AI Writes Half the Web. Original Expertise Is Where Content Marketing Wins.

Nearly half of the new articles published online are written mostly by AI. Based on research conducted in 2026, Graphite, an SEO firm, analyzed 55,400 randomly sampled articles from Common Crawl, the open archive of the web, and ran them through three AI detectors. AI-generated and human-written articles were roughly 50/50. A separate Ahrefs study of 900,000 new web pages found that 74.2% contained AI-generated content.

Most notably, this AI-generated content marketing hasn’t been paying off. Orbit Media's 2026 survey of 1,042 content marketers found that 92.4% now use AI for blogging. Yet only 14% report "strong results," six points below the previous low in the survey's history. In Orbit's words, AI use "correlates with nothing."

Speed to produce content increased. Results decreased. 

AI content produces more of the same

Large language models are excellent at summarizing existing information. That creates a problem for content marketing. 

If your team and three competitors ask the same AI tools about the same topic, you will likely get four versions of the same answer. 

Google’s guidance for creators emphasizes two questions:

• Does the content offer original information, reporting, research, or analysis?

• Does it show firsthand expertise and depth of knowledge?

Volume alone does not make content valuable. Graphite researchers found that AI-generated articles, despite being published at high volume, rarely appear in Google or ChatGPT results. They suggest this may be why the growth of AI content has flattened.

Writing capacity is no longer the constraint. The competitive advantage is creating useful information that does not already exist online.

The Proprietary Input Test

Before a piece of content gets budget, ask one question: could a competitor produce this with a prompt?

If the answer is yes, it's a commodity. If the answer is no, it's because the content draws on at least one of three inputs AI can't generate on its own:

1. Original data. Customer surveys, original research, order patterns, support tickets, return reasons, usage data, benchmark results.

2. People only you have access to. Your product engineers, category managers, technical sales reps, and your customers.

3. Experience you've lived. The launch that failed, the test that surprised you, the specification that took three rounds to get right.

Content that passes this test earns links, citations, and mentions because it's the original source. 

Three content formats that perform

These are the formats I've seen perform, across B2B business solutions and B2C consumer products.

1. Buying guides written by product experts

AI is good at writing generic "how to choose" articles. A buying guide created by a product designer or category manager addresses specifications that matter to end users, common mistakes or misperceptions about the category, and leaves out marketing filler. Put their name, title and short bio on the guide. Quote them directly. Include the tradeoffs they would explain to a customer.

2. Webinars led by topic experts

The more specific the topic, the better. A webinar on a specific regulatory change, an installation problem, or a new barcode format may draw a smaller audience than a broad topic, like industry trends, but it attracts the right people who are actively dealing with that topic. Plus, the questions they ask can become inspiration for your future content.

At a niche wholesaler with more than 500 product categories, we hosted 20+ customer workshops, webinars, and events each year. Sales used these events to build relationships. Marketing focused on learning, firsthand, the obstacles customers were facing, in their own words.

Those conversations revealed:

• The topics customers cared about most
• The issues that connected customers across industries
• The problems our content and solutions needed to address

For example, safety and regulatory compliance were consistent concerns across nearly every customer segment.

3. Surveys on a timely industry issue

Original research gets cited because it gives people data they can reference and quote. Orbit Media found that marketers who publish original research improve their likelihood of strong results by 50%.

You don't need a research firm to get started. You need:

1. A clear question the industry cares about
2. A customer or prospect list to survey
3. The discipline to report the full results (not just the data that supports your message)

At a consumer goods manufacturer, we built a brand community of more than 5,000 members to capture customer feedback, product ideas, and conversations about our products.

Because members played a role in product development as a part of this community, they became our strongest brand advocates. Unprompted, they answered customer questions, addressed concerns, and defended the brand in online conversations. Their loyalty came from having a voice in shaping the products we produced.

The community increased our Net Promoter Score by 40% and became a reliable source of customer insight that competitors could not easily replicate

Where AI fits in content marketing

Using AI the right way and for the right type of work is critical to getting the results you want.

Use people to provide the expertise, experience, and original insight. Use AI to organize, analyze, draft, and repurpose that material.

In practice, that looks like:

• Capture: Record an expert interview, customer conversation, or webinar. AI can transcribe it and identify the strongest quotes, examples, claims, and statistics.

• Analyze: Use AI to group open-ended answers to survey questions into themes. It can flag patterns worth investigating, while a person validates the data and conclusions.

• Draft: Have AI create a first draft using only the interview transcript, webinar recording, or survey data. The expert then reviews it for accuracy, context, and missing details.

• Repurpose: One webinar becomes a long-form article, product-page FAQs, a LinkedIn post series, video clips, and a sales enablement one-pager.

Writing used to be the most expensive part of content production.

Now, the expense is one hour of an expert’s time. With AI handling the structure and production, one hour can provide original material for weeks of content.

What to measure instead of content volume

If your goal is to build authority, do not focus only on how much content you publish. Track the signals that show people trust and use your expertise.

• Earned links and citations: Track when trade publications, partners, industry sites, and other credible sources link to your content or cite your research.

• Mentions in AI answers: Ask ChatGPT, Gemini, Perplexity, and Google's AI Overviews the questions your buyers ask. Track whether your brand, research, or subject-matter experts appear in the answers.

• Branded search: Watch for more people searching for your company by name. That is a sign your expertise is reaching the market. 

• Sales use: Are sales reps sending your buying guide, webinar, research, or other content to prospects? If they do, that's a strong signal the content helps answer real buyer questions.

What this means for marketing leaders

Content budgets built for production capacity and writing volume need to shift toward access to expertise.

That requires three changes:

1. Budget for expert time, not word count. Set aside time quarterly with product, engineering, technical sales, and other subject-matter experts. Treat that time as a content investment, just like media spend.

2. Fund one original research project a year. Pick a question your industry is actively debating, gather the data, and become the company that provides a credible answer.

3. Redefine the content team’s role. The content marketer's role is transitioning from writer to editor, interviewer, and research producer. AI handles the drafting, while people provide judgment, expertise, and accountability.

When half the web is written by the same tools, generic content becomes the default. The advantage goes to companies that publish insight only they can provide.

What valuable knowledge does your company have that has never been published?

AI-assisted research and drafting; independently reported, verified, and edited by the author.