I walked into a boardroom in Central last year—glass walls, expensive coffee, that particular Hong Kong humidity that makes your shirt stick to your back within five minutes. The CMO had just been fired. I was there to figure out why the pipeline had gone flat despite a 40% increase in marketing spend.
The team proudly presented their Q1 roadmap. It was a masterpiece of addition. New TikTok channel. New attribution software. Three new content series. A video agency retainer. A podcast. A rebrand. The slide deck was 47 pages and looked incredible.
I flipped to the last page. Pipeline: down 12% quarter-over-quarter.
"You're not marketing," I said. "You're hoarding. And the algorithm can smell it."
The Addition Addiction
Every new marketing leader I've met has the same instinct: prove value by creating. More content, more channels, more tools, more headcount. It feels productive. It looks busy. It fills the quarterly report with colorful initiative names.
But in 2026, the B2A (Business-to-Agent) economy doesn't reward noise. It rewards signal. And signal requires silence around it. Every new blog post that doesn't serve a specific purpose is just static that confuses the AI models trying to figure out what your company actually does.
When I take over a marketing function now, I don't build for the first 90 days. I cut. Here's exactly how.
Move One: The Two-Week Publishing Freeze
The first thing I do—literally day one—is stop all content production. Full stop. No exceptions. The content team usually looks at me like I've just suggested we burn the office down.
"But our publishing calendar..."
"But our SEO targets..."
"But we've already commissioned three articles..."
I don't care. For two weeks, nobody writes anything. Instead, we run what I call a Citation Share Audit.
I sit the team down and ask: "What percentage of the time do the major AI models cite us versus our top three competitors when buyers ask high-intent questions?"
They never know. They'll give me Google Search Console impressions. They'll give me brand sentiment scores. They'll give me opinions. But they won't have the number.
So we map out twenty prompts that actual buyers actually ask. Not "What is [category]?"—that's research phase. I'm talking about "Which platform is better for [specific use case]?" "How much does [solution] cost for a 200-person team?" "What are the compliance risks of [approach]?"
We run them through ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. We document who gets cited, in what position, and for what specific claim. It takes about four hours. It produces the only baseline that matters.
Last time I did this, the client discovered they were invisible on 17 out of 20 prompts. Their biggest competitor was cited as "the leading provider" on 14. The team had been publishing three articles a week for a year and hadn't realized the algorithm had decided they didn't exist.
That audit becomes the North Star for the next six months. Everything else is just decoration.
Move Two: Feed the Winners, Starve the Losers
Once the audit is done, I pull the website analytics. Not traffic—conversion. I want to know which pages actually move someone from "curious" to "qualified pipeline."
It's almost always the same pattern. Two, maybe three pages carrying 80% of the weight. Usually a comparison page ("Us versus Them"), a pricing rationale doc, a deep technical integration guide, or a specific customer case study that hits a nerve. The team always knows which ones. Ask them privately and they'll tell you immediately. But ask them what they're working on this quarter and they'll list everything except those pages.
The corporate instinct is bizarre: protect the mediocre, neglect the excellent. I've seen teams leave a high-converting comparison page untouched for eighteen months while they churn out fifty generic top-of-funnel blog posts that get 12 views each.
So I do the opposite. I identify the top two pages and dedicate the next two weeks to fortifying them. Fresh 2026 data. Updated competitive matrices. New customer quotes. Expanded schema markup so AI models can extract the facts instantly. I treat these pages like crown jewels because they are.
One client had a "Salesforce Integration" page that drove 40% of their enterprise leads. It hadn't been updated since 2022. We spent three days on it—new screenshots, updated API references, a comparison table showing exactly where their integration was deeper than the competitor's. Pipeline from that page alone jumped 30% in six weeks. No new content created. Just old content defended.
Move Three: The 30% Purge
This is where the room gets tense.
Every B2B content library I've audited has between 40 and 200 articles that are actively harming the business. Not neutral. Not "maybe useful someday." Actively harmful.
The 2021 "What is [Industry Jargon]" guides that AI Overviews have made completely redundant. The thin, 300-word top-of-funnel fluff that ranks for nothing and converts for less. The five articles that all say the same thing with slightly different titles. The legacy positioning pieces that describe a company that no longer exists.
In the LLM SEO era, these aren't harmless ghosts. They're signal pollution. When an AI model scrapes your domain to understand what you do, it ingests everything. The contradictory, outdated, low-quality noise dilutes your entity signal. The algorithm gets confused about who you are and what you're authoritative about.
So I delete 30% of the content. Redirect or kill. No nostalgia.
The internal politics are always brutal. "Jim wrote those posts!" "We spent money on that campaign!" "What if someone still finds it useful?"
I don't care about Jim's feelings. I care about the Signal-to-Noise Ratio. If a page isn't actively supporting your current positioning and driving qualified intent, it's not an asset. It's a liability wearing an asset's clothes.
Last time I executed this purge, the team was in mourning for a week. Then the metrics shifted. AI citation rates started climbing within a month. The models could finally "hear" what the company actually did because we'd stopped shouting contradictory messages from the archive.
The Pattern
Look at what these three moves have in common:
- Stop guessing. Start measuring what the algorithms actually think.
- Stop building new. Defend what already works.
- Stop hoarding. Purge what confuses the signal.
None of this requires budget approval. None of it requires new products. None of it requires hiring an agency. It requires something harder: the willingness to look like you're doing less while actually doing more.
The board won't applaud a 30% content reduction in the quarterly deck. It doesn't look sexy. It looks like retreat. But the pipeline numbers six months later tell the real story.
That CMO in Central? The one who got fired? His replacement executed this exact playbook. No new channels. No new tools. Just a freeze, an audit, a fortification, and a purge. Pipeline was up 34% in Q2.
Sometimes the most aggressive growth strategy is a pair of scissors.
— James, Mercury Technology Solutions, Hong Kong, May 2026
