AEO expert explains why traffic tanks when you 10x content — and what to do about it

My path here wasn't linear. I started as a high school psychology teacher, moved into SEO and content marketing about a decade ago, and spent years learning the craft inside agencies and in-house teams before going independent.

The shift to AI in marketing has been the most interesting moment of my career so far. A lot of what I built my practice on — traditional SEO with ranking and traffic as the goal — started changing fast. Clients' content was showing up in ChatGPT and Perplexity answers without driving a single click back. CMOs had to defend content investment in QBRs and often lacked strong answers.

The shift to AI in marketing has been the most interesting moment of my career so far.

So, I rebuilt the methodology around what I now call Content IP and Discovery Loops, frameworks designed for a world where content needs to earn both AI citations and human conversions.

I'm honestly still figuring a lot of it out in real time, alongside my clients. But that's part of what makes this moment so interesting.

How to measure AI visibility

I rebuilt how we measure content visibility. For years, we defaulted to rank tracking and organic traffic. That's still useful, but it's no longer the whole picture. Buyers now conduct a huge percentage of research inside AI tools, and Google Search Console doesn't show any of that.

So, for every client engagement, I now build what I call an AEO prompt tracking spreadsheet. We identify 60 to 150 buyer-intent prompts that a real prospect would type into ChatGPT, Perplexity, Claude, or Gemini. Then, we track if the client gets cited, which competitors the AI cites instead, and what sources the AI pulls from. We rerun it on a regular cadence and watch the trendlines.

For every client engagement, I now build what I call an AEO prompt tracking spreadsheet.

The results changed things in a few meaningful ways:

1. It gave my clients a measurable answer to a question their CEOs and boards started asking: "Are we showing up in AI?" Before this, most marketing teams were guessing.
2. It changed what we create. When you can see that AI tools are pulling from G2 reviews, Reddit threads, and analyst reports more than from your blog, your content priorities shift fast. We started investing more in earned mentions, original research, and Content IP that AI cites, instead of just publishing more SEO posts hoping to rank.

In fact, I've seen organizations improve AI citation from 8% to 40% and 6x pipelines in a quarter.

How to implement an AEO workflow

Here's the step-by-step breakdown:

Step 1: Prompt sourcing. I start with the sales team and their call recordings. I pull every question prospects ask, every objection, and every moment someone says "I was just looking into..." or "I asked ChatGPT about..." And I supplement this with support tickets, customer success notes, and Reddit and G2 threads. The goal is 60 to 150 buyer-intent prompts a human would type into an AI tool.

Step 2: AI-assisted prompt expansion. I take that seed list and use Claude to expand it. Variations in phrasing, different stages of the buying journey, and different personas asking the same underlying question. AI is useful here because the work involves pattern matching, not original thinking. I create a structured spreadsheet of prompts categorized by buying stage, persona, and intent.

Step 3: Manual citation testing. I run each prompt through ChatGPT, Perplexity, Claude, and Gemini. I record whether the organization gets cited, which competitors get cited instead, and the sources the AI pulls from (G2, Reddit, the client's blog, analyst reports, podcasts, etc.). This part is still sometimes manual because automation tools aren't reliable enough yet. I use AI to tag and categorize the results, but a human must read the answers.

Step 4: Pattern analysis. I use Claude to analyze the full dataset and identify patterns. Where is the organization winning? Where are they losing? What sources do the AI tools trust most in this category? Which competitors keep showing up and why? AI earns its keep here because synthesizing 600 AI responses into themes is a job no human wants to do by hand.

Step 5: Strategic interpretation. This part is fully human. I review the data, talk to the team, and make decisions about what to do. Are we losing because we don't have content on a topic, or because our content isn't structured for humans to read or AI to pull from? Do we need to invest in earned mentions in the sources AI is citing? Do we need to commission original research to give AI something proprietary to point to? This is the meeting where we develop strategy, and AI is not in the room for it.

Step 6: Content production with AI in the loop. Once we know what to create, I use AI for research synthesis, first drafts, structural outlines, and reformatting. Human work involves the angle, the voice, the point of view, and the editing pass that makes it sound like a real person wrote it. AI handles maybe 40% of the time investment. The 60% that remains human makes the work worth publishing.

Step 7: Re-tracking on a cadence. Every 30 to 60 days, I rerun the prompt set and compare it to the baseline. We monitor citation rate, source mix, and competitor visibility over time. This transforms the workflow into a real system instead of a one-time audit.

The whole thing replaces what used to be a vague conversation about "AI search" with a measurable, repeatable process.

Why AI supports compression and synthesis, but not conviction or taste

AI now weaves into almost every part of my work, but its role varies significantly by task.

I rely heavily on AI for research synthesis, first drafts of long-form content, competitor analysis, transcript analysis from sales and customer calls, content briefs, and much of the connective tissue work that used to consume hours. It also handles audience segmentation, voice-of-customer mining, and extracting themes from support tickets and Gong calls.

I rely heavily on AI for research synthesis, first drafts of long-form content, competitor analysis, transcript analysis from sales and customer calls, content briefs, and much of the connective tissue work that used to consume hours.

And it performs production work now, such as transforming a research report into a newsletter draft, generating prompt libraries, or rewriting in brand voice after I document it well.

Explicitly human tasks include strategy, point of view, and Content IP creation: the frameworks I build, the names we give core problems, the angles we take in thought leadership, and the judgment calls about what to publish and what to kill. AI performs poorly at these because the value lies in the conviction behind the choice. A framework only works if a real market perspective informs it. AI averages toward the middle, which is exactly what we aim to escape.

Relationships are also entirely human: the discovery calls, the workshops, the moments where I push back on a CMO about to make a bad decision. I delegate none of that.

The principle I revisit is that AI excels at compression and synthesis, while humans remain superior at conviction and taste. So I use AI to reach a starting point faster, then I spend my time on parts that require judgment.

Where AI misses the mark in content production

AI's biggest miss is content ideation. I expected AI to partner with me in coming up with sharp angles for posts and campaigns. It's not. The output is competent, generic, and slightly too clean. The good ideas still come from sales calls, support tickets, weird conference conversations, and the back of my notebook.

AI helps develop an idea once I have one, but it's not a useful source of original thinking.

The second miss is brand voice. Even with thorough voice documentation, AI-generated content in a client's voice needs a heavy human pass to sound like them. It hits the surface patterns and misses the rhythm. The soul. I expected this to be 80% solved by now. It's closer to 40%.

The third miss is real strategic judgment. Positioning, ICP refinement, pricing, and market entry. AI gives me a well-organized version of what I already know, plus a few suggestions that sound smart and fall apart under stress-testing. Strategy still requires a human who has sat in the room.

Strategy still requires a human who has sat in the room.

The fourth miss is AEO citation prediction. We can track what's getting cited and reverse-engineer patterns, but predicting what will get cited is still mostly guesswork.

And the most surprising miss: Maybe I suck at it, but AI has not made me faster at the actual writing. It's made me faster at everything around the writing. The writing itself takes me about the same amount of time, because the bar is still high and getting AI output to that bar takes roughly as long, or longer, as writing it myself.

Why content volume has become a risk

I built much of my early career on volume. Publish more, rank for more, get more traffic, win. It worked for a long time, and many marketing playbooks (mine included) quietly assumed production capacity was the constraint. If you could publish twice as much, you'd grow twice as fast.

AI broke that assumption in two directions at once.

1. AI made volume cheap. Everyone can publish 10x more now. This means volume itself is no longer a competitive advantage. The marketers chasing it are pointed straight off the edge of a cliff.
2. AI changed what gets rewarded. Both Google and LLMs pull from a smaller, more trusted set of sources. They want substance, originality, and a real point of view. Volume without those things isn't just neutral, it's actively harmful. I've watched teams tank their organic visibility by scaling AI content production, and I've watched leaner teams with sharper takes get cited everywhere.

So, that's the big risk I'm seeing right now: The rush to publish AI-generated content at scale.

It tanks organic visibility, getting flagged by Google's helpful content updates, and producing zero AI citations because no original thinking exists for AI tools to point to.

That's the big risk I'm seeing right now: The rush to publish AI-generated content at scale.

Volume without substance is performing worse than ever, and AI made that gap more visible, not less.

Why winning content requires content IP

If you want winning content, build content IP. This involves naming frameworks for your audience's core problems. Interview your 3S teams (sales, success, and support) monthly. Pipe every call into Slack, providing weekly summaries relevant to marketing.

And pick a fight. Most B2B content is consensus content. It agrees with what every analyst, competitor, and thought leader says. Impactful pieces name something others get wrong and offer a clear alternative.

You don't have to be mean about it. You do have to be willing to be wrong in public.

How content production workflows should be redesigned with AI

As far as content production, most teams still run a workflow built for 2018: Brief, draft, edit, publish, distribute. Teams bolted AI onto the drafting step and called it transformation. It wasn't.

The redesign includes three notable layers:

1. The brief layer is a structured input (not a Google doc). It includes the buyer prompt that the piece aims to win, the Content IP it supports, the point of view, and source material from transcripts and internal experts. AI assembles it. A human approves the angle before drafting begins.
2. The drafting layer uses AI heavily but with strict role clarity. AI does synthesis, structural drafting, and reformatting. A human handles the angle, the voice, and the editing pass. The human is the author throughout, not a reviewer at the end.
3. Most teams in the QA layer still focus on typos and brand guidelines. We added explicit checks for readability, brand voice fidelity, and whether the piece takes a unique position within its category. If it doesn't, we kill it before it ships. The kill rate is high, but that's on purpose.

The result is fewer pieces published, higher quality for each, and a measurable lift in both AEO citations and pipeline.

Bolting AI onto a 2018 workflow just gets you 2018-content-spam-cannon results faster.

What CMOs must do to keep up with the AI transformation

If you're a CMO and we're getting coffee, here's the advice I'd share:

1. Stop chasing AI tools and start auditing your content's substance. The teams losing right now are losing because their content has nothing distinctive for AI tools, Google, or buyers to point to. If you don't have a real point of view, no tool is going to manufacture one for you. Stop chasing the "perfect" AI stack.
2. Get serious about Content IP. Named problems, named frameworks, proprietary research, original takes on your category. AI cites this, it ranks in search, wins on LinkedIn, gets newsletter subscribers, and travels on sales calls. AI also cannot replicate this part of marketing, which makes it more valuable, not less.
3. Measure what matters now. Add AEO visibility tracking to your reporting. Track LinkedIn's impact on the pipeline. Determine the correlation between newsletter subscribers and close rates or sales velocity. Know which buyer prompts you show up for in ChatGPT, Perplexity, and Claude, and which competitors AI cites instead. If you can't answer that question in your next QBR, you'll soon be asked it.
4. Use AI where it's good and stop forcing it where it isn't. It's good at synthesis, research, transcript analysis, drafting, and reformatting. Pipe every sales call through Claude and into Slack for your team to see daily, minutes after it happens. AI is bad at original ideas, brand voice, and strategic judgment. Build your workflow around that distinction instead of trying to AI-ify everything.
5. Publish less and make each piece carry more weight. The volume game is over (though every AI content software vendor will claim otherwise). The winning teams publish fewer, sharper pieces and get more out of each one through repurposing, distribution, and earned mentions.
6. And finally, be honest with your team and your CEO about what's changing. Many marketing leaders quietly hope this blows over. It won't. Buyers have already changed how they research. The holy trinity is discovery (Google/LLMs), social (LinkedIn, YouTube) and relationship content (Newsletter/Podcast).

Follow along

You can follow Brendan Hufford's work on LinkedIn. And check out Growth Sprints.

More expert interviews to come on The CMO Club!