Technical SEO Audits with AI | John Caiozzo
With John Caiozzo, senior director of SEO at Tenstrat · Streamed live · 48 min
What this episode covers
Paul Andre de Vera talks with John Caiozzo about running technical SEO audits with AI, using Screaming Frog's built-in AI integration for on-page checks. They cover writing bounded prompts to control token costs, using AI for Jira tickets and structured data with guardrails, and automating meeting notes. The main conclusion is that AI handles repetitive grunt work while strategic decisions still need human judgment.
Key takeaways
- ✦Use Screaming Frog's AI integration with a focused, bounded prompt to check on-page elements like titles, meta descriptions and headings.
- ✦Send only relevant page content to the AI model, since full rendered pages consume many tokens quickly.
- ✦Verify AI audit output carefully, because hallucinations are most likely around JavaScript-rendered content the model cannot see.
- ✦Give AI explicit guardrails for structured data so it reproduces the wording on the page instead of adding details.
- ✦Let AI automate research and follow-ups, but make strategic recommendations yourself based on business context.
Key moments
- 23:55Writing Effective AI Prompts
“So this is really one of the key things here is what is the prompt going to be?”
- 25:42Common Prompt Mistakes to Avoid
“Another thing that can happen is you can be overly broad too.”
- 34:45Where AI Audits Go Wrong
“Yeah, so that the mistakes are almost always going to be around JavaScript rendering.”
- 39:16Structured Data Hallucination Risks
“With structured data, you have to be very careful about hallucinations.”
- 42:56Why Strategy Needs Human Judgment
“It doesn't understand what's going to be the most impactful thing to do on a strategic basis.”
- 43:30Where Beginners Should Start
“look at your ongoing tasks that you do, your daily things, how you can start automating them.”
Quotes are from the episode transcript, lightly cleaned of auto-caption errors.
Questions answered in this episode
- How can I run a technical SEO audit with AI using Screaming Frog?
- Use Screaming Frog's AI section to choose a platform and model, write a focused custom prompt, then crawl the site so the prompt runs on each page and export the feedback to Excel.
- What are the common mistakes when writing AI prompts for SEO audits?
- Getting too detailed by asking for many checks at once can bottleneck the work, and being overly broad without guardrails can produce weak results. Sending full rendered pages also inflates token costs.
- Can AI be trusted for structured data and audit findings?
- Caiozzo says AI needs guardrails, since it can hallucinate or add text to structured data, and its audit mistakes are mostly around JavaScript rendering, so results need checking.
More episodes on AI Search & LLMs

Keyword Research Beyond Tools for AI Search | Venkata Pagadala
Venkata Pagadala

AI SEO: How to Rank in AI Search | Ross Hudgens
Ross Hudgens

AI Search Optimization: How to Test ChatGPT and Gemini | Sanjay Singh
Sanjay Singh

SEO Reporting with AI and Search Console | Aaron Abbott
Aaron Abbott

SEO Reporting: Leading the Internal Search Narrative | Tyson Stockton
Tyson Stockton

LLM SEO: Track What AI Says About Your Brand | Bernard Huang
Bernard Huang