Google's AI Mode Just Crossed 1 Billion Monthly Users. Your Category Just Joined The Race.

Tuesday's I/O keynote was the moment AI search became default-on at consumer scale. Google said AI Mode has surpassed 1 billion monthly active users in just over a year, calling it the biggest upgrade to the Search box in 25 years. They paired the announcement with a new model (Gemini 3.5 Flash, generally available the same day), a new shopping layer (Universal Cart, rolling out across Search, Gemini, YouTube and Gmail) and a new class of always-on Information Agents that monitor the web 24/7 on the user's behalf. The user's interface to the internet just changed. Your category's interface to the user just changed with it.

A day earlier, OpenAI quietly shipped the other half of the story by signing Dell for on-premises Codex deployment, the first major hybrid distribution deal for any frontier model in enterprise. The pair of events tells you something the press releases do not say out loud. Consumer AI is now mainstream and enterprise AI is now installable. Two different gates fell open in the same week, and any brand still treating AI visibility as a 2027 problem just lost the calendar fight by about 18 months.

TL;DR

Google's AI Mode passed 1 billion monthly active users in roughly one year, with queries doubling quarterly since launch (Google I/O 2026).

Gemini 3.5 Flash went GA at I/O, hitting Terminal-Bench 2.1 76.2%, GDPval-AA 1656 Elo and MCP Atlas 83.6%, at less than 50% the cost of competitor frontier models.

Google's Universal Cart will roll out across Search, Gemini, YouTube and Gmail this summer, putting product discovery and checkout in one persistent surface.

OpenAI and Dell announced on May 18 that Codex will run on-premises through the Dell AI Data Platform and Dell AI Factory, with 4 million developers already using Codex weekly.

Google launched Information Agents that monitor the web 24/7 for changes on the user's chosen topics, plus SynthID Verification with 50 million global verification uses to date.

1. AI Mode Passes 1 Billion Monthly Users. The Channel Is No Longer "Emerging."

May 19, 2026 | At Google I/O 2026, Sundar Pichai announced AI Mode has crossed 1 billion monthly active users in roughly one year, with queries doubling quarterly since launch. Google described it as "the biggest upgrade to the Search box in over 25 years." The new Search box now accepts text, images, files, video and Chrome tabs as input, and routes responses through Gemini 3.5 Flash with agentic capabilities baked into the response itself. Source: https://blog.google/innovation-and-ai/technology/ai/google-io-2026-all-our-announcements/

What I Think: The CMO who told me in February that "AI search is interesting but it's still niche" can no longer use that line. A billion monthly users in a year is the most decisive consumer-internet number since TikTok in 2020. The interesting detail buried in the keynote is that queries are doubling quarterly. That is not "adoption" any more, that is people learning a new behaviour and ditching the old one inside the same financial year.

Why It Matters: For every CMO across every vertical, the first audit is the same one. Pull the AI Assistant channel report inside Google Analytics that shipped last week. Compare against this time last year. If your share of AI Mode traffic looks anything like the share you held of Google Search a year ago, you are exactly where you should be. If it doesn't, you have one quarter to fix the citation surface of your top 20 pages.

2. Gemini 3.5 Flash GA. Frontier Intelligence At Half The Cost.

May 19, 2026 | Google made Gemini 3.5 Flash generally available across the Gemini API, Google AI Studio, Android Studio and the new Google Antigravity platform. The model hit Terminal-Bench 2.1 76.2%, GDPval-AA 1656 Elo, and MCP Atlas 83.6%, and runs at less than 50% of the cost of competitor frontier models on output tokens per second according to Google's own benchmark figures. VentureBeat covered the same announcement with the headline that 3.5 Flash could "slash enterprise AI costs by more than $1 billion a year" for high-volume deployments. Source: https://venturebeat.com/technology/google-says-gemini-3-5-flash-can-slash-enterprise-ai-costs-by-more-than-1-billion-a-year

What I Think: This is the price war finally arriving inside the frontier tier. Until last week the frontier conversation was about which model is smartest. Starting Tuesday it is also about which model is the cheapest per inference at scale. Marketing departments running customer-facing assistants and enterprise teams running internal copilots will both feel the impact inside 90 days, because the procurement teams just got a credible threat point against the incumbent's pricing.

Why It Matters: Two things to do this week. One, if your team is paying a per-token premium to OpenAI or Anthropic for retrieval, summarisation, or extraction work where 3.5 Flash is adequate, you have a renegotiation case. Two, if you are building a customer-facing assistant on a fixed model contract, lock in the price benefit by routing the cheap paths through Gemini and saving the expensive paths for harder reasoning. Your AI bill is now a competitive variable, not a fixed cost.

3. Universal Cart Lands Across Search, Gemini, YouTube And Gmail.

May 19, 2026 | Google introduced Universal Cart, a single shopping basket that follows the user across Search, Gemini, YouTube and Gmail. The user can add items while reading a Search result, watching a YouTube video, chatting in Gemini or scanning a Gmail receipt. Google said the feature rolls out across Search and the Gemini app this summer, with YouTube and Gmail following. Source: https://blog.google/innovation-and-ai/technology/ai/google-io-2026-all-our-announcements/

What I Think: The unification of cart across discovery surfaces ends one of the oldest battles in e-commerce. Every product page on the web has spent 20 years optimising for "add to my site's cart." Universal Cart says: it does not matter whose cart, it matters whose product. The brand that wins is the brand the user adds, not the brand whose checkout flow is best designed. For B2C marketers, that flips the priority list.

Why It Matters: Your product feed becomes the front door, not your category page. Make sure your Merchant Center catalogue, your structured data, and your product page schema all carry the attributes the agent uses to pick the item: price, availability, return policy, key feature differentiators, and verified reviews. If your product feed is stale, your category page can be perfect and you will still lose the cart.

4. OpenAI x Dell. Codex Comes On-Premises.

May 18, 2026 | OpenAI and Dell Technologies announced a partnership to bring Codex to hybrid and on-premises enterprise environments. Codex will connect with the Dell AI Data Platform that many businesses already use to govern enterprise data on-prem, and the two firms will explore Codex integration with the Dell AI Factory for production AI workloads. OpenAI separately confirmed Codex has more than 4 million developers using it weekly, making it one of OpenAI's fastest-growing enterprise products. Source: https://openai.com/index/dell-codex-enterprise-partnership/

What I Think: This is the deal that closes the last big procurement objection for regulated industries. Financial services, healthcare, public sector and defence all said the same thing for 18 months: we will deploy when the data does not leave our firewall. As of last Monday, it does not have to. The brand-side implication is bigger than it looks. If your customers are large enterprises, their developers will be deploying Codex inside their own data centres in the next two quarters. Your B2B content needs to be readable and citable by the same model their developers use.

Why It Matters: Two practical reads. One, if you sell into regulated verticals, your prospects' technical buyers now have a clean ChatGPT story they can show their CISO. Your competitive positioning needs to be visible to that buyer at the moment their procurement team runs the diligence prompt. Two, the same shift moves enterprise GEO from "be visible in ChatGPT.com" to "be visible inside the Codex instance running in the customer's own data centre." The source set still matters, the surface where the model reads you has just multiplied.

5. Information Agents Are Now A Default Feature Of Search.

May 19, 2026 | Google launched Information Agents, autonomous agents running inside Google Search that monitor the web 24/7 on the user's chosen topics and surface changes back to them. The feature rolls out to AI Pro and AI Ultra subscribers in the summer. Combined with the new Antigravity 2.0 developer platform and a $100/month AI Ultra subscription tier with 5x the usage limits of AI Pro, Google is staking out the prosumer monitoring layer. Source: https://www.seroundtable.com/google-search-gemini-3-5-flash-agents-41346.html

What I Think: The "search query" was a transactional, one-shot interaction for 25 years. Information Agents make it a standing instruction. The user no longer asks "what is happening with X" once and then forgets. They tell Google "watch X for me forever" and the model does. That changes the half-life of every piece of content you publish, because the agent will revisit, compare and re-cite. A page that is true today but not tomorrow becomes a long-term reliability risk for your brand inside the agent's source set.

Why It Matters: Two implications. One, freshness becomes a structural requirement, not a polish item. Date your pages, refresh them quarterly, and version material that genuinely changes. Two, if you do not have a way to publish updates to material claims (pricing, availability, features, capabilities) that the agent can re-read, you will accidentally teach the model to flag your brand as unreliable. The cost of a stale page just went from "lower CTR" to "negative trust signal."

6. AI Content Verification Becomes Default Plumbing.

May 19, 2026 | OpenAI announced advances in content provenance, previewing a public verification tool that lets anyone check whether an uploaded image was generated by ChatGPT, the OpenAI API or Codex. The check inspects Content Credentials and SynthID signals. On the same day, Google said SynthID has now been used for 50 million global verifications to date and confirmed cross-platform adoption with OpenAI, Kakao and ElevenLabs. Source: https://openai.com/index/content-provenance/

What I Think: This is the quiet story of the week and arguably the most consequential for brand integrity. Every brand creative team has spent two years arguing about where the line is on AI-generated imagery. The vendors just collapsed the argument by making detection trivial. By the autumn, a journalist, a regulator, or a competitor can paste your hero image into a public tool and get an instant verdict. The conversation about disclosure is over. The conversation about consistency just began.

Why It Matters: Your brand's published asset library now needs a provenance audit. Pull every hero image from the last 12 months, identify the ones generated by AI, and decide which ones get disclosed and which ones get re-shot. The cost of a "gotcha" moment in the trade press is no longer a brand risk, it is a measurable risk with a tool URL attached. Have the policy written this month, not next quarter.

Travel Desk

For travel, this week's news was the Google AI Mode hotel-booking partner list, announced at the same I/O keynote. Google named Booking.com, Expedia, Marriott International, IHG, Choice Hotels and Wyndham as launch partners for hotel booking inside AI Mode, with Google explicitly not acting as merchant of record. The partner owns the customer, the booking, and the loyalty point. The day before, the Skift x Curacity number from 11 May was still hanging in the air: 94% of hotels are effectively invisible in AI search, and on Hyatt-specific queries NerdWallet pulls 13.6% of AI citations against Hyatt.com's 10.3%. This week's Geotravel LinkedIn deep dive (18 May, tactic #6) confirmed the underlying mechanic: 44% of all AI citations come from "best-style" listicles and listicles with a comparison table in the first viewport get cited 3x more than the same listicle without one. The travel brands that built the citation surface this winter are the ones Google named on Tuesday. Full travel breakdown in this week's GeoTravel newsletter: Geotravel.ai.

The Strategic Reality Check

Three things shifted across the broad AI-search landscape in seven days.

One: the AI search channel is no longer emerging. It is mainstream. A billion users in a year, queries doubling quarterly. Your "we will revisit GEO in 2027" plan has been overtaken by the calendar.

Two: the frontier price war has started. Gemini 3.5 Flash at less than half the cost of competitor models reframes every AI spend conversation inside marketing, customer service, sales engineering and search operations. The CFO will ask the question even if you do not.

Three: the agent layer is becoming default plumbing. Information Agents, Universal Cart, content provenance and on-premises Codex are not separate stories. They are the same story told four ways. The user's interface to the internet is now an agent that watches, cites, buys, verifies and updates. Brands that publish for human readers only are now publishing to the wrong audience.

Three truths every CMO should write down this morning:

1. AI Mode is a default channel, not an experiment. Treat it like Google Search in 2008. Decide what you are willing to spend on visibility, and do it now.

2. The cost of frontier inference dropped this week. Re-open your AI vendor contracts. The renegotiation case is on the public benchmarks.

3. Provenance is no longer optional. Every brand image you publish needs an attribution decision, because by autumn every reader can check.

What You Should Do This Week

Run your top 20 brand queries inside Google AI Mode today. Note which sources Google cites, which brands appear in the answer, and how far below the answer your own site sits. That report is your baseline. Without it, your next quarterly board meeting becomes a debate instead of a status update.

Pull the AI Assistant channel inside Google Analytics for April and May. Stack it against your direct traffic for the same months. The numbers most teams find inside this report make the AI visibility budget conversation a thirty-minute decision instead of a six-month negotiation.

Re-open your AI vendor contracts this week. Gemini 3.5 Flash hitting frontier benchmarks at less than half the price of competitor frontier models is the public benchmark you need. Bring it to the next renewal call. The savings, if any, fund the visibility investment in step one.

Audit your published brand imagery for AI-generated assets. Decide which ones get disclosed, which ones get re-shot, and write the policy now. The verification tools shipped last week. By autumn the journalists, regulators and competitors will be using them.

Stop publishing static category pages this quarter. Date them. Add an update calendar. Build a quarterly refresh into the editorial brief. Information Agents will re-read your pages on the user's schedule, not yours. Stale pages stop being a CTR problem and start being a trust problem.

The interface to the internet changed this week, and so did the price of using it. Move now or move when the agents have already learned to skip you.

Think a billion users is "still niche"? Tell that to the brand whose category just folded into Universal Cart. Hit reply and explain your plan.

Want to audit your AI visibility before the next board meeting? Let us talk: https://calendly.com/vincent-getinference/

See you next week.

Vincent Chief AI Enthusiast, SimplyAI: AI SEO (GEO)