Judith Lewis: AI reports look good, but break under pressure

Jul 1, 2025 · 33:21 · Webinar & Conference Session
Judith Lewis Founder · Decabit Consultancy

Key takeaways

AI language models hallucinate data by tokenizing language and filling gaps with plausible but false information, like ChatGPT fabricating monthly traffic numbers and bounce rates for a client report.
Professional SEOs have handled ambiguity for decades through trial and error with Google's search algorithms; this trained judgment lets them catch AI errors that untrained clients miss.
Create reporting templates that tie every metric back to client business goals (revenue, market share, newsletter subscribers) and use them monthly to demonstrate your strategic value, not just data.
Use AI tools and trusted platforms like SE Ranking or Semrush to gather and visualize data quickly, but always double-check findings and add human commentary before presenting to clients.
AI cannot understand your client's internal politics, business structure, or real needs; only human expertise can translate data into actionable insights that protect your job and client relationship.

Chapters

Opening: You're smarter than AI
Judith Lewis introduction and background
How AI hallucinations happen: tokenization and language gaps
Real example: ChatGPT invents traffic data and audit findings
Why human oversight catches AI errors ChatGPT misses
Creating value-focused reporting templates
AI as a tool: data crunching, not strategy
Ethical considerations and energy use
Future of AI in SEO and final message
Q&A: Double-checking AI conclusions and contextual data
Q&A: Ethical and practical deployment considerations

Q&A

If we let an AI tell the story of our SEO results, how do we double-check its conclusions without doing all the work ourselves, and do you trust AI to understand contextual data?

Don't trust AI conclusions; trust data and analysis instead. SEOs have handled ambiguity for decades and can spot problems ChatGPT can't, like identifying broken Google Tag Manager tags. Train instances with guardrails, but always double-check with multiple tools like Screaming Frog, Majestic, and Google Search Console.Judith Lewis

What are the ethical and practical considerations when deploying AI for content creation and technical optimization, especially for search engine compliance and user trust?

AI-generated code can be unreliable and introduce security risks like injected malware. Energy use is significant—AI analysis burned substantial resources with poor results. Use AI for data visualization and crunching, but always verify code and compliance manually. Experts should remain primary decision-makers, not AI, because guardrails still fail and costs (energy, errors) are high.Judith Lewis

Quotes

You're smarter than an AI. Yes, I said it. You are smarter than an AI.Judith Lewis
It did it really, really, really fast. I mean, if I didn't know better, I would be like, 'Oh my god, I'm out of a job.'Judith Lewis
We can trust SE Ranking. We cannot trust Chat GPT to give us the truth because it is not a tool. It is an LLM.Judith Lewis
No matter what your imposter syndrome says, you are better than an AI, believe it or not, because we need to use the right tools in the right way.Judith Lewis
You are the most intelligent tool in the toolbox. You're the sharpest tool in the toolbox. Don't let AI make you feel less.Judith Lewis
Schedule a demo