Anyone can produce content now. Earning attention is the harder part, and the part the AI tools can't shortcut on their own. We write stories your audience actually reads, ship them at the cadence your business needs, and place them where the readers (and the search engines that quote them) already are.
The thesis
The internet is louder than it has ever been, and audiences are blunter about it. Ad-blockers are routine. Banner-blindness is universal. Sponsored posts that smell like sponsored posts get scrolled past before the message lands.
The work that lands is the work that earns its place. A story with something to say, evidence the reader can check, written for the person scrolling rather than the platform serving the ad. Published once, it keeps working. Articles compound. Campaigns expire.
Most businesses know this. Almost none of them do it consistently.
Pulling Rabbits exists for that gap.
The workflow
By the time a writer opens a draft, half the pipeline is already done. The topic has been diagnosed against the client's commercial reality, the brand context loaded, intent and format classified, the spine condensed from sources, the research bundled and source-checked. Five layers of decisions sit behind every paragraph the reader sees.
That is the difference between a piece that reads like it was thought about and a piece that reads like it was generated. The pipeline is not a chatbot. It is ten discrete stages, each with its own brief, its own specialised model, its own evaluation gate, and a senior editor on the output before anything hands off.
Diagnosis before production.
Before any brief is written, we understand the business. Client, product, audience, market position, where AI actually pays back in the workflow. Output: a commercial diagnosis. Not every brief belongs in the pipeline; this stage decides which ones do.
Market and competitor intelligence. AI-fit mapping per workflow.
Strategist runs the diagnosis. Editor scopes the brief.
What is this brand, really?
Once we know the why, we structure the how. A read of the brand, the audience, the existing content footprint, voice, position, off-limits topics. The context model holds the rules every downstream step is graded against.
Brand-context model briefed per client.
Strategists set the rules. Editors gate the brief.
Sort by intent, format, funnel.
Topics get sorted by reader intent, funnel position, format, urgency and content type. Decides what's worth producing and what isn't. Long-form vs landing vs LinkedIn vs newsletter, top vs bottom of funnel.
Topic-intent classifier tuned to category.
Editor signs the classification before research kicks.
Distil the spine before writing.
Sources, references and prior work compressed into the spine of the piece. The argument, the order, the evidence backbone. We know what we're saying before the first draft.
Multi-source synthesis with citation tracking.
Editor accepts the spine. No spine, no draft.
Active facts, not training-set memory.
Live web research, internal data, expert input. Pulls current statistics, named sources and verifiable claims. Anything that goes in the piece can be footnoted back to a real document.
Live web retrieval. Source-grounded chains.
Researcher validates. Editor adjudicates the claim.
Draft. Voice. Rhythm.
Production. LLM produces a draft against the context model, classification, spine and research bundle. Editor finishes it. Voice tuned per brand. House style trained into the model, not retrofitted in edits.
Per-brand writing model. Style and tone constraints locked.
Editor writes the parts the model can't.
Fact, brand, claim, originality.
Fact-checks, brand-voice check, plagiarism, claim verification, schema compliance, accessibility, reading level. Each layer is a separate model with a separate brief. Nothing publishes that fails any of them.
Independent QA agents per check type.
Senior editor on every flag the agents raise.
Visuals tuned to the brand.
Cover, in-line, social and chart graphics. Aesthetic tuned to brand identity, not generic stock. Hero image moves with the brand book. Charts are real charts on real data, not decorative slop.
Brand-tuned image pipeline. Style references locked.
Designer reviews. Charts approved against the source.
Findable. Quotable. Cited.
On-page, schema, internal linking, meta, structured data. The page is engineered to be found by Google and surfaced by ChatGPT, Perplexity, Gemini, Claude. The multi-engine reality of modern discovery.
Per-engine optimisation passes.
SEO lead approves the structure and link graph.
Ship. Amplify. Measure.
Final formatting, CMS deploy, social amplification, distribution to syndication partners, performance tracking. The first hour after publish is when most pieces fail. We don't lose the launch window.
Channel-specific cut-down generation.
Channel managers ship and watch the first wave.
What it produces
Bylined, evidence-led, defensible. The kind of piece a journalist would file.
Conversion-tuned, brand-voice, on-page SEO baked in from draft one.
Sequenced content built around lifecycle journeys, not blast templates.
Help centres, glossaries and FAQ schema engineered to be answered cleanly by AI.
The origin
Pulling Rabbits is the content production stack that runs YLO's SEO and editorial work. It is now available as a standalone engagement for brands, agencies and operators who need the same discipline without bringing it in-house.
Tell us the brand, the brief and the cadence. We come back with a sample piece run through the pipeline and a clear monthly figure.
