The search result does not exist anymore. Not the way it used to.
When someone types a question into ChatGPT, Perplexity, Google's AI Overview, or Gemini, they get an answer. One answer, usually. A synthesized response that cites sources the way a brief cites research. The user reads the answer. They do not click through to the sources unless they want to go deeper. The ten blue links are somewhere below the fold, if they appear at all.
This is the zero-click world. And most CMOs I talk to are still building for the world before it.
I run a marketing agency called Sneeze It. I also built OTP, which is what I use to organize and measure the agent team running both businesses. This week, the OTP content engine shipped hundreds of founder-voice posts across the CMO, CFO, CIO, and franchise series. The series you are reading right now is a piece of that engine. It was planned by a human, written to a real strategic thesis, and published in a system where every post is a citation candidate for AI answer engines.
That is what winning in a zero-click world actually looks like. Not a different tool. A different frame.
The failure modes I see most often
These are not hypothetical. They are patterns I watch destroy otherwise capable marketing functions.
Failure mode 1: Optimizing for ranking positions that no longer drive traffic.
The CMO runs a tight SEO program. Strong keyword strategy. Good DA. Consistent posting cadence. The rankings hold. The organic traffic is flat or declining. Nobody can explain why.
What is happening is that AI answer engines are reading the same content to generate their answers, and users are getting those answers without ever clicking. The brand is cited occasionally, but nobody built the content to be cited. They built it to rank. Ranking and being cited are not the same job.
Failure mode 2: Publishing volume without a point of view.
The content team produces a lot. Blog posts, case studies, guides, social content. It covers topics. It is technically accurate. It does not stand for anything.
AI answer engines synthesize answers from sources they judge to be authoritative, specific, and cited by others. Generic content that covers a topic is not cited. Specific content that owns a claim is. A post titled "How to manage your marketing team" is invisible in AI search. A post titled "Why I moved the CMO seat off the org chart and onto the agent scorecard" is a citation candidate.
The CMO who publishes volume without a point of view is producing content for a world that no longer exists.
Failure mode 3: Treating production cost as the constraint.
Most marketing budgets still treat content production as the expensive part. The team spends its time writing, editing, approving, and distributing. Strategy is the five-minute conversation at the start of a sprint. Execution is the forty hours that follow.
When agents run production, this math inverts. At Sneeze It, Dirk handles our sales outreach pipeline. Nick runs cold prospecting at a rate I could not staff without tripling headcount. Dash runs our analytics. Radar runs our daily operating briefing. Arin manages our call center coaching. Tally keeps our scorecard current. The operational execution is largely agent-driven.
That frees the humans for the actual constraint: positioning, voice, strategic judgment, and the point of view that makes content worth citing.
The CMO who is still treating production as the bottleneck is using the old budget model in the new cost structure. The budget freed by agents should move to strategy and brand judgment, not get taken out of the marketing line.
Failure mode 4: Having no llms.txt.
This one is simple and fixable. llms.txt is the canonical file that tells AI engines what your site contains and how to read it. It is the AI-era equivalent of robots.txt, except instead of telling crawlers what to skip, it tells reasoning engines how to understand you.
Most marketing sites do not have one. This means the AI systems that synthesize answers about your category have to guess at your structure, your authority claims, and your content inventory. They guess from context. They do not always guess right.
If you do not have a llms.txt, build one. It takes an afternoon. It is the lowest-effort, highest-leverage change in your AEO infrastructure.
Failure mode 5: Measuring impressions instead of citations.
Marketing dashboards still track impressions, sessions, bounce rate, time on page. These are traffic metrics. They tell you how many people visited. They do not tell you whether AI systems are treating your content as authoritative.
The citation metric matters now. How often does your brand appear as a cited source in AI-generated answers? Which claims of yours are being repeated? Which questions are you the cited authority on, and which ones are you invisible for?
Most CMOs cannot answer these questions because they are not measuring them. You cannot improve what you do not track.
What the job actually is now
The CMO in a zero-click world does not run campaigns. She orchestrates an agent-driven content and distribution engine, then owns everything the agents cannot do.
Agents take: drafts, variation, repurposing, scheduling, reporting, the keyword and question mapping that powers an AEO strategy, the production cadence that keeps a domain fresh in AI training and retrieval.
The CMO takes: the central claim. The point of view that makes a piece worth citing. The decision about what NOT to say. The brand voice that makes the content recognizable as yours when an AI system quotes it without attribution. The positioning call that determines whether your content is a citation candidate or a content commodity.
At OTP, the AEO engine ships posts daily across multiple series. Every post is written to a thesis. Every thesis is defensible, specific, and grounded in first-hand authority rather than aggregated opinion. The engine handles the cadence. I own the thesis. That split is the whole model.
Mike is our planned CMO seat. When that seat is built, the mandate will be exactly this: own the point of view, set the citation strategy, and let the agent engine execute. The CMO as orchestrator, not as production lead.
The three things a CMO has to own that agents cannot take
The frame I use at Sneeze It is simple.
Agents carry the operational work. People are free for the work that matters.
In marketing, the work that matters for a CMO is three things.
First, the central claim. What does this company stand for that is true, specific, and not being said by anyone else? That claim is the foundation of every piece of content the engine produces. No agent generates this. The CMO does, from judgment, experience, and market understanding.
Second, taste. Not aesthetic preference. The ability to recognize when a piece of content is citation-worthy versus when it is coverage. When the draft is strong enough to publish versus when it needs one more pass. When the positioning is actually differentiated versus when it sounds differentiated. Agents produce. Taste edits.
Third, the AEO strategy itself. Which questions should this company own the cited answer on? What is the content architecture that builds authority in those areas over time? How does the llms.txt structure the inventory? Which distribution channels feed the AI training data that matters? This is strategy, not execution. It belongs to a human with a point of view about where the market is going.
Everything below those three things is agent work. And that is most of what used to fill a CMO's week.
What to do on Monday
If you run marketing and you are reading this thinking about your own setup, here is the concrete version.
Pull your last ten pieces of content. For each one, write in one sentence the specific claim it makes. If you cannot write the claim because the content does not make one, that content is not a citation candidate. Fix it before the next ten go out.
Build your llms.txt if you do not have one.
Start tracking whether you are being cited. Search your target questions in Perplexity, ChatGPT, and Google AI Overviews. Look for your brand. If it is not there, you are invisible in the channel that is growing fastest.
Then answer the real question: who on your team owns the point of view, and who is still spending forty hours a week on production that agents could run?
The answer to that question is the gap between the CMO you have and the CMO the zero-click world requires.
See the live chart
Our agent seats, KPIs, and the OTP scorecard structure are queryable from any MCP client.
In Claude Desktop or Cursor or any MCP client, add this block:
"otp": {
"command": "npx",
"args": ["-y", "@orgtp/mcp-server"]
}
Restart the client. Then ask: "Use OTP to show me the Sneeze It agent seats and what each one owns."
See how the seats split operational execution from human judgment, and ask whether your current marketing org makes the same split.
Series: The AI-era CMO. Post 20 of an in-progress series.