Scaling AI content is the strategy. The strategy is the problem.

Conductor's 2026 State of AEO/GEO CMO Investment Report surveyed 250+ executives across 12 industries and asked them what their top AI search content priority was. The answer, across every maturity tier, was scaling AI-generated content. Above structured data. Above original research. Above authoritative long-form.

That should worry you more than it seems to be worrying enterprise marketing leaders.

Because in the same report, the experts Conductor brought in to comment on the data — Aleyda Solis, Eli Schwartz, Lily Ray, Pedro Dias — all said variations of the same thing: this is a bad idea, it's already failing, and Google has started issuing manual actions against the sites doing it most aggressively. The strategy the CMOs ranked first is the strategy the practitioners ranked as a likely cause of catastrophic visibility loss.

That's not a disagreement. That's a coordination failure dressed up as a survey result.

What the report actually says

Conductor's headline finding is that scaled AI content production is the dominant AI search investment across maturity levels. Schwartz's read is that leaders are "somewhat skeptical about the effectiveness of mass amounts of AI content, but are afraid of being left behind if they don't do this."

Read that again. The strategy isn't being chosen because it works. It's being chosen because not choosing it feels worse.

This is what fear-driven procurement looks like when it makes it into a board deck. The CMO doesn't need to believe scaling AI content will work. They need to be able to say, when asked, that the organisation is doing something. Doing something is the deliverable. Whether the something is correct is a downstream problem for whoever inherits the traffic graph in eighteen months.

The Mt. AI graph

Glenn Gabe coined the term "Mt. AI" for the shape these sites print in Search Console: a sharp ramp as Google ingests a flood of new URLs, a plateau as freshness signals temporarily prop them up, then a cliff edge when Google's quality threshold catches up.

The CMOs in the Conductor report are buying a graph that peaks before their next performance review and craters after it.

Pedro Dias documented that from June 2025 onwards, Google started issuing manual actions for "aggressive spam techniques, such as large-scale content abuse" — explicitly aimed at sites mass-publishing AI output. UK, US and EU sites all received Search Console notifications. This isn't speculation. This is enforcement that's already running.

Dan Taylor's analysis nailed the mechanism. The problem isn't that the content is AI-generated. It's that the content was published without an editorial floor underneath it. The freshness boost masks the absence of strategy for a quarter or two. Then the assessment fires and the whole pile gets reweighted.

The CMOs in the Conductor report are buying a graph that peaks before their next performance review and craters after it.

Why "scaling" is the wrong word

The word "scaling" implies that what you'd do at small volume is fine, and the challenge is multiplying it. That's not what's happening. The thing being scaled isn't a working content programme. It's the absence of one.

I've watched this pattern play out at every size of business over the last 18 years. Someone discovers a production lever — outsourced writers in 2014, content spinners in 2016, programmatic SEO in 2020, AI in 2024 — and the leadership response is identical. Press the lever harder. Don't audit the output. Wait for the traffic.

The lever changes. The mistake doesn't. The mistake is assuming volume is the bottleneck, when the actual bottleneck is judgement. AI scales the production step and leaves the judgement step exactly where it was, which was already understaffed.

When Shelley Walsh writes that she can get "astounding results from Claude when I input quality, unique research, but I do have to invest a huge amount of guidance to get anything worth publishing" — that's the whole story. The model is a force multiplier on the editorial judgement you put in front of it. If there's no judgement to multiply, you've just built a faster way to produce indistinct content.

The competitive logic is backwards

Here's the part of the Conductor framing that breaks if you push on it. The implicit pitch is: scale AI content or fall behind competitors who are scaling AI content.

But if every enterprise in the category is scaling AI content using broadly similar models with broadly similar prompts against broadly similar topic lists, what you produce is a flat landscape of interchangeable pages. No site in that landscape has an advantage. They've all done the same thing. The category as a whole has become noisier, which means the signals that actually differentiate — original research, first-party data, distinct points of view, named author authority — get *more* valuable, not less.

The arbitrage isn't doing what everyone else is doing faster. The arbitrage is doing the thing the AI-scaled competitors can't do because their editorial floor is on the ground.

Lily Ray's line — "just because it's easy doesn't mean it's a good idea" — does a lot of work here. Easy is the giveaway. Easy means non-differentiating. If the lever you're pulling is one any competitor with a corporate card can pull tomorrow, it's not a strategy. It's a tax.

What the gap actually rewards

The competitive position worth building right now isn't "we publish more." It's "we publish things our competitors structurally can't."

That includes original research — proprietary surveys, internal data releases, longitudinal analysis on something only you have access to. It includes named subject-matter expertise — bylined pieces from people whose work is independently verifiable elsewhere. It includes points of view that take a position the rest of the category is too cautious to take. It includes assets useful enough that a senior person at a target account would forward them to colleagues — the Rand Fishkin "drop into Slack with a note" test.

None of this scales the way an AI content programme scales. That's the point. The asymmetry is what gives it value. If you can produce twelve pieces a year that meet that bar, you'll out-compete a competitor producing twelve hundred that don't. The Conductor report frames this as a maturity question, but it's really a discipline question. The mature programmes aren't the ones publishing more. They're the ones publishing less of better things.

What this means if you're sitting in the room

If you're the marketing lead being shown a deck that recommends scaling AI content production as your AI search strategy, here's the question that usually breaks the slide.

Ask what the editorial review process looks like at the proposed volume. Not "do you have one" — every deck claims to have one. Ask how many human hours of senior review per published piece, who that senior reviewer is, and what their review criteria are. If the answer is vague, or the hours-per-piece figure is under one, you don't have an editorial floor. You have an aspiration.

Then ask what happens when Google's next quality assessment fires. The agency's answer will tell you whether they've read the same enforcement data the rest of us have, or whether they're selling you the pre-June-2025 version of the playbook.

The honest version of this strategy isn't "scale AI content." It's "use AI to handle the parts of production that don't require judgement, so the humans can spend more time on the parts that do." That version produces less output. It also produces output that's still indexed in two years' time.

The Conductor data shows the industry has chosen the other version. That's a problem if you're inside the industry. It's an opportunity if you're competing against it.