Google rewires search around AI at I/O 2026

What happened

Per the Google AI blog, the company used I/O 2026 to position its next phase of Search as "the best of a search engine with the best of AI." The framing matters more than the marketing copy. Google is no longer treating AI as a feature bolted onto blue links. It is treating the assistant and the index as a single product.

The post collapses the distinction Google has spent two years maintaining: that AI Overviews, AI Mode, and Gemini were complementary surfaces sitting alongside classic Search. At I/O 2026, that scaffolding is gone. There is one entry point, and the model decides how much of the answer is generative, how much is retrieved, and which sources appear.

For brands that have spent the last 18 months optimising for two parallel systems, one ranked by PageRank-era signals and one ranked by model preference, the consolidation is the story.

Why it matters for your brand

The era of treating AI search as a side channel is over. If you are a CMO at a bank, a reinsurer, an industrial group, or a UN agency, your share of voice on Google is now mediated by a model that decides, prompt by prompt, whether to quote you, summarise you, or ignore you. The ten blue links were a stable distribution surface. This is not.

For financial services brands, the implication is sharpest. Compliance teams have historically signed off on web copy knowing exactly how it would render in a SERP. In a merged surface, the same paragraph can be paraphrased into an AI answer that strips qualifiers, omits risk language, or attributes a claim to the wrong entity. Brand and legal need to align on how AI-generated summaries of regulated content are monitored. "We did not say that, the model did" will not satisfy a regulator.

For multilaterals and policy institutions, the shift changes who gets cited as an authority. When Google's index and Google's model operate as one product, the model's training preferences become a ranking signal in everything but name. Institutions like the World Bank, the IMF, UNDRR, and the OECD have historically dominated topical queries on development, climate risk, and standards. They retain that advantage only if their content is structured for retrieval (clear claims, attributable statistics, schema-marked publications) rather than for human readers paging through a PDF. The PDF era of policy publishing is the single biggest liability in this transition.

For major industrial groups, the merged surface punishes thin product taxonomies. Holcim, Siemens, Schneider, ABB, and their peers publish across dozens of business units, often with inconsistent naming. A unified AI search will consolidate or fragment that brand presence based on how cleanly the model can resolve entities. If "Holcim Solutions & Products" and "Holcim Building Envelope" read as separate companies to Gemini, they will be treated as separate companies in answers. Entity disambiguation is now a marketing problem, not just an SEO problem.

For philanthropic and policy institutions, the change reframes what "thought leadership" means. A 60-page report that gets 400 downloads is worth less than a 600-word explainer that gets quoted in 40,000 Gemini answers. The unit of distribution is the claim, not the document. Foundations and think tanks that publish for peer audiences via gated PDFs will see their citation share decline in favour of organisations that publish atomised, attributable claims on indexed HTML pages.

Content strategy needs to absorb three changes at once. First, every published asset must be readable to a model: structured, claim-dense, free of "as discussed above" cross-references that lose meaning when a paragraph is extracted. Second, brand systems need to assume any sentence on your site may be quoted in isolation, which means every sentence must carry your attribution and your qualifying context. Third, measurement has to move past click-through. If your CFO is still asking for organic sessions as the proof point, you are measuring the wrong surface.

The signal in context

Google has resisted this merger for two years because it threatened the ad business that funds the company. The fact that I/O 2026 made the merger official tells you the commercial model has been figured out, or that competitive pressure from ChatGPT, Perplexity, and Anthropic's enterprise search products has forced the question. Either way, the surface that drove the majority of B2B discovery for two decades is being rebuilt from the index up.

The broader trend is convergence. Microsoft merged Bing and Copilot. OpenAI added search to ChatGPT. Anthropic added web search to Claude. Perplexity built the merged product from day one. Google was the last major player still defending the separation between retrieval and generation, and now it has stopped defending it. For anyone building a brand visibility strategy, the assumption that "SEO" and "AI search optimisation" are two disciplines is no longer tenable. They are one discipline, and the brands that restructure their content operations around that fact in the next two quarters will be cited; the ones that do not will be summarised.