How Ahrefs Automates Content Engineering With Claude Code
Post from the Ahrefs blog published on April 28, 2026 by Ryan Law (Director of Content Marketing, Ahrefs) describing an in-house content engineering system built around Claude Code: an editorial pipeline that produces publish-ready drafts in 6 to 12 minutes. Pivot thesis: *« AI content is not, by default, good.
By **Ryan Law** — Director of Content Marketing chez **Ahrefs**. Praticien senior du content marketing SEO ; le billet est un retour d'expérience personnel// Source ahrefs.com ↗/Reading 2 min/.md// Auto-verified translation
Ryan Law, Director of Content Marketing at Ahrefs, describes in this April 28, 2026 blog post the content engineering system he built around Claude Code to produce publish-ready article drafts in 6 to 12 minutes. His founding thesis defuses the hype from the outset: « AI content is not, by default, good. This process works well because it mirrors our existing human editorial process ». In other words, quality doesn't come from the model but from the faithful reproduction of a human editorial process already proven.
The system rests on about 23 skill files, each corresponding to a specific editorial step: keyword research, topic gap analysis, structural outlining, research compilation, draft generation, formatting. A master skill, blog-pipeline, chains them together to produce a complete article. At the core of the setup, the Ahrefs MCP lets Claude pull real SEO data (keyword metrics, parent topic, long-tail themes, SERP overviews, search intent, competitive analysis) rather than inventing figures.
if you get an article at the end of a ten minute run, and it's bad, it's hard to diagnose precisely where and why the process went wrong
— **Ryan Law** — Director of Content Marketing chez **Ahrefs**. Praticien senior du content marketing SEO ; le billet est un retour d'expérience personnel , ahrefs.com
Law lays out seven design principles: mimicking existing human workflows; outputting each step separately for troubleshooting (« if you get an article at the end of a ten minute run, and it's bad, it's hard to diagnose precisely where and why the process went wrong »); creating test cases via Anthropic's skill-creator; plugging in quality data sources; front-loading human direction via context parameters; building interactive HTML previews for review; and making the system forkable and customizable by each team member.
The numbers remain deliberately modest: ~15 articles published and ~30 updated, with development starting in February 2026 (the prior process, from August 2025, took several days plus manual intervention). Law pairs all this with explicit caveats that reinforce his credibility: « experience matters » (the process reflects decades of expertise), topic selection is limited to well-mastered informational SEO content, and Ahrefs has no plan to scale massively — it maintains an evergreen library.
The overall philosophy is to automate only « the formulaic parts of work » — the formulaic drudgery — to free up time for research, thought leadership, webinars and system optimization, without replacing human effort. This post has become the field reference cited by 2026 SEO workflows (notably Pasquale Pillitteri), a concrete illustration of the skills-over-prompts and systems around the model doctrine: the advantage comes from the orchestrated pipeline, not the raw model.
Key takeaways
Date / source.April 28, 2026, Ahrefs blog (ahrefs.com/blog). Author: Ryan Law (Director of Content Marketing, Ahrefs).
Central thesis (to remember verbatim).« AI content is not, by default, good. This process works well because it mirrors our existing human editorial process. » ### Pipeline architecture
~23 skill files. , one per editorial step: keyword research, topic gap analysis, structural outlining, research compilation, draft generation, formatting.
Orchestration by a master skill blog-pipeline that sequences the others → complete article.
Ahrefs MCP. = live SEO source of truth (keyword metrics, parent topic, long-tail, SERP overview, intent, competitive analysis) instead of hallucinations. ### The 7 design principles 1. Mimic human workflows (skills adapted from existing Ahrefs editorial documentation). 2. Output each step separately → troubleshooting (save intermediate outputs). 3. Create test cases via Anthropic's skill-creator. 4. Plug in quality sources (Ahrefs MCP, competitors, product docs). 5. Front-load human direction (context parameters). 6. Interactive HTML previews for review before publication. 7. Forkable / customizable by each team member. ### Numbers & caveats
6 to 12 minutes. for a publish-ready draft (canonical figure quoted everywhere).
~15 articles published. , ~30 updated. Development started February 2026; the prior process (August 2025) meant several days plus manual intervention.
Explicit caveats.« experience matters »; topics = well-mastered informational SEO; no massive-scale plan → evergreen library.
Philosophy: automate « the formulaic parts of work » (drudgery) → free up time for research / thought leadership / webinars / system optimization. ### To leverage in engagements / presentations
Field proof-point. for the editorial-cycle gain: 6-12 min/draft, anchored on live data (MCP), with a human safeguard.
Direct illustration of the skills-over-prompts + systems around the model doctrine (the edge = the orchestrated pipeline, not the raw model) — overlaps with Lattice, PROJ-AI, Dropbox/Okumura.
The save-every-step idea = reusable agentic-pipeline observability as a pattern (cf. decision traces / agent debugging).
Key figures
~15 articles published + ~30 updated via the pipeline