Google I/O 2026: what the AI search overhaul means for your GEO

Google I/O 2026 changed the AI search rules. On May 19, 2026, Google announced that AI Mode crossed 1 billion monthly active users, up from 100 million in late November 2025, a 10x increase in roughly six months. AI Overviews, the summaries that sit above organic results, now trigger on approximately 48% of all tracked search queries, a 58% year-over-year increase. For marketing managers, growth leads, and fractional operators who depend on organic search as a pipeline channel, the implication is immediate: most of your target queries now surface an AI answer before a blue link. If your brand is not named inside that answer, you are invisible.

This post covers what actually changed at I/O 2026, what it means for Generative Engine Optimization, and the six specific moves Xpand Media is shipping for clients right now. Xpand Media is a Dubai-based AI growth agency that runs GEO, performance marketing, and AI automation as one integrated motion for B2B SaaS, DTC, and services brands across MENA and APAC.

What did Google announce at I/O 2026 that affects search visibility?

Google called I/O 2026 the biggest transformation in Search in more than 25 years. Three announcements matter for GEO practitioners. AI Mode expanded from a US-only paid feature to nearly 200 countries and territories across 98 languages with no subscription required. AI Mode queries are more than doubling every quarter since launch. Gemini 3.5 Flash is now the default model globally, which means a faster inference loop and more queries routed through AI Mode rather than traditional results.

The third announcement is the biggest signal for GEO: Search agents. Google is launching information agents that operate in the background 24/7, proactively retrieving information and completing tasks on behalf of users. The first wave targets AI Pro and Ultra subscribers this summer, but the architecture is clear. Google described it as entering the era of Search agents. That means less reactive query-response and more proactive brand retrieval. The brands those agents retrieve from their index will be the brands that built citation authority before the agents launched. That window is now.

What do AI Overviews do to organic click-through rates?

The data is unflattering for publishers but clarifying for GEO. When an AI Overview appears, only 8% of users click a traditional search result, down from 15% without one, a relative drop of 47%. Google offsets this with a different claim: AI Overviews drive over a 10% increase in overall search usage for query types where they appear. More queries, fewer clicks per query. The net effect for most B2B content teams is negative on traffic. The counter-move is citation, not avoidance.

If your brand gets named inside the AI Overview, not below it, that changes the math. Users follow named sources at a higher rate than they click anonymous blue links, which is why LLM referral traffic converts at 15.9% from ChatGPT and 10.5% from Perplexity compared to a 1.76% conversion rate from organic search. Less total volume, much higher intent. The GEO service Xpand runs is built around earning that named citation, not chasing the click below the AI answer.

How did the rest of the AI search landscape shift in May and June 2026?

Google was not the only engine moving. Perplexity launched the Comet browser worldwide for free, with built-in AI answers that increase user questions per session versus standard browsing. Perplexity also unveiled a hybrid local-cloud inference orchestrator at Computex 2026, a system that decides autonomously which AI workloads run on-device and which route to frontier models. The practical outcome: Perplexity gets faster and more available on low-connectivity devices, expanding its reach into mobile-first markets including MENA and Southeast Asia.

On the ChatGPT side, the May 2026 update increased direct linking to external websites inside AI-generated responses, with multiple business case studies reporting measurable increases in referral traffic from ChatGPT citations. AI search now represents 9.3% of total AI crawler traffic, up from 7.5% the prior month. Optimizing for ChatGPT alone misses the share held by Claude, Gemini, and Perplexity, so a per-engine strategy beats a single-engine bet.

Why does the 11% citation overlap between ChatGPT and Perplexity matter for your content strategy?

The most important number in the GEO statistics landscape right now: there is only an 11% domain overlap between ChatGPT and Perplexity citation sources. ChatGPT uses Bing's search index. Perplexity uses its own vector indexing. Google AI Overviews draw from Google's index. These are different retrieval systems with different freshness criteria, different domain authority weightings, and different preferences for content structure.

Getting cited in one engine does not give you coverage in the others. A GEO strategy that optimizes only FAQPage schema for Google AI Overviews will miss the Bing index signals ChatGPT requires. A strategy that submits the sitemap to Bing Webmaster Tools but ignores the vector indexing cues Perplexity prefers will win on ChatGPT and underperform on Perplexity. The GEO vs SEO comparison breaks down the per-engine differences in detail. The section below covers which moves address which engines.

What are the six GEO moves Xpand Media is shipping after I/O 2026?

Xpand's GEO playbook prioritises six moves in response to the I/O announcements, based on what delivers the fastest citation lift given the specific changes Google, Perplexity, and ChatGPT shipped in May and June.

1. Add a schema callout block to every top landing page

Most sites we review lack a schema callout block on the homepage. A schema callout is a 1 to 2 sentence, AI-extractable definition that sits at the top of the page, marked up as a SpeakableSpecification in your Article schema. Google's AI Overviews pull this block first when building summary answers. Adding one is a fast, low-effort win for Perplexity citation. The format is specific: one sentence that names the entity, names the service, includes one citable number or threshold. No setup. No hedging.

2. Build FAQPage schema across at least three pages

FAQPage schema is not a new move, but the I/O 2026 updates raised its priority. Pages backed by FAQPage schema across a cluster are cited noticeably more often than a single optimised page. The engine needs cluster coverage, not a single entry point. Every service page on the Xpand GEO service ships FAQPage schema by default. The FAQPage schema snippets template takes under an hour to implement per page.

3. Submit sitemaps to Bing Webmaster Tools, not just Google Search Console

Most growth teams submit to Google Search Console and ignore Bing. ChatGPT search uses Bing's index, so Bing indexing is a prerequisite for ChatGPT citations. This is the most commonly skipped step in Xpand's audits and the one with the fastest time to impact. Submit your sitemap at Bing Webmaster Tools, verify the domain, and request indexing for your top 10 pages manually. ChatGPT citation frequency for those pages should improve within 14 to 21 days.

4. Rewrite answer capsules to match AI Mode's conversational pull format

AI Mode now carries conversational context between follow-up questions. When a user asks a follow-up from inside an AI Overview, the engine pulls from the same page that generated the first answer. Your opening 70-word capsule needs to anticipate the likely follow-up structure, not just answer the primary query. In our experience, a tight 70-word answer capsule is frequently pulled verbatim into AI Overviews when queried in AI Mode. The capsule format: definition sentence, then a specific number or threshold, then a named outcome. The GEO Content Checklist walks through the capsule format with before and after examples.

5. Update your llms.txt file to reflect the new AI agent endpoints

Google's Search agents will read site-level metadata to decide which domains to pull from proactively. The llms.txt standard sits at the site root and declares your entity, your services, and your key pages in a format AI crawlers can parse in one pass. With Search agents now confirmed for summer 2026, the window to update your llms.txt before agents start building their default retrieval lists is closing. Xpand's llms.txt Starter Template is the fastest path if you don't have one.

6. Run a per-engine citation audit and weight your content by engine mix

Given the 11% citation overlap between ChatGPT and Perplexity sources, a single content piece cannot carry GEO for your full target query set. Xpand runs a 30-prompt citation audit across Google AI Overviews, ChatGPT search, Perplexity, Claude, and Microsoft Copilot at the start of every engagement. The audit reveals which engines already cite you, which cite competitors, and which query types show no brand citations at all. That gap map drives the content calendar. One new piece per underserved engine per month, optimized to that engine's citation preferences, compounds faster than generic content aimed at all engines at once. Book a GEO strategy session to run the 30-prompt audit on your domain.

What does this mean for the AI Overviews versus organic search debate?

The debate is settled for informational and commercial queries. TechCrunch's headline from May 19, 2026 was 'Google Search as you know it is over'. Navigational and transactional queries still resolve to blue links in a large share of cases. The practical implication for Xpand clients: stop treating AI Overviews as a threat to be managed and start treating them as real estate to be claimed. An AI Overview citation is higher-intent traffic, at a better conversion rate, with less competition than a position-3 blue link in a saturated SERP. The state of GEO 2026 data backs this.

The counter-argument is volume: fewer queries produce clicks at all when an AI Overview answers the question. That is true. In our experience, volume decline from AI Overviews on top-of-funnel informational queries is real and is largely not recoverable through SEO. The recoverable pipeline lives in the citation, not the click. Build the schema, write the capsule, run the FAQ schema, and earn the named mention inside the answer. The click follows when the user wants to go deeper. That is the new funnel.

AI Mode queries are more than doubling every quarter since launch. We are entering the era of Search agents.

Where should you start this week?

1. 1Run the 30-prompt citation audit on your domain: ask ChatGPT, Perplexity, Google AI Mode, Claude, and Microsoft Copilot 6 queries each that your target persona would ask. Note which competitors are named and where you are absent.
2. 2Add a schema callout block to your homepage and your top service page. One sentence: entity name, service name, one specific number. Mark it up as SpeakableSpecification in your Article schema.
3. 3Submit your sitemap to Bing Webmaster Tools if you have not. Manually request indexing for your top 10 pages.
4. 4Check that FAQPage schema exists on at least three pages. If it exists on only one, expand it to two more pages this week.
5. 5Update or create your llms.txt file. Add or update the services and key pages sections to reflect your current offering.
6. 6Re-run the 30-prompt audit in 14 days and compare citation counts. The moves above should show measurable lift in Perplexity within 14 days and in ChatGPT within 21 days.