AI content pipeline
for SaaS companies.

Yes, when it's built and edited the way Google's guidance says content should be built — original insight, real utility, human judgment on quality. HCU's guidance is about content that doesn't serve the reader, not the mechanism used to draft it. Our pipeline ranks post-HCU because every post is drafted from real keyword research mapped to a real ICP job story, edited by a human for accuracy and voice, structured with clear author bylines and dates, and published with proper internal linking and structured data. Content sites that got hit by HCU published templated slop at scale with no editorial oversight — the exact opposite of what we do. Zero clients have lost pages to core updates in the last four update cycles.

Four things, all non-negotiable: (1) Every post is drafted against a brand-voice profile trained on your best-performing existing writing — not a generic LLM voice. (2) Every draft is edited BY A HUMAN before publish — not proofread by another model. (3) Structural integrity — real subheads, real internal links, real citations, real examples specific to your product, not the “Introduction / Benefits / Conclusion” skeleton. (4) The topic itself has to be based on real ICP job-story research, not “write 100 posts about our category.” If any of those four break, we stop shipping until it's fixed.

Typical edit passes touch 15-25% of the words on a first-cluster post, dropping to 5-15% by month three as the voice profile tightens. The editor's job is: (1) fact-check every specific claim (numbers, product features, competitor references), (2) rewrite anything that sounds LLM-generic, (3) add company-specific examples the model doesn't know, (4) fix structure — subhead hierarchy, internal linking, and reading flow. If a draft needs more than a full rewrite, we throw it out and re-draft — we don't polish bad drafts.

The brand-voice profile is more than a prompt. Setup week 1 ingests 20-40 of your best-performing existing posts and (if you have one) your style guide. We extract tone patterns, sentence length distribution, product-vs-education balance, forbidden words, house style on em-dashes vs. colons vs. Oxford commas, and preferred phrasings. Every draft is scored against the profile before it reaches the editor — drafts that miss the voice bar get re-drafted, not sent through. The profile updates monthly based on which drafts your editor accepted vs. rewrote heavily.

Native publishing to Sanity, Contentful, WordPress, Ghost, and Webflow — direct API publish with SEO metadata, internal linking, and structured data on every post. Draft workspace: Notion. Editorial calendar: Airtable. Keyword + SERP data: Ahrefs, SEMrush, Google Search Console. Attribution: Google Analytics 4, PostHog, Amplitude, Mixpanel. Slack for editorial review notifications. Anything with an API or webhook we can wire — custom integrations get scoped and priced in the setup fee.

Yes, when it's done with a human review pass and unique-per-page content. The pages that got sites de-indexed in the 2024 core updates were programmatic pages with near-identical thin content — the doorway-page pattern Google has always penalized. What we build for comparison + integration pages: unique intro tailored to the specific X vs Y, real feature-comparison table with cited sources, use-case-specific examples, and a human editor sign-off before publish. Every page passes the “would a real person land here and get value” test. That's the delta between programmatic pages that rank and programmatic pages that get deindexed.

Five that separate the good from the dangerous: (1) Show me a client's pages that survived the last three Google core updates. (2) Who edits every post — a human editor, or another model? (3) How do you build the brand-voice profile, and can I see one you built? (4) What's your attribution model — vanity traffic or trials/signups/deals? (5) Do you kill clusters that rank but don't convert? (Most agencies won't — they're paid on volume.) We're happy to answer all five on the discovery call.

3 to 4 weeks. Week 1: brand-voice profile + first cluster keyword research. Week 2: outlines for the first 8 posts, editorial calendar approved. Week 3: first drafts + edits + CMS publish for the first 4 posts. Week 4: remaining posts published, first performance data. 30 days of tuning included after that. Exit: $500 deposit fully refundable if we can't confirm scope or commit a kickoff date within 7 days. Post-launch you own the voice profile, all published posts, keyword clusters, and editorial calendar — we export on request and shut off billing the following month. No multi-year contract, no exit fee.