We Built a System That Learns Which Content Performs — Here's What We Found
By Archie Cortés, Founder of AutoPilotPR — Puerto Rico's only AI-native marketing agency. We built something we've never seen another agency build. I want to explain why we built it and what it's shown us.
Most content marketing operates on instinct and hope. A writer produces something that feels good. It gets published. Maybe it performs, maybe it doesn't. The agency moves on to the next piece.
After enough cycles of this, you accumulate a content library that has no institutional memory. No record of what worked, what flopped, or why. Every new content brief starts from zero.
We decided to fix that — for ourselves, and for our clients.
The result is what we call the Gene Registry, hosted at evolution.autopilotpr.com. It's the operational backbone of how AutoPilotPR produces and optimizes content. Here's what it is, how it works, and — more importantly — what it's taught us.
Table of Contents
- The Problem: Agencies Produce Content Blindly
- What the Gene Registry Is
- How It Works: Content as Organisms
- What We've Learned So Far
- Why This Matters for Puerto Rico Businesses
- What It Looks Like in Practice
- FAQ
- Want a Content System That Gets Smarter Over Time?
The Problem: Agencies Produce Content Blindly
Let's be direct about how most content marketing actually works.
A client brief arrives. A writer researches the topic, drafts something, an editor cleans it up, and it goes live. Traffic is monitored loosely. If a piece does well, there's a vague sense that "this topic resonates." If it doesn't, it's attributed to the algorithm or bad timing.
No structured feedback loop. No extraction of what specifically worked. No systematic testing of whether a different hook, structure, or angle would have performed better. The knowledge that could improve the next piece simply doesn't get captured.
The average content piece takes 4 hours to produce (Orbit Media Content Marketing Survey, 2024) and the median blog post generates fewer than 500 page views. Most of that effort produces content that ranks for nothing, converts nobody, and accumulates as digital weight on a domain.
This is a resource problem, a process problem, and — most fundamentally — a data problem.
What if you could treat your content library as a learning system? What if every piece of content taught you something about what to do differently next time?
That's the question the Gene Registry answers.
What the Gene Registry Is
The Gene Registry is a content intelligence system — a structured database where every piece of content we produce is logged, tracked, and evaluated based on real-world performance.
Think of it as the institutional memory for content strategy. Instead of "this type of post tends to do well," we have: "posts with this specific hook pattern, in this format, covering this category of topic, with this distribution approach achieve X% higher engagement than the average."
The name comes from evolutionary biology — specifically the idea that successful traits in organisms are encoded in genes and passed down. We apply that framing to content: successful patterns are encoded as "genes" and applied to future content. Unsuccessful patterns are deprioritized or modified.
Non-technically: it's a system that remembers what worked, figures out why, and applies those lessons to the next thing we write.
It's live at evolution.autopilotpr.com.
How It Works: Content as Organisms
Here's how we think about it operationally.
Every piece of content is logged as an "organism" when it's published. The organism record includes:
- The hook type (question, stat, counterintuitive claim, story, etc.)
- The content format (listicle, how-to, case study, comparison, opinion)
- The primary keyword and intent category
- The distribution channels used
- The author voice profile
- Target audience segment
- Internal linking structure
Fitness is scored by real engagement data. After a piece has been live for 30, 60, and 90 days, the system pulls performance data:
- Organic search ranking for target keyword
- Organic traffic volume
- Average time on page
- Scroll depth
- Conversion events (form fills, CTA clicks, contact inquiries)
- Social shares and backlinks earned
Each organism receives a fitness score — a composite metric that weighs these factors based on what we're optimizing for (usually consultation requests or organic traffic, depending on the client).
Patterns are extracted and encoded as genes. When high-fitness organisms share characteristics — the same hook type, the same structural pattern, the same topic category — those characteristics get flagged as strong genes. The system surfaces them when new content is briefed: "Posts with [pattern] have 40% higher fitness in this category."
Low-fitness organisms are autopsied. When content underperforms, the system helps identify which variables likely contributed — low keyword search volume, weak hook, thin content, poor distribution, or simply a topic that doesn't resonate in this market.
Over time, the Gene Registry becomes smarter. The more content is tracked, the more signal accumulates. The more signal, the more precise the recommendations.
What We've Learned So Far
We've been running this system long enough to surface patterns that have changed how we approach content production. Here are the honest findings:
Hook type matters more than topic.
The first line of a piece of content drives whether someone reads past the first paragraph. We've tracked enough organisms to see that counterintuitive-claim hooks ("Most Puerto Rico businesses are doing SEO completely wrong") consistently outperform question hooks ("Are you making these SEO mistakes?") in our market — by a significant margin.
The reason is probably that questions feel like quizzes. Counterintuitive claims feel like insider information. Our audience of business owners responds to the latter.
Specificity drives time-on-page.
Pieces that include specific numbers, named tools, and real case examples consistently score higher on average time on page and scroll depth than pieces written at a general level. "5-9% revenue increase" lands harder than "significant revenue increase." "68% of online experiences begin with search" lands harder than "most people use search engines."
Vague content doesn't keep readers. Precise content does.
Local signals are the highest-leverage SEO variable in Puerto Rico.
For every piece of content targeting Puerto Rico businesses, adding specific local context — municipality names, industry-specific Puerto Rico statistics, references to local regulation like Act 60 — dramatically improves both ranking and conversion. Generic content that could apply anywhere converts poorly in a market where buyers want to know you understand their specific context.
We now brief every piece of content with a mandatory local signal requirement. It has meaningfully changed ranking outcomes.
Comparison content converts at 2–3x other formats.
"Agency A vs. Agency B," "Option 1 vs. Option 2," pricing comparisons, "before and after" structures — these consistently generate more consultation requests per reader than purely informational or how-to content. The likely reason: comparison content attracts people who are closer to a decision. They're not learning; they're evaluating.
Content that doesn't link internally to a conversion point is dead weight.
We audited a full year of content from before the Gene Registry was live. Pieces with clear internal links to a contact or consultation page generated 4–5x more conversion events than pieces without them — even when traffic was similar. This sounds obvious. In practice, it's the most commonly skipped step in content production.
Why This Matters for Puerto Rico Businesses
The Gene Registry isn't just an internal tool — it's what makes our client work meaningfully different from what a traditional agency delivers.
Every piece of content we produce for a client is informed by pattern data from the full organism library. That means your hotel marketing blog post benefits from what we've learned works in restaurant marketing, law firm marketing, and professional services content — because the underlying hook and structural patterns cross-cut.
Most agencies start from zero every time. New client, new brief, new content program — all based on the writer's instincts and the account manager's gut feel. Our system means we start from a data-informed baseline, even for clients in a new vertical.
The system personalizes over time. As we produce content for a specific client and track their organism library, the system identifies which patterns perform best for their audience specifically. A restaurant's audience responds differently than a law firm's. The Gene Registry captures that distinction and adjusts future briefs accordingly.
You get compounding returns, not just deliverables. Traditional content marketing produces content you own. Our system produces content you own plus a growing understanding of what works for your specific business and audience. That institutional knowledge becomes a competitive advantage that's hard to replicate.
Puerto Rico's market is small enough that ranking advantages matter enormously. There are only so many people searching "hotel marketing Puerto Rico" or "law firm Puerto Rico" in any given month. The business that ranks for those terms and converts readers efficiently captures a disproportionate share of that demand. The Gene Registry exists to make that happen faster.
What It Looks Like in Practice
When a client engages AutoPilotPR, the Gene Registry is part of their content program from day one.
Every piece of content produced goes through a briefing process that surfaces relevant patterns from the organism library. Writers receive structured briefs that specify hook type, format, required local signals, internal link targets, and the specific conversion outcome the piece should drive.
After 30, 60, and 90 days, each piece's fitness score is calculated and logged. Patterns that emerge from high-performing content are encoded and applied to the next content cycle. The program gets smarter with each piece published.
Clients see this reflected in their monthly reports — not just "here's how your content performed" but "here's what we're applying next month based on what we learned." That's the difference between an agency that produces and one that learns.
You can explore the system at evolution.autopilotpr.com.
FAQ
Q: Is the Gene Registry something clients can access directly?
Currently, the Gene Registry is an internal intelligence system with a public-facing overview at evolution.autopilotpr.com. Client-specific reporting is delivered through monthly performance reports that surface relevant Gene Registry insights for their program. Direct client access is on our roadmap.
Q: How is this different from just looking at Google Analytics?
Google Analytics tells you what happened — traffic, bounce rate, sessions. The Gene Registry tells you why it happened and what to do next. It connects content characteristics (hook type, format, structure) to outcomes (rankings, conversions) in a structured way that raw analytics data doesn't capture. It's the interpretive layer that turns data into decisions.
Q: How long does it take for the system to generate useful patterns?
With a new client, we draw on the existing organism library from day one. Client-specific patterns typically emerge after 8–12 pieces of content with performance data. By month 4–6, we have meaningful client-specific signal that meaningfully shapes the content program.
Q: Does this work for service businesses in Puerto Rico, or only tech companies?
The Gene Registry was built specifically with Puerto Rico service businesses in mind — hotels, restaurants, law firms, real estate agencies, professional service firms. The local market signal component is particularly valuable in Puerto Rico, where generic content strategies fail to account for the bilingual, local, and regulatory-specific context that buyers expect.
Q: Can other agencies or businesses use a similar approach without AutoPilotPR?
In principle, yes — the methodology is systematic content tracking tied to real performance data. In practice, building and maintaining the organism library, scoring system, and pattern extraction requires infrastructure that most businesses and agencies don't build. That's why we built it as a proprietary system rather than a spreadsheet. We've invested significantly in making it work reliably at scale.
Q: What happens to the content assets if we stop working with AutoPilotPR?
All content produced for clients is owned by the client. If the relationship ends, you keep every blog post, landing page, and content asset. What you don't retain is access to the Gene Registry pattern library — that stays proprietary. But the content itself, and whatever rankings and traffic it's generated, is yours.
Want a Content System That Gets Smarter Over Time?
Most content marketing is a cost. We built a system that turns it into an asset — one that compounds, learns, and improves with every piece published.
If you're a Puerto Rico business that wants content marketing that actually works — and gets more effective over time — we should talk.
Talk to AutoPilotPR about your content strategy →
Explore the system: evolution.autopilotpr.com | Learn more: AutoPilotPR Marketing Agency | SEO for Puerto Rico