copied straight from a PDF, a government website, or whatever – and it sends that to the Claude API with a structured prompt. The output is clean, WordPress-ready HTML with proper heading hierarchy, a suggested Yoast SEO keyphrase, a meta description, and section-level copy buttons so you can grab individual chunks.

It also keeps a localStorage history of the last 20 reformats so nothing gets lost mid-session.

The prompt engineering side

Getting Claude to output consistent, structured HTML took iteration. The key was being explicit about the output format upfront – specifying WordPress block comment syntax, heading levels, and the exact structure of the keyphrase suggestion. Vague prompts give vague output. Specific prompts give usable output.

I also had to handle the edge cases: bursaries with no closing date, ones with multiple eligibility criteria, and ones where the source text is a wall of unstructured paragraphs. The prompt handles these gracefully now.

What this actually unlocked

We went from spending 15-20 minutes per bursary listing to under 2 minutes. That’s not a small win – at scale it’s the difference between keeping content updated and letting it go stale. Stale content is invisible content, and invisible content doesn’t help Grade 12 learners find funding.

This is the kind of automation that actually matters: not replacing people, but removing the parts of the work that are repetitive and low-value so the team can focus on what requires judgement.

Should you build something like this?

If you run a content-heavy WordPress site and you’re still formatting posts manually, yes. The Claude API is straightforward to integrate, the cost is negligible at this scale, and the time savings compound quickly. The hardest part is writing a good prompt – but once it’s working, it just works.

The tool is internal to UniApplyForMe for now, but I’m considering open-sourcing a generalised version. Watch this space.