Generative Engine Optimization: From SEO to GEO

Overview

September 15, 2026. Cloudflare flips a switch, changing its network-wide default to block AI training bots while allowing search retrieval bots to pass through. That infrastructure tweak fundamentally changed how brands reach audiences. We have officially moved from SEO—Search Engine Optimization—to GEO, Generative Eng

In this show

• September 15, 2026. Cloudflare flips a switch, changing its network-wide default to block AI training bots while allowing search retrieval bots to pass through. That infrastructure tweak fundamentally changed how brands reach audiences. We have officially moved from SEO—Search Engine Optimization—to GEO, Generative Engine Optimization. Ranking for target keywords is out. Being the clearest existing answer a model can lift and paraphrase is in. To understand how models like Gemini, Claude, and Perplexity ingest data, and how companies are adapting, I am joined by Search Dynamics Architect, and our Generative Content Director.
• It completely depends on the bot and the platform's specific ingestion method. You have to understand that the crawler ecosystem has fragmented. Copilot is grounded by the Bing index. ChatGPT uses OAI-SearchBot for retrieval-time fetching. Claude relies heavily on the Brave index through ClaudeBot, while Gemini utilizes Google-Extended to separately control inclusion in Gemini and Vertex AI training. The crawlability of your site—clean HTML, page speed, and strict sitemap hygiene—is still the pre-condition. But the success signal has changed. We are no longer chasing time-on-page or bounce rate.
• Exactly, and that changes how we author content entirely. Look at the software industry, specifically HubSpot versus Salesforce. HubSpot engineered standalone AEO—Answer Engine Optimization—pages built explicitly as citation bait. They wrote discrete, definitional statements like, 'Cloud-based inventory management software is a system that tracks stock levels.' Salesforce, conversely, relied on a traditional glossary page for terms like CRM. Because Salesforce did not tightly author its own comparison narratives, models like Claude simply synthesized outside reviewers.
• That is the golden rule of GEO: if you do not author the comparison, the model writes one from whoever else has. HubSpot dominated the AI narrative because they built answer-first paragraphs. AI models synthesize brand identities using fundamentally different primary source types than traditional search engines. They heavily index structured data and schema markup to signal entity clarity. If your schema does not explicitly link your official profiles—your LinkedIn, Crunchbase, Wikipedia, and Wikidata using the 'sameAs' property—the model struggles to baseline its knowledge of your brand.
• Speaking of Wikidata, that is perhaps the most overlooked lever in GEO. In classic SEO, we treated Wikipedia simply as a high-authority backlink. Today, Wikipedia and Wikidata are foundational reputation engines. Now, as of March 2026, Wikipedia explicitly bans AI-generated article text, and you cannot directly edit your own brand's Wikipedia page without massive conflict-of-interest flags. But Wikidata is an open, editable structured database. You can create and edit your entry directly using a neutral account name, provided every claim has an external reference. It builds the one-line synthesis that feeds Google's Knowledge Graph.
• And those external references bring us to earned PR, which has undergone a seismic shift. Classic SEO treated all high-authority coverage as interchangeable currency. A link from CNN and a link from the AP were both just high-authority backlinks. GEO has to account for platform-specific licensing. OpenAI signed a massive fifty-million-dollar-a-year deal with News Corp, meaning the Wall Street Journal and New York Post have outsized weight in ChatGPT. Google licensed Der Spiegel and The Guardian.
• Which means your PR strategy must track outlet by-platform licensing status, not just prestige and reach. You want your brand mentioned in the publications that these AI platforms actually pay to cite. And the framing of that coverage matters immensely. You cannot just pitch narrative brand storytelling anymore. You have to push for specific, quotable factual claims in the coverage—hard numbers, named comparisons, clear positioning statements. You are optimizing for citation frequency and share of voice across AI answers, treating each media placement as a potential Wikidata source reference.
• The currency of the web is no longer the hyperlink; it is the cited fact. Three takeaways for adjusting your marketing activities. First, restructure your owned content into answer-first paragraphs with discrete, definitional statements. Second, manage your entity recognition directly by building out structured properties on Wikidata, which feeds directly into the Knowledge Graph. Third, target your PR efforts toward licensed news outlets, pitching hard, quotable data rather than vague narratives. Huge thanks to our Search Dynamics Architect and our Generative Content Director for breaking down the GEO playbook.