How a Two-Person B2B Content Team Can Score Topics Before Writing a Single Word

Most B2B content teams spend 60-70% of their production budget on articles that never compound. We've watched it happen across dozens of editorial calendars: ten posts go out, two drive meaningful traffic growth, and the other eight sit there consuming crawl budget and editorial hours that could've gone somewhere better.

The fix isn't publishing more. It's also not another round of on-page optimization on existing posts. The actual unlock is killing bad ideas before they reach a writer's screen.

A two-person team with $150/month in tooling can build a topic pre-qualification model that filters out low-ROI ideas at the ideation stage. We estimate this cuts wasted production spend by 30-40% per quarter, based on what we've observed when teams stop treating every keyword as equally worth pursuing.

Here's the thinking behind the model, the three signals that power it, and how to wire the whole thing into a weekly editorial workflow.

Why Traditional Topic Selection Fails Small Teams

The standard playbook goes like this: pull keywords from Semrush or Ahrefs, sort by volume, eyeball difficulty scores, pick topics that "feel" right. Maybe run them past a manager for gut-check approval.

This process has two structural problems. First, search volume is a lagging indicator. By the time a keyword shows meaningful monthly volume, your competitors have already published on it. Second, difficulty scores don't account for the new citation layer that AI Overviews and LLMs have introduced. A keyword might look achievable in traditional SERPs but be completely locked down in AI-generated answers.

The result? Teams produce content that ranks on page two, never gets cited by an LLM, and generates a slow trickle of traffic that doesn't justify the $300-800 it cost to produce. Multiply that by eight posts a month, and you're looking at $2,400-6,400 in quarterly waste on content that was doomed at the idea stage.

The Three Signals That Actually Predict Compounding Returns

We've narrowed the pre-qualification model down to three publicly available signals. Each one captures a different dimension of topic potential, and together they give you a score that's far more predictive than volume-plus-difficulty alone.

Signal 1: AI Overview Citability (Weight: 40%)

This is the most important signal, and most teams completely ignore it.

Check whether an AI Overview appears for your target keyword. If one does, look at what's being cited. Research from Onely shows that LLM citation preference follows a clear hierarchy: topic match, recency, position, and content completeness are the dominant factors. Formatting has negligible impact, which means you can stop obsessing over bullet points and schema markup for this purpose.

The recency signal is particularly telling. Content published within the last two years dominates AI Overview citations, with nearly half coming from the most recent year alone. So if you're scoring a topic, ask: is there an AI Overview? What format does it cite (long-form guides get roughly 45% of informational citations)? And can you realistically publish something within the recency window that LLMs prefer?

Score it simply. AI Overview exists: 1 point. Your brand is already cited: 0.5 bonus. Format matches your production capability: 0.5 bonus. No AI Overview at all: 0 points, and you should think hard about whether this topic has a future.

Signal 2: LLM Citation Velocity (Weight: 35%)

This is where things get interesting. Static citation presence tells you where things are. Citation velocity tells you where things are going.

B2B buyers are already deep into LLM-assisted research, and AI-referred visitors spend significantly more time on-page than traditional search visitors. That behavioral difference matters because it signals higher intent and better downstream conversion potential.

Brand mentions carry weight here even without backlinks. LLMs give authority to brands that get talked about, not just linked to. A topic pre-qualification model should track three things: how many unique brands are being cited for a given topic, the month-over-month growth in those citations, and whether your brand currently appears in the set.

We use a rough threshold. Citation growth above 15% month-over-month signals a topic with expanding opportunity. Below 5% means the citation set has stabilized, and breaking in will be expensive. Between 5-15% is situational; it depends on the other two signals.

Signal 3: Keyword Velocity (Weight: 25%)

Absolute search volume is almost useless for predicting compounding returns. Velocity, the rate at which search interest is changing, is far more predictive.

Predictive SEO models analyze query velocity alongside seasonal patterns and social media activity to forecast where demand is headed. A high keyword velocity for a specific term signals that you should be creating content on that topic now, before the competition catches up.

Google Trends remains one of the best free tools for measuring this, especially for identifying emerging search demand. But you need to look at the slope of the trend line, not the absolute interest score. A keyword going from 200 to 600 monthly searches in 90 days is a much better bet than a stable keyword sitting at 5,000.

Score this one on growth rate. Over 100% growth month-over-month gets the maximum score (breakout territory). 30-100% growth is solid. Under 30% means the topic is either mature or declining, neither of which is where a two-person team should be spending limited cycles.

Building the Actual Scoring Model

Forget building something fancy. A Google Sheet works. We're serious.

Create a row for each candidate topic. Add columns for each signal's sub-scores, then a weighted total. The formula is straightforward:

Total Score = (AI Overview Score × 0.40) + (Citation Velocity Score × 0.35) + (Keyword Velocity Score × 0.25)

Normalize each signal on a 0-to-3 scale so the math stays clean. Your maximum possible score is 3.0, minimum is 0.

Set the decision threshold at the 40th percentile of your scored topics for any given batch. If you're evaluating 15 topics per week, the top 6 move to production. The bottom 9 get shelved. Not deleted. Shelved. You'll re-score quarterly, because signals shift.

One thing we'll be honest about: the weights we've suggested (40/35/25) are a starting point based on what we've seen work. Your mix might need adjustment. If your brand already has strong LLM citation presence, you might weight Signal 2 lower and Signal 3 higher. The calibration happens over 4-6 weeks of comparing your predicted winners against actual performance.

The Weekly Workflow for Two People

This model is useless if it takes half your week to run. Here's how we'd structure it for a team of two.

Monday (1 hour, one person): Generate 10-15 topic candidates. Sources: customer questions from sales calls, competitor content gaps, trending subtopics in your niche subreddits or Slack communities. Don't over-think this step. Volume of ideas matters here because the model filters for you.

Tuesday (30 minutes, one person): Score each topic. Pull up Google for AI Overview checks, run keyword velocity through Google Trends and your SEO tool, check LLM citation data through free monitoring tools. Enter scores into the sheet.

Tuesday afternoon (15 minutes, both people): Review the ranked list together. Discuss any borderline topics. Pick the top 4-6 for the week's production queue. This conversation should be fast; the numbers do most of the talking.

End of month (1 hour, both people): Review which published topics actually drove citations, traffic growth, and conversions. Adjust signal weights if needed. This feedback loop is what turns a static scoring model into a calibrated one.

Total time investment: about 2 hours per week. That's less than most teams spend in a single editorial planning meeting that produces no framework and no accountability.

What This Costs

The tooling stack is cheap.

Google Search Console and Google Trends are free. A solid B2B content research workflow can run on Semrush's free tier or SE Ranking at $55-100/month. AI Overview monitoring can be done manually in a browser (it's tedious but functional) or through Ahrefs' free tier. LLM citation tracking has several free-tier options available now. And the scoring sheet lives in Google Sheets or a basic Airtable base.

Total: $75-150/month. Add $20 for a Zapier or Make.com automation that pings Slack when a tracked keyword breaks out, and you're still under $200.

Compare that to the cost of producing even one wasted article. If your fully loaded cost per post (research, writing, editing, publishing) is $400, and you kill just two low-ROI topics per week, that's $3,200/month in avoided waste. The tooling pays for itself in the first week.

Where This Model Breaks

We'd be doing you a disservice if we pretended this works perfectly.

It doesn't account for strategic topics that matter for brand positioning but have zero search demand yet. Sometimes you need to write about something because it establishes authority in a space nobody else is covering, and that won't score well in a velocity-based model. Keep a separate "strategic bets" slot in your editorial calendar, maybe 1-2 posts per month, that bypasses the scoring model entirely.

It also struggles with highly seasonal topics where velocity data can be misleading. A topic spiking in January might be a seasonal pattern, not genuine emerging demand. The monthly calibration step catches some of this, but not all.

And the LLM citation data layer is still maturing. The tools for tracking it are getting better fast, but the data isn't as clean as traditional search metrics. Expect some noise in your Signal 2 scores for the next 6-12 months.

The Compounding Math

Content marketing ROI compounds over time, with long-term averages reaching several multiples of initial investment. But that compounding only kicks in for topics that actually gain traction. Every low-ROI article in your archive is a missed compounding opportunity, not just a sunk cost.

If a two-person team publishes 12 posts per month and this model filters out the bottom 30% at the idea stage, they're now producing 8 high-potential posts instead of 12 scattered ones. Production cost drops. Quality per post goes up (because you're spending the same editorial hours on fewer pieces). And the compounding curve steepens because every published article has a higher probability of gaining citations and traffic.

That's the real ROI of pre-qualification. Not the money saved on unwritten articles, but the faster compounding on the ones you do write.

The teams that figure this out in 2026 will look back in 18 months and wonder how they ever ran an editorial calendar without it. And the ones that keep spray-and-praying with volume will keep wondering why their content budget keeps growing while their traffic stays flat.