Frequently asked questions

A service I run. In the founding cohort, I onboard every tenant personally, tune the brand voice in the first month, design the AEO prompt set, and stay on a direct Slack channel for ongoing tuning and support. You @mention the bot to draft and approve content. Everything underneath that is my responsibility.

If you want a tool to operate yourself, Jasper, Copy.ai, or Letterdrop are better fits. Content Loop Studio is for teams that want output, not a dashboard.

I cap the founding cohort at 10 customers for two reasons. First, I’m one person, and onboarding well takes 6-10 hours of focused work plus 3-5 rounds of voice tuning over the first month. Ten is what I can deliver without compromising quality. Second, the founding rate is meaningfully discounted, and the cap is what makes the unit economics work.

Once we hit 10, new customers join at Operator pricing ($5,500/mo + $5,000 setup). Founding members keep their rate for 24 months from signup.

Practically: you give me read-only access to Search Console + GA4 (a service account, about 10 minutes), a GitHub PAT scoped to your marketing site repo, and your chat tenant (Microsoft Teams or Google Chat). I do the rest: Cloudflare deployment, Workers configuration, Vectorize index creation, GitHub webhooks, chat bot wiring.

You spend about 3 hours in week one helping me get the brand voice doc right (the only thing I can’t write for you without your input). After that, it’s about 30 minutes a week to review drafts.

They probably won’t be. Week-one drafts are the worst the agent will produce. The Writer hasn’t tuned to your voice yet. I’d plan to reject or heavily revise the first 2-3 drafts as we calibrate the brand voice doc against actual output. By week 3-4, drafts are usually shippable with light edits. By week 6-8, drafts are shippable without me as a checkpoint.

If after the first month you still don’t see usable output, that’s a signal Content Loop Studio isn’t the right fit and we should part ways with a refund of unused time. I’d rather know that early.

For v1:

• Static site generator: Hugo. Astro and WordPress are on the roadmap (likely Q3/Q4 2026). Hugo is the only supported target today.
• Repo: GitHub. GitLab and Bitbucket are possible but require code changes. Book a call to discuss.
• Hosting: Cloudflare Pages. Other static hosts (Netlify, Vercel) could work but require integration work. Pages is the default because branch previews are critical to the approval flow.
• Chat: Microsoft Teams or Google Chat. Slack is the most-requested addition. Not yet supported.
• Analytics: Google Search Console + GA4. Other analytics platforms (Plausible, Fathom, server logs) don’t yet integrate.

If you’re on a different stack and serious about Content Loop Studio, the discovery call should cover whether the gap is bridgeable in your timeframe.

Week one: yes, probably. Week six: shouldn’t. Two layers of defense against AI-cadence:

First, the Writer loads your brand voice document and 3-5 example posts on every draft. Explicit reference material, not generic style rubrics. The tuning happens in the brand voice doc. We iterate on it during onboarding until drafts read like your team wrote them.

Second, every draft passes through a humanizer post-process step before opening a PR. It detects 29 known AI-tell patterns: significance inflation, AI vocabulary (“crucial”, “landscape”, “delve”), em-dash overuse, signposting, copula avoidance, rule-of-three filler, and others. It rewrites them in your voice. You can opt out per-tenant if you want raw Writer output. It’s on by default.

The honest disclosure: there’s a “tell” floor with any AI-drafted content right now. Skilled readers can sometimes still detect AI-cadence. The goal isn’t to deceive. It’s to produce content good enough that the question “did AI write this?” stops being interesting because the substance is right.

Two layers of mitigation. First: the Analyst agent is grounded in your real Search Console and GA4 data, plus semantic search over your existing site. Topic ideas and content briefs come with specific numbers cited from those sources. Not invented.

Second: the Writer’s system prompt explicitly forbids invented statistics. When the Writer needs a number, it should pull from the brief (which came from grounded data) or cite a source it has access to. If it makes one up anyway, that’s a bug. Flag it on Slack and I’ll tighten the prompt that week.

We’re not perfect on this. I won’t pretend we are. But it’s a known concern that I track and tune.

Your data lives on Cloudflare infrastructure (Workers, Durable Objects, R2, Vectorize) under per-tenant structural isolation. Specifically:

• Your site content is indexed into a per-tenant subset of a shared Vectorize index, filtered by tenant ID metadata.
• Your brand voice doc and AEO snapshots are stored in R2 under tenants/{your-id}/....
• Your conversation history and approval state live in your tenant’s Durable Object SQLite database.
• LLM calls go to Anthropic. Data retention is per Anthropic’s terms (we don’t opt into anything beyond defaults).

I do not use your data to train models. I do not share it with other customers. On contract end, your data is exported on request and deleted within 30 days. The License page has the formal terms.

You can. There’s no quota on drafts in any tier, only the platform-side budget caps (per-request and per-day in USD) which I configure conservatively for cost safety. If you want to ramp up to 3-5 posts a week, we adjust the caps and the bot keeps drafting.

The bottleneck usually isn’t the bot’s capacity. It’s your team’s review time. Most founding cohort customers settle into 1-2 posts a week.

One site per tenant in v1. If you have a marketing site, a docs site, and a careers site, that’s three tenants, which lands in Scale-tier territory. We can run them under one contract with shared brand voice tuning if the voice is consistent across sites.

• Week 0: Discovery call. Live demo against your real Search Console data (read-only, temporary scope).
• Week 0.5: Proposal + MSA (within 48 hours of the call).
• Week 1: Setup. Cloudflare deployment, GitHub wiring, chat bot, GSC/GA4 service accounts.
• Weeks 2-4: Brand voice iteration. First drafts ship around week 2. Quality climbs through week 4.
• Week 5+: Steady state. You’re approving drafts in chat. I’m in your Slack for anything that needs tuning.
• Month 3: First quarterly strategy review.

Total from signed contract to “shipping content without my involvement on every draft” is typically 3-4 weeks.

In the founding cohort, you have a direct Slack channel with me. If the bot’s not responding, a draft errors out, or an AEO baseline times out, message me on Slack. Same-business-day response, usually within a few hours during my work hours (PT).

Underneath, every operation produces a Langfuse trace that I can pull to diagnose. The platform has automated alerts for cost spikes and Anthropic 5xx storms. I get paged before you notice in most cases.

Past the founding cohort, Operator and Scale tiers have defined SLAs in the contract.