Netlinking and visibility in generative AIs are often managed separately, even though both strategies frequently target the same sites. Speakers at the conference on GEO (Generative Engine Optimization), Getfluence and Resoneo, propose designing a single third-party article to be ranked on Google and cited by AIs. In many companies, netlinking, i.e., obtaining links from other sites, is tracked based on the link and position on Google, while AI visibility is tracked based on citations and share of voice. Companies thus sometimes pay twice for the same asset: the third-party article.
ChatGPT relies heavily on Google, but not exclusively
According to a Resoneo study published on Search Engine Land, covering 1,200 ChatGPT responses, 88,000 results, and 26,900 pages analyzed in July and August 2026, about 75% of results retrieved in Thinking mode, the paid version, come from Google. The rest comes from OpenAI’s own index, named Labrador. In the free version, the proportion reverses: Labrador provides 74.7% of results, and news, via Oxylabs, 22%. OpenAI’s index also overlaps very little with Bing, as only 1.5% of its URLs appear in Bing’s top 20. Therefore, one must both rank well on Google and enter OpenAI’s index.
Citations are also becoming more concentrated. Across all responses studied, 61,332 URLs were surfaced, but only 759 pages were actually opened, and these are cited in 74% of cases, compared to 7% for others. According to Resoneo’s work on Meteoria data, the average number of domains cited per response has dropped from 19 to 15 since GPT-5.3 Instant in March 2026. An Ahrefs study from May 2025 on 75,000 brands places web mentions of the brand as the top correlation with visibility in AI Overviews (0.66), ahead of brand-name link anchors (0.53). These are correlations, not causal links.
Place the brand in the first third of the article
Generative engines rephrase the user’s question into several secondary queries, called query fan-out. A Surfer study covering 173,902 cited URLs indicates that a page ranked for the main query and at least one sub-query has a 51.2% probability of being cited in Gemini, compared to 19.6% for a page ranked for neither. The speakers therefore recommend structuring the article with one sub-query per subheading and answering from the first sentence.
Position within the text also matters. According to Kevin Indig, 44.2% of citations come from the first third of the page, compared to 24.7% from the last third. The speakers cite the case of a third-party comparison regularly featured in B2B SaaS, where the brand ranked first gains 16.5 points of AI visibility. They advise getting ranked in others’ lists rather than writing your own and avoiding relegating the brand to a final sidebar. Brand anchors, however, remain marginal in campaigns, at 3%, 9.5%, or 21% according to Getfluence 2026 reports, even though they constitute the second signal most correlated with AI visibility in the Ahrefs study.
Measure at the right pace
Google and AIs do not read an article at the same pace. In the OVHcloud case, analyzed by Oncrawl from Cloudflare logs in September 2026, the article is indexed and ranked on the day of publication, but the ChatGPT-User bot only visits at D+3, its visits peak at D+7, and the first mention of the brand occurs at a median delay of D+11. An initial assessment against competitors is only possible at D+20. The speakers acknowledge that these observations, based on two articles and a single brand, outline a trend rather than a rule. Measuring too early risks wrongly concluding that GEO doesn’t work.
For a campaign run in 2026 by Getfluence for an insurer specializing in decennial insurance, combining links and nine AI-First™ articles, the average position of priority keywords moved from 33rd to 2nd place between December 2025 and August 2026. On ChatGPT, visibility increased from 10% to 54% for ‘devis décennale’. Each month, iteration focuses on articles and attributes not yet covered.
