Build an AI content site on Cloudflare Pages without hiding the real costs
Plan a static AI content site on Cloudflare Pages, connect your domain, and budget separately for generation, data, review, and future automation costs.
Start with a site you can operate, not just deploy
A useful first version is a small collection of source-backed tutorials with a clear topic and a repeatable publishing process. For an AI tools site, choose one reader and one job: for example, helping a small team automate its content research. Prepare a homepage, a topic index, three useful tutorials, and an explanation of how articles are researched and corrected. A large collection of empty category pages adds work without answering a reader's question.
This is a planning tutorial based on primary documentation, not a claim that we tested a particular traffic level or earned advertising revenue. The practical goal is to separate the public reading experience from the machinery that gathers sources and drafts articles. That separation makes the budget and failure modes easier to understand.
Understand what free hosting covers
Cloudflare's Pages pricing documentation distinguishes static assets from requests that invoke Functions. Static asset requests are free and unlimited on its free and paid plans. Pages Functions use Workers quotas and pricing. An article rendered into HTML at build time can therefore have a different cost profile from a page that calls a model whenever someone visits.
The Pages limits documentation currently lists 500 builds per month, one concurrent build, a 20-minute build timeout, 20,000 files per Free-plan site, and a 25 MiB maximum per asset. These are documented limits checked on November 18, 2026, not permanent promises. Batch approved changes into sensible publishing windows instead of rebuilding the entire site after every feed item.
Create a cost sheet before adding automation
Use a worksheet with one row per service and columns for billing unit, monthly allowance, observed usage, hard limit, and owner. An existing domain still has renewal costs. Content generation, keyword providers, storage, email, and monitoring may each have their own terms. Do not combine them under a single 'free website' label.
A useful planning formula is total monthly operating cost = domain cost allocated monthly + hosting and compute + data access + model generation + monitoring + editorial work. Fill it with your actual quotes and usage. Until those values exist, leave the forecast blank rather than inserting a plausible revenue or cost estimate.
- Set a monthly ceiling for model calls and record failed calls as usage too.
- Keep research jobs separate from public page requests.
- Store the sources and article revision so a failed build does not require repeating all research.
- Include the time needed to check facts and repair broken instructions.
Deploy the static version first
Generate complete HTML for every published article. Confirm that the title, introduction, navigation, and source links are present in the built output. Keep credentials and unpublished drafts out of the public files. Open several pages locally, including a missing URL, before connecting the deployment.
Cloudflare Pages can connect to GitHub or GitLab and deploy repository changes. Its Git integration also provides branch previews. Use a preview to inspect an approved content change before it reaches the production branch. Treat a successful deployment as confirmation that files were served, not evidence of Google indexing or search performance.
Connect your domain without guessing at DNS
Start in the Pages project's Custom domains settings and register the hostname there. Cloudflare documents two different paths: an apex domain such as example.com must be a zone in the relevant Cloudflare account with the required nameservers; a subdomain can use a CNAME from an external DNS provider. Follow the path matching your domain setup.
Do not create only a CNAME and assume the Pages association exists. Cloudflare explicitly requires adding the domain through its Pages setup. After activation, verify the preferred hostname, HTTPS, internal links, and canonical URLs together. Keep an inventory of existing email and verification DNS records before changing nameservers.
Make the first publishing loop measurable
Run one complete cycle: discover a candidate, collect evidence, draft a useful answer, check it, deploy it, and record the resulting URL and revision. Keep research failures in a review queue. A missing source should stop an unsupported claim from becoming a public fact.
Your launch checklist should answer four questions: Can a reader use the tutorial? Can you explain each factual claim? Can you recover the previous version? Can you identify what generated each bill? Once that small loop works, increase throughput according to review capacity and observed operating cost. Article count alone does not demonstrate that the system is working.