Brands Recommended by AI: 27% Vanish the Next Day
A study of 210 shopping prompts shows that 84% of AI recommendations change daily, and 27% of brands disappear the next day. Traditional SEO no longer works — your brand needs GEO.
Taqi Molavi published a study on X covering 210 shopping prompts. The results: in ChatGPT and Google AI Mode, 84% of brand recommendations change every day, and 27% of brands that show up on day one disappear completely by day two. Meanwhile, AI Overviews now cover 85% of shopping searches.
What does this mean? A brand's position in AI search engines isn't a stable asset — it's a constantly fluctuating state. A brand that ranked in the answer yesterday may be completely ignored today. Evaluating brand visibility using the results of a single ChatGPT search is basically self-deception.
Taqi Molavi's recommendation is this: rebuild your knowledge graph structure and your site's facts. That requires a new optimization methodology — traditional keyword density and backlink counting no longer apply. Generative search engines digest pages like human experts do; they look for verifiable brand entities, authoritative citation networks, and clear semantic structure. Without these signals, AI engines will skip right over your page when generating an answer.
To that end, Taqi Molavi released two tools.
The first is the Molavi Web Intelligence Engine (MWIE), available at molavi.pro/tools/web-intelligence. It's a completely free deterministic analysis engine that consumes no LLM tokens and doesn't rely on any proprietary API. It can audit websites for technical SEO, brand entity recognition, and GEO (generative engine optimization). The tool doesn't require a credit card and has no usage cap. All computation runs on the open-source MWIE core, using DOM parsing and mathematical signals.
The tool covers four dimensions: entity disambiguation, citation attribution, direct Q&A modules, and high content-to-HTML signal density. The architecture diagram is below:
The four pillars illustrated:
The second is the MCP GEO Server, open-sourced at github.com/tmolavi/mcp-geo-server. It's a Model Context Protocol server that exposes the SAGE audit engine to AI agents. Tools like Claude Desktop and Cursor can call it directly to run a comprehensive audit — including technical SEO, AEO (answer engine optimization), and GEO readiness — on a specified URL. It provides MCP tools such as audit_url, technical_seo, and geo_readiness, and also offers a Python interface for easy integration into your own workflows.
For readers who aren't familiar with the technical details, put simply: if you have a website and want to know how it performs in AI search engines like ChatGPT and Perplexity, MWIE lets you test it for free. If you're a developer who wants your own AI agents to audit websites automatically, the MCP GEO Server plugs straight in.
Amid the rapid churn of AI recommendations, brand visibility may become the new traffic bottleneck. A single optimistic test result isn't worth trusting — you need continuous monitoring and systematic optimization. The tools are already open source, and the testing entry point is free. All that's left is to act.
发布时间: 2026-11-20 08:34