When Content Costs Almost Nothing to Produce, What Is Marketing Actually For?

By Victor Hoang · Updated Oct 11, 2026

For twenty years, a lot of marketing strategy rested on a quiet assumption: content is expensive, so producing a lot of it is an advantage. A company that published three useful articles a week beat the one that published three a quarter. The cost of a page was the moat.

That moat is gone. A competitor with a modest budget can now publish a hundred competent articles a day. So can you. And so can the thousand companies behind them. If your marketing plan is still measured mainly in pieces shipped, you're competing on the one input that just got close to free.

This isn't a reason to stop marketing. It's a reason to ask what marketing is for when production is no longer the hard part. Our view: marketing's value moves from making content to making things worth believing. That means original data, a point of view someone could disagree with, distribution you own or earn, and trust built over time. Those four are the scarce part now, and AI can't mass-produce any of them for you.

Key Facts

• Graphite's analysis of 55,400 English-language web articles, published in May 2026, found that AI-generated articles reached 36% of new articles within 12 months of ChatGPT's launch and 48% by 24 months, and have stayed near 50% for the last five quarters.
• The same firm's October 2025 study of Google results found that 86% of articles in the first two pages of results were written by humans and 14% were AI-generated. The share cited by ChatGPT with web search was 82% human and 18% AI.
• Google's spam policy lists using generative AI tools to generate many pages without adding value for users as an example of scaled content abuse, and says the policy applies no matter how the content is created.
• Pew Research Center found that in March 2025, US adults clicked a traditional result link in 8% of visits when Google showed an AI summary, against 15% of visits when it didn't.
• In LinkedIn's 2024 summary of its research with Edelman, 75% of decision-makers and C-suite executives said a piece of thought leadership led them to research a product or service they weren't considering. Fewer than half rated the overall quality of what they read as good, and only 15% called it very good.

Why Volume Stopped Being an Advantage

Two things happened at once. The supply of content exploded, and the attention it competes for stayed flat.

On supply, Graphite's May 2026 analysis found that the share of new web articles that are primarily AI-generated climbed to about half and then stopped climbing. That plateau is the interesting part. The firm says it doesn't know whether those articles earn comparable traffic, and it hypothesizes that practitioners found AI-only articles don't perform well in search. Treat that as a hypothesis, not a result. Its measurement also relies on AI detectors, which have error rates, and Graphite is a growth agency with its own view. Still, the direction is clear. The web now has far more competent text than anyone can read.

On the demand side, what readers do with that flood is changing. Pew tracked the browsing of 900 US adults in March 2025. When Google showed an AI summary, users clicked a traditional result in 8% of visits. Without one, the figure was 15%. The summary answers the easy question for them, which means the generic explainer you'd have published to catch that search is exactly the content that gets absorbed and not clicked.

Put those together. More supply, fewer clicks to the average page, and a platform that has said in writing what it thinks of mass-produced pages. Google's spam policies name scaled content abuse, meaning many pages made mainly to manipulate rankings, and list generative AI used to produce many pages with no added value as an example. The policy doesn't ban AI. It says the method of creation doesn't decide the question. What decides it is whether the page helps someone.

That is the practical test for volume: if the page would exist in the same form for any competitor who typed the same prompt, it adds nothing, and it will be treated that way by search engines and by readers.

What Marketing Is For When Production Is Cheap

Strip away the production work and marketing has always had three jobs: get found, be believed, and make buying easier. AI collapses the cost of the first draft. It doesn't collapse the cost of any of these.

Google's own guidance on helpful content puts the quality bar in questions. One is whether the content provides original information, reporting, research, or analysis. The page also says that, of the qualities it describes, trust is the most important. That's a search engine telling you which assets it can't get from a model alone.

We'd name four scarce assets.

1. Original data. A language model can summarize what is already published. It can't tell you what your 2,000 customers did last quarter, what your support tickets say about onboarding, or what you found when you tested three pricing pages. Data you generate yourself is the one input competitors can't copy by prompting, and it's the one thing other writers have to cite you for. Your customer conversations are the richest unused source most companies have, as we argue in why customer conversations are an underused data set.

2. A point of view. AI output trends toward the average of what's been written. A real position, one that names who is wrong and what you'd do instead, is the opposite of average. It also helps with trust, because a view you can disagree with is a view someone stood behind. We cover the executive side of this in thought leadership for CEOs.

3. Distribution. Content nobody sees has no value, whatever it cost to make. When search sends fewer clicks, the channels you own and earn matter more: an email list, a community, a speaking circuit, partner audiences, a sales team that forwards your work. Cheap content makes distribution relatively more expensive, because everyone's competing to be seen.

4. Trust. This is the one that compounds, and the one that can't be bought in a batch. The LinkedIn and Edelman 2024 research is a useful reminder of the stakes. Three-quarters of senior buyers said a single piece of thought leadership pushed them to look at a product they hadn't considered, and 70% of C-suite leaders said a piece at least occasionally made them question whether to stay with an existing supplier. Content that earns attention can move real purchasing. The same research found the quality of what buyers read is mostly unimpressive, which is the opening.

Where the Money Moves

If production is cheap, budget should follow the scarce parts. Here's how the shift looks in practice.

The last row has support from the field. In the Content Marketing Institute and MarketingProfs 2026 B2B trends research, fielded in mid-2025 among 1,015 B2B marketers, 75% of the more mature teams tracked business impact for thought leadership, compared with 63% across all respondents. The report's own conclusion was that the teams doing best weren't the ones producing more content or chasing the algorithm. It also found that 95% of B2B marketers say their organizations use AI-powered applications, so the tooling itself no longer separates anyone.

The Operator Question: Where Does AI Still Earn Its Place?

None of this says to stop using AI in marketing. It says to use it where it removes cost without removing the thing readers came for.

AI is good at: repurposing a webinar into five formats, producing first drafts from an expert's outline, localizing, summarizing research for a brief, generating variants for testing, and handling the formatting work around a finished piece. A well-run process there can cut the cost of each piece sharply. We lay out one version of that for SEO in the AI content operator model.

AI is weak at: knowing what's true about your business, taking a risky position, noticing what your customers said last Tuesday, and having a reputation. Those stay human, and they're where the budget you save should go.

A simple test for any AI-assisted piece: What does this contain that a competitor couldn't have produced with the same prompt? If the answer is "nothing," either add the missing ingredient (a data point, a customer story, a position) or don't publish it.

A Decision Framework: Is This Piece Worth Publishing?

Use this before anything ships, whether a person or a model drafted it.

A piece that fails three or more of these shouldn't ship, however cheap it was to make. Cheap isn't the same as free: every weak page spends a little of your credibility, and credibility is the asset you're trying to protect.

A Worked Example (Illustrative)

These numbers are invented to show the logic, not to benchmark anyone.

A B2B software company publishes 40 AI-assisted articles a month at roughly $150 each in editing time, or $6,000. Organic traffic is flat. Sales says the blog never comes up in deals.

The marketing lead proposes cutting to 12 articles a month, with heavier editing at about $200 each, and spending the rest of the same $6,000 differently:

The company ends up with fewer pages and more things a journalist, analyst or buyer would want to quote. The measures change too: citations from other sites, newsletter growth, and the number of deals where a prospect mentions a specific piece. Output volume drops, and that's the point. Nobody on that team is graded on pieces shipped anymore.

The tradeoff is real. Fewer articles means fewer long-tail searches covered, and some of that traffic may be lost. The bet is that the traffic you'd lose was the generic kind that AI summaries are already absorbing.

What to Do Next

For a CEO, CMO, or head of growth, here's a 90-day sequence.

1. Audit your content by uniqueness, not volume. Sort the last six months of output into "only we could have written this" and "anyone could." The second pile tells you what to stop.
2. Stop reporting output counts as a result. Replace "articles published" with citations earned, subscribers, branded search, and deals where content was mentioned.
3. Start one standing data program. Pick a source you own (product usage, support tickets, customer survey, sales calls) and publish what it shows on a fixed schedule. If you're unsure where to look, begin with what your buyers are asking AI and what your customers say in their own words.
4. Name your experts and protect their time. Two hours a month from a real practitioner beats ten extra drafts. Give them the byline.
5. Write down your point of view. Three positions your company holds that competitors don't. If you can't, that's the real marketing problem, and no tool will fix it. Our piece on brand strategy for CEOs is a place to start.
6. Build a hedge against search. An owned audience is the one channel an AI summary can't intercept.
7. Put a person on verification. Cheap drafts raise the odds of confident errors going live. Assign a named reviewer to every number.

The same logic applies beyond marketing. When AI makes the routine part of a function nearly free, the value of that function moves to judgment, evidence and relationships, which is why we keep writing about undocumented know-how as the real bottleneck. And as with support, where we ask whether it should be the first team automated, the question isn't whether to use AI. It's which part of the work customers pay attention to, and whether you've protected it.

Learn More

• The Verification Tax: Why AI Time Savings Shrink Once Someone Checks the Work
• Why Customer Conversations Are the Most Underused Data in Your Revenue Engine
• Thought Leadership for CEO: Building Your Voice and Industry Influence
• Brand Voice Guidelines for AI-Generated Content