How to Get ChatGPT to Recommend Your Business (Full GEO Tutorial)
This tutorial explains Generative Engine Optimization (GEO), a new discipline focused on making websites more visible to AI language models like ChatGPT, Claude, and Gemini. The presenter covers both on-site techniques (robots.txt configuration, schema markup, FAQ pages with authoritative citations) and off-site strategies (reviews and brand mentions on high-authority platforms). The video also demonstrates how to use Claude connected to Google Analytics via Zapier MCP to track LLM referral traffic.
Summary
The video opens with the claim that 60% of consumers now consult ChatGPT before making purchasing decisions, establishing the premise that traditional SEO must now be complemented by GEO (Generative Engine Optimization) — optimizing content for large language models (LLMs) such as ChatGPT, Perplexity, Claude, and Gemini.
The first on-site technique covered is ensuring LLM crawlers can access your website by checking and configuring the robots.txt file. The presenter shows how to visit a site's /robots.txt path and, if using a custom file, explicitly allow crawlers from ChatGPT, Perplexity, Claude, and Google. WordPress sites handle this automatically by default.
The second technique is schema markup written in JSON-LD format. Because LLMs read raw HTML code rather than rendered pages, schema markup 'spoon-feeds' structured information to crawlers. The presenter demonstrates using an AI prompt (directed at Claude) to analyze a website URL and auto-generate 6–10 FAQ pairs in JSON-LD format. This should be done for every page on the site, and a dedicated FAQ page is strongly recommended.
Three content-writing tips are provided to improve GEO rankings: (1) Use expert quotations attributed to credible sources, since LLMs treat attributed quotes as more authoritative and citable. (2) Include specific statistics and data points rather than vague claims, as quantifiable information is easier for LLMs to extract and cite. (3) Cite authoritative external sources like Gartner, Harvard Business Review, or government data to signal content reliability. The presenter also advises writing each paragraph as a standalone unit rather than using first-person pronouns like 'we' or 'our,' since LLMs extract content in discrete blocks.
For off-site GEO, the presenter highlights two strategies: accumulating reviews on platforms like Trustpilot, B2B review sites, and Google; and generating brand mentions on high-domain-authority platforms that LLMs actively crawl, such as Reddit, LinkedIn, and YouTube. Notably, Instagram and TikTok are excluded because their login walls prevent LLM crawlers from accessing content.
Finally, the presenter demonstrates tracking LLM referral traffic by connecting Claude to Google Analytics via Zapier MCP. After setting up the integration, users can prompt Claude in natural language to retrieve and visualize referral traffic data from LLMs — eliminating the need to manually navigate Google Analytics. The presenter admits his own LLM referral traffic is currently low but commits to improving it over the next 6–12 months.
Key Insights
- The presenter argues that GEO differs fundamentally from SEO in that it relies on 'attributions' and 'brand mentions' rather than backlinks — LLMs don't require a hyperlink back to a site to recognize and cite it, just a mention of the brand name.
- The presenter claims that Instagram and TikTok do not contribute to GEO rankings because their login-wall architecture prevents LLM crawlers from accessing content, making LinkedIn, Reddit, and YouTube the superior platforms for GEO off-site strategy.
- The presenter advises that website copy should be restructured so every paragraph reads as a standalone unit, avoiding first-person pronouns like 'we' and 'our,' because LLMs extract and digest content in discrete paragraph-level blocks rather than reading full pages holistically.
- The presenter demonstrates that Claude can be connected to Google Analytics via a Zapier MCP integration, allowing users to query their analytics data in plain natural language and receive reports or charts — a workflow the presenter frames as saving 'hundreds of hours' compared to manual Google Analytics navigation.
- The presenter contends that including specific, attributed statistics (e.g., citing a named study with a precise percentage) dramatically increases the likelihood that an LLM will extract and cite a website's content, because quantifiable data points give the model something concrete to reference.
Topics
Full transcript available for MurmurCast members
Sign Up to Access