The Content Engine
I build a content engine that turns your data, writing, and client work into content that sounds like you and ranks.
Teams come to me after an agency stalls or a content hire gets buried building the system instead of publishing. The last build drove 20x search impressions in 110 days.
Built by a 20 year Associated Press content systems veteran.
One engine build for Juniper Real Estate, a small San Diego brokerage, measured in Google Search Console. Read that case study →
The job underneath
The expertise is already in the firm. Turning it into live pages that keep working still means a lot of manual work.
You prompt a draft, then research the claims, fix the voice, make or find images, upload them, relink, fill SEO fields, and rewrite for LinkedIn. The next chat has none of that context, so the same manual work starts over.
See the hand-work side by side →Capacity shows up on the calendar. Months go to voice docs, templates, and workflows before a steady publishing rhythm does. The pages you wanted stay thin.
Compare hire, agency, and the engine →Teams keep funding the next post while the pages that already ranked get ignored.
See how archive refresh works →The engine is built for that work: sources you approve, voice locked from your work, a quality gate before you review, and images and SEO fields included in the push.
Your calls, writing, and client work teach the engine how your team thinks and sounds. I turn that into voice rules and a quality gate, so every draft reads like your work before it reaches you. When the build is done, you own every piece it publishes. The engine itself, the files and the publishing workflow, runs under a license your team uses in house.
It starts with strategy, the angle your market is missing and only you can own. For Juniper, the brokerage above, that was local pages backed by real data nobody else in the city would publish. Strategy decides what to say. The engine makes sure it ships, in your voice.
I spent 20 years at the Associated Press building content systems that had to be fast, accurate, and in one voice. The quality gate below is that newsroom discipline written down: it checks every draft for machine writing tells and unsupported claims, and sends failures back to be rebuilt before you ever read them.
Your team asks for work in plain language, the same way you would brief a person. There is no prompt to learn. Here is the range of what you can ask for, and you approve every piece before it goes live.
You would askWrite the permit guide from the county data I dropped in, then let me review it before publishing.
You would askGrade our services page, then fix what you find.
You would askWhat should we publish next quarter, based on what buyers keep asking?
You would askTurn that guide into a LinkedIn post and a follow up email.
What changes in practice: the heavy work happens before you review, you approve instead of rewriting at the end, and one push ships the page with images, links, and SEO fields attached.
Same job, two paths
Same article. Different amount of hand work.
You still approve before anything publishes. The busywork before and after that review is what drops out.
Before you write anything new
Most teams only publish the next post. Old pages go quiet, numbers go stale, internal links rot, and last year's best article stops matching this year's facts.
The engine goes back through the archive first. It refreshes the facts, relinks related pages, and brings strong URLs current. Search already knows those pages exist, so updating one usually costs less than starting from scratch.
Juniper is where I watched that pay off. I rebuilt every page instead of only adding new ones, 23 in a single week, and the rebuilt community pages climbed alongside the new work. The Mission Valley page alone pulled 7,000 impressions in that window. If you have years of posts, white papers, or research reports, keeping that library current can matter more than another new article.
How to think about the spend
Most teams weigh this against a content hire, an agency retainer, or another year of ChatGPT drafts. Here is that decision side by side.
| ChatGPT or generic AI | Content hire, first 6 months | Agency retainer, 6 months | The Pilot | The engine build | |
|---|---|---|---|---|---|
| Typical spend | Cheap tool, costly staff time | $45k to $70k loaded | $30k to $90k | $12,500 fixed, four weeks | From $27,500, one build |
| Sounds like you | Only if re-taught each time | After months of ramp | Sometimes | Yes, from your own material | Yes, locked in the system |
| Keeps old pages current | No | If they find the time | Rarely the focus | Can be part of the four weeks | Built to refresh and relink |
| Publishes into your CMS | No | By hand | By hand, or their stack | Files handed to your team | Wired to your stack |
| What you own after | Chat history | What they wrote before leaving | Their process stays theirs | The finished work | A license to run it, plus the content |
| Best for | One off drafts | Long term headcount | Ongoing outsourced production | Judging quality before a build | Running a system in house |
The hire and agency numbers are planning ranges, not a read on your payroll. On a fit call I swap in your real ones.
I run this engine for my own site. Here's the content creation workflow that turns research, voice edits, and quality gates into published pages. I've also shipped it for a real estate brokerage, a product consultancy, a housing data product, and a job search tool. Here are two of them.
Juniper Real Estate
An agency stalled for months. I gave Juniper a wedge, local content backed by real data that nobody else in San Diego was publishing, then rebuilt every page with the engine and shipped 23 pages in a week. Daily impressions went from 40 to over 1,500, clicks grew 4x, and ChatGPT, Perplexity, Claude, and Gemini now cite the pages. One local business found a guide and reached out.
Level Up Product
Jon Shutt built products at Disney, MLB, and Perry Street for 20 years, then let his consultancy sit for a year. I locked his positioning, built an engine on his own stories, and handed him an engine he runs by voice. Five approved posts shipped before my work wrapped.
People still Google, but more of them now ask ChatGPT, Perplexity, Claude, and Gemini, and those engines answer by citing pages they can parse. The engine writes structured pages tagged with schema, built to be that source. Juniper's already turn up in live answers.
A real estate brokerage got cited for coffee. The engine found live demand around North Park and thin coverage answering it, then wrote the guide that answered it better, with enough hard neighborhood detail for an AI engine to pull from. Nothing about that needed private data, and it did not need access to Juniper's site. I can produce finished pieces and hand them over as files for your team to publish.
A lot of the edits already feel really close to what I would want anyway. At this point, it's likely just nitpicking.
Ryan didn't just build pages, he helped me figure out what we should actually be known for. Instead of chasing the same market updates every agent in San Diego posts, we focused on the local questions people were really asking.
I'm Ryan. Most people call me Berg. One person builds your engine, start to finish, so there are no handoffs and nothing gets lost between a strategist and a dev shop. More about how I work.
I move fast and I like to keep it fun and as painless as I can. The one thing I ask for is a few working sessions up front so I can lock your wedge and voice before the engine starts producing. Get that right and production gets much easier.
Positioning and voice. I find your wedge, or help you find one if it isn't clear yet, mine your transcripts and past writing, then lock your voice and the roadmap.
I build the engine, tune the gate on real drafts, and wire up publishing automation.
Training for your team, first pieces shipped together, then a clean handoff with documentation.
You run it from day one, or I run production for a stretch and step back once the numbers show.
Before the engine
The engine only pays off if search engines can crawl, read, and rank your site. Whether it turns up in Start with Berg's review or the crawl that scopes your Pilot, the fix, broken crawling, missing schema, thin pages, usually folds into the work already underway. When the technical debt is bad enough to eat that whole timeline on its own, I scope it separately first as the Foundation, then the Pilot or the engine follows.
You own every piece the engine publishes. You hold a license to run the engine itself in house, with no seat fee that stops working the month you stop paying. I build it once, train your team, and can stay on after launch. The spend next to a hire or an agency is up in what this replaces.
Not ready for four weeks yet? Start with Berg is $6,000 for two weeks and ships real fixes before you commit to either offer below.
Four weeks of finished work on your own material, so you can judge the quality before anyone builds a system around it. A crawl sets the mix between new content and any site repairs your pages need before they can rank. $5,000 of the fee credits toward the engine build below if you start within 30 days, and you keep everything that shipped either way.
Most engagements start here so the proof arrives before the system does.
What you get in the build:
Scope depends on how many voices, how much automation, and whether distribution is included. Scoped after a call.
An extra marketer and builder on your team without the headcount. I run production, publishing, and SEO upkeep, and send a performance readout every month. Higher in the band it becomes strategy too: where to aim the engine next, which channels to add, and tuning live pages so more of the traffic converts. Where you land depends on how much ships each month.
Branded charts and custom graphics are optional extras. They increase the build scope and price.
Light technical groundwork, site config, speed, broken link cleanup, folds into the Pilot above. If the site needs more than four weeks of repair, that is the Foundation, scoped and quoted on its own before anything else starts.
You own
Every finished piece the engine publishes: articles, pages, posts, and the charts that ship with them. On full payment those deliverables are yours.
You license
The engine itself: the skill files, voice rules, quality gate, workflows, templates, and publishing setup. Your team runs it for your business. Changes to the engine itself come through me.
The license is for internal use. It is not for resale, sharing outside your team, or building a competing product. The SOW spells out the rest.
I also take a limited number of founder scale engines for solo consultants. Ask on the call.
ChatGPT hands you a draft in a chat window and stops there. You still do the research, check the claims, fix the voice, upload it, fill in the SEO fields, and rewrite it for LinkedIn, and next month none of what you taught it carried over. A writer does learn your business, then takes that learning with them when the contract ends. The engine keeps it: locked voice rules built from your own material, drafts from sources you approved, a quality gate before anything reaches you, publishing into your CMS with images, links, and metadata attached, and old pages brought current instead of only new ones written. Your team runs it under license after the build, so you are not renting my chat window either.
No, you don't need to be technical. The engine is a custom skill that runs in Claude Code and Cursor, the same AI tools I build software with. I set it up, and you drive it in plain language: ask for a piece, review the draft, approve it, and it publishes to your site. It runs as real files under a license, not a subscription you rent every month, so your team keeps running it long after the build ends with nothing extra to pay.
A QA layer for text written by AI. Multiple automated checks run before a draft ever reaches you: em dashes, banned words, robotic sentence patterns, claims without evidence. If a draft fails, the engine rebuilds it. You only review work that already passed, and you stay the final check before anything ships.
Google says it rewards helpful content however it gets made and goes after spam produced in bulk. The engine is built for the first and blocks the second: every piece runs on your real expertise and data, not filler. The Juniper numbers on this page came from pages the engine built, measured in Search Console, and AI engines cite those same pages as sources.
Three to four weeks for most engines. The exact shape depends on how many voices I lock and how much automation and distribution I wire in.
Either. Some clients take the keys the day training ends. Others keep me on as a fractional marketer and builder: I run production, and at the higher end the ongoing strategy too, while their team ramps up, then step back once the numbers prove out.
WordPress is proven in production, including automated Yoast and RankMath SEO metadata, FAQ schema, and image uploads through the API. The engine also works beyond WordPress: this site runs on Astro and Netlify, and I have published through fully custom apps like CasitaScore and Kinship Careers. If your stack has a way in, the engine can publish to it.
The cheapest way in is Start with Berg, $6,000 for two weeks of review and real fixes. Most engagements then move to a four week Pilot at $12,500 that proves the work on your own material, with $5,000 of that credited toward the full build if you start within 30 days. Engine builds start at $27,500, usually less than the first few months of a senior content hire. If you want me to run it after launch, that runs $12,000 to $18,000 per month with a three month minimum. The exact number comes out of the fit call.
You own every finished piece the engine publishes. The engine itself, the skill files, voice rules, quality gate, workflows, and publishing setup, stays under a license your team uses for your own business, with no monthly seat fee. Changes to the engine itself come through me. The license is for internal use, not for resale or building a competing product. Full terms live in the SOW.