Top SEO Myths Debunked With Real Data on Ahrefs Podcast
Key Takeaways
The "SEO is declining because AI is rising" narrative relies on a false zero-sum assumption. This mirrors past mistakes (e.g., apps vs. the web). AI usage can grow significantly without the web or SEO declining in parallel.
Many claims about SEO traffic collapse are based on weak self-reported survey data. Asking users to estimate past search or AI usage is methodologically unreliable compared to panel-based traffic data.
Large-scale traffic data shows only a modest decline, not a collapse. Analysis of panel data from Similarweb covering 40,000 major U.S. sites across 2024–2025 showed total organic traffic down ~2.5%, not the commonly cited 25%, with results directionally validated against internal and Google data.
AI introduces new user behaviors and changes how visibility works, rather than simply replacing search. A joint OpenAI and Harvard study found roughly a third of AI prompts reflect entirely new behaviors (e.g., writing, self-expression).
In AI systems, visibility matters beyond the "top three" results because models draw on many citations and core training signals, making rankings noisier and broader than in traditional search.
Summary
Ethan Smith — Co-Founder & CEO of Graphite — joins Tim Soulo for a 60-minute walkthrough of debunking the biggest myths in the SEO/AEO field.
I looked at 40,000 sites, and of course, some sites went up, some sites went down. Total traffic is down, but it's down 2.5% overall and not 25%.
Ethan Smith, CEO of Graphite
You'll hear why the rise of ChatGPT isn't the "Google Killer" everyone predicted, how small websites are actually gaining traffic in the AI era, and the exact strategy you need to ensure your brand is cited by LLMs.
The conversation also digs into the flaws of AI visibility tracking, including why small sets of custom prompts introduce bias and why self-reported prompt volume data remains too noisy to trust. Ethan shares practical alternatives, from converting search keywords into prompts to mining Reddit and sales calls for real questions, and explains how both core model training and RAG citations shape whether your brand appears in answers.
