Most SEO teams already use AI to write content. Almost none of them can explain the system behind it.
In a recent SEJ webinar, Darrell Tyler, senior director of organic growth at CallRail, shared a statistic from his own conversations across the industry: About 85% of the SEOs he speaks with use AI for content, and only about 12% have documented systems governing that use.
This gap is the whole problem. The adoption has already taken place. What separates the teams now is whether the AI runs on foundations or is free.
Darrell has gone through the four levels that transform an AI subscription to a real benefitwhy your content reads generically without them, and the an audit that shows where your gaps are.
Watch the webinar on demand now and get the complete framework.
85% of SEOs use AI for content. 12% have a system behind it.
The adoption is settled. In Darrell’s conversations across the industry, the vast majority of SEOs are already using AI for content in one form or another. The distribution appears one level lower: only about 12% have documented systems for how this AI is actually used.
“If your use of AI is the same as your competitor’s, you don’t really have a strategy or advantage, you just have a subscription,” Darrell said.
The symptoms of an underbuilt operation are ones that most practitioners recognize. The outcome varies between team members, as each executes their own prompts. The quality degrades on a large scale: the first articles look great, then from article 97 onwards there is a visible decline because the work started to optimize registered tokens instead of business results. Publish 500 articles on a weak base and you’ve produced 500 misaligned pages on brand, not 500 wins.
Darrell named this large-scale inconsistency, invisible quality atrophyAnd optimization drift. Scaling AI without the systems to support it is not growth. This costs real traffic and real time spent correcting published work.
The first step is an honest audit of your team’s real situation. Complete the AI Maturity Audit within the on-demand session.
Why your AI content reads like everyone else’s
Why does AI content seem generic?
Because AI starts from the same blank page that your competitors use. If you write an article about what call tracking is and a competitor writes the same article with a similar prompt, you are both sending roughly the same result. Darrell calls this input “blank AI” and it’s a large part of why AI content is organically touched. This matches everything that has already been published.
The phrase he wants you to walk away with: “You can’t get out of an undocumented context by quick means. »
Rapid engineering is real, but it doesn’t save an AI that has no context about your business. The model is not the bottleneck. The platform is not the bottleneck. The operation around AI is. Without documented context, AI writes from what exists on the Internet, which is the same source your competitors come from.
Action Item: Before scaling, document the context that makes your content unique: your brand and product positioningyour first party data and the insights that only your team can provide you.
Find out what documented context looks like in practicein the on-demand webinar.
Teach the AI your business before asking it to write
What is AI Ops for SEO?
It’s the system that governs how AI produces consistent, high-quality, brand-aligned work at scale. Darrell’s framework has four layers, borrowed in spirit from MLOps and RevOps and geared toward content.
The knowledge layer is your AI’s source of truth about your business: brand and product ontologies, style guidelines, competitive intelligence, and first-party data like reviews, customer testimonials, and call transcripts. He calls this the most important layer, because it’s the one that corrects the uniformity of the AI. The AI stops writing from just the topic and starts writing from your positioning.
The workflow layer is where an individual’s capability becomes an organizational standard: SOPs, prompt libraries treated like production code, templates. The governance layer is the people side: quality assurance frameworks, review checkpoints, and feedback loops that build confidence in results over time. The application layer, the tools and models themselves, is the least important. Models are engines that you exchange when a better one ships. Your system does not change when the engine changes.
First-party data is the part most teams ignore and the part that gives the advantage. Reviews, customer testimonials, and call transcripts give the AI direct experience to write about, which is exactly what organic search rewards.
The contents of each layer, what to put in the knowledge base, how to structure workflow SOPs, and how governance checkpoints are removed as trust is established, are explained in full on-demand. Find out what’s going on inside each layer.
Stop measuring content by volume. Start measuring results.
How can we measure AI content, if not in volume? By the results it generates. A competitor can buy the same AI subscription tomorrow. They can’t buy the knowledge layer, workflows, and governance that you’ve built and iterated on for a year. This is the part that composes.
Darrell’s advice on tools is to remain LLM independent by design. Run today’s job using the best-performing model, and when the leader changes, swap out the engine, not the operation. Keep your assets, style guidelines, prompt libraries, and positioning documents living independently in a version-controlled environment rather than locked to a single platform.
The role changes with this. Less writing from scratch, less manual research, more strategy, development of knowledge layers and governance. The technician becomes a system architect.
And the dashboard changes. SEO ROI is measured by effectiveness, conversions, and revenue, not by the number of articles you have published.
Watch the webinar on demand for complete deployment, from audit to operationalized workflow.
Q&A: Most useful questions from the webinar
Q: I feed the AI with links from my site. Is this enough to build a knowledge layer?
Darrell replied: It’s a beginning, not an end. Retrieved links cover what’s already public, but the value of the knowledge layer is in what’s not on your website. He highlighted internal context like a brand manifesto, the audience you’re trying to attract, and positioning that never appears on a public page. Feed the links, then dig deeper into the context that the AI can’t find on its own.
Q: The winning prompt on ChatGPT is not the best on Claude. How can I handle this?
Darrell replied: A prompt is only half of a good result. The other half is a single context. If you have a strong sense of what greatness looks like, build on that and ask AI to help you close the gap. He argued that when you provide the same unique context, you get a more balanced result no matter which model you run, making rapid differences between platforms less significant.
Q: Beyond impressions and clicks in Search Console, how do I know if my AI content is hurting more than it helps?
Darrell replied: Go to GA4 for the page and read the engagement signals. Average engagement time and views per user tell you how content actually performs once someone lands, not just whether Google served it. His informal litmus test: Ask someone outside the work to read it, and if they have difficulty, the content is probably not strong enough.
Q: A year later, my AI content is still poor. Is it the prompts or the template?
Darrell replied: Not the model. Start with the prompt, then look more carefully at the context you gave the AI to do the job. His analogy: Ask two people to build a house, and the one who asks if you want brick or wood, who first gathers the context, brings the vision to life. He who runs away and builds immediately does not do so. Audit the prompt, but audit the context behind it, because that’s the combination that improves the outcome.
