AI search and Google are not the same: identical content, completely different sources
AI search systematically favours third-party authoritative sources while Google keeps a more balanced mix — lifting an SEO playbook across will not work.
Key numbers
- Controlled experiments across verticals, languages and query paraphrases compared AI search (ChatGPT, Perplexity, Gemini, Claude) with Google: AI search shows a systematic, overwhelming bias toward earned media over brand-owned and social content.
- Engines also differ sharply among themselves — domain diversity, content freshness, cross-language stability and sensitivity to phrasing all vary by engine.
- Pew field data cited in the paper: AI summaries appeared on about 18% of observed searches; link clicks fell to 8% when a summary was present versus 15% without, only ~1% of clicks happened inside the AI box, and ~26% of those searches ended with no click at all.
- Market data cited in the paper: by mid-2025, 34% of US adults had used ChatGPT (nearly double 2023); in July 2025 ChatGPT held ~81% of AI-chat share, Perplexity ~8%, Copilot ~5%, Gemini ~2%, Claude ~1%.
Method and sample
arXiv paper “Generative Engine Optimization: How to Dominate AI Search” (2025-09-10): large-scale controlled experiments across verticals, languages and paraphrases comparing how AI search and Google source information — media-type mix, domain diversity, freshness, cross-language stability and phrasing sensitivity. The Pew and market-share figures are secondary sources cited by the paper.
What it means for brands
Ranking well on Google does not mean being cited in AI answers. The two systems prefer different sources: AI leans much harder on third-party authority, and each engine has its own taste. To be cited you need machine-scannable content that can be lifted directly — plus coverage in the third-party sources AI trusts.
How SourceSight handles it
SourceSight runs 10 engines side by side and reports mentions and citations per engine rather than blending them into one score; the off-site list is built per market because AI's preference for third-party sources cannot be skipped.
Limits
The experiments run in a controlled setting that differs from real user query distributions; the Pew and market-share numbers are secondary citations, not measurements by this paper. Engine behaviour keeps changing, so the size of the bias shifts over time.
References
Run it on your own project
What the same prompts return across 10 engines is only knowable by running them.
Related research
What 2 million AI citations show: prompt-content match matters most, checklists matter less
Once you control for the domain, the usual AEO checklist (FAQ blocks, structured data, Core Web Vitals) reverses or collapses to zero — domain authority and prompt alignment dominate.
How engines work · 2026-10-07Whose side is the AI shopping assistant on? Change its role and the penalty for sponsored results disappears
The same sponsored listings get treated very differently once the prompt names the platform as the agent's principal — disclosure rules built for humans do not carry over.