AI Developers Caution Against Optimization Claims, Citing Data

Image credit: Search Engine Journal

Global AI developers expressed significant caution regarding the predictability and control of artificial intelligence, directly contrasting confident optimization claims made by many consultants and practitioners.

This divergence highlights a growing gap between creators of advanced AI systems and those marketing specific, often unsubstantiated, tactics for optimizing AI interactions, according to industry observers.

Prominent AI figures, including Dario Amodei of Anthropic, Neel Nanda of Google DeepMind, and Ilya Sutskever, formerly of OpenAI and now with Safe Superintelligence, repeatedly voiced concerns about the inherent unpredictability of large language models.

Despite these warnings, numerous consultants continue to assert deterministic outcomes for AI optimization strategies, frequently citing precise percentage uplifts for methods like schema markup, content chunking, and specific citation techniques.

These claims often lack rigorous empirical testing and are frequently based on self-produced data, according to critics.

A study conducted by Ahrefs, led by researchers Louise Linehan and Xibeijia Guan between August 2025 and March 2026, found no meaningful increase in AI citations from the addition of schema markup.

The Ahrefs research further indicated a slight decline in AI Overviews for content employing such markup, challenging a widely promoted optimization strategy.

Google’s official documentation, updated on May 15, 2026, explicitly refuted several common prescriptions for AI optimization.

The company stated that practices such as using llms.txt files, content chunking, rewriting content specifically for AI, implementing special schema, and generating inauthentic mentions are not required or helpful for improving AI interactions.

Pedro Dias, a prominent voice in the digital marketing community, highlighted the situation as an industry-scale Dunning-Kruger effect, where individuals furthest from the technical workings of AI systems often exhibit the highest confidence in their optimization claims.

The disparity underscores a need for greater scrutiny and evidence-based approaches in the rapidly developing field of AI optimization.

Source: Search Engine Journal