Google I/O 2026 opens with Gemini Intelligence and Googlebook, ChatGPT shifts ads to cost-per-click bidding, Snap kills Perplexity's $400m AI search deal, Coty hands Consumer Beauty content to Pencil

Welcome to the Daily Marketing Brief. Your daily AI news and tactics for marketers who move fast. I'm your host Jen Bryan and here's today's update. The week ahead is set up to be the most consequential for marketing technology in 2026 so far.

Google I.O. opens today morning California time. Google Marketing Live lands Wednesday.

Chat GPT is now running cost per click ads against Google search inventory, and Snap has quietly walked away from a $400 million AI partnership that was supposed to anchor its default chat experience. None of that is theoretical. All of it sits inside this quarter's media plan. In today's briefing, four stories that should shape how a sharp operator approaches the next 10 days.

We open with Google I.O. because the keynote is 24 hours away and shapes everything Google ships through Q4. We cover OpenAI's switch to cost per click pricing inside ChatGPT, which puts a real per-click number on AI search advertising for the first time.

We look at Snap's amicable breakup with perplexity and what it tells us about the limits of platform scale AI partnerships. And we close with Cody's tie-up with Pencil, which is the most credible enterprise Gen AI content deployment we have seen in beauty to date. First up, Google I/O 2026 opens today with Gemini Intelligence, Google Book on Aluminium OS, Android 17, and Android XR Glasses. What happened?

Google I.O. 2026 opens today at 10 a.m.

Pacific time at the Shoreline Amphitheater with the developer keynote at 1 p.m. Pacific. Twelve days ago, on May 12th, Google streamed the Android Show I.

O. Edition and formally previewed four big strands. First, Gemini Intelligence, the new umbrella term for Google's best AI features across Android phones, watches, cars, glasses, and laptops, with the first wave landing on the latest Pixel and Samsung Galaxy this summer. Second, Google Book, a new premium AI laptop category running a merged Android and Chrome OS platform called Aluminium OS, with hardware arriving this autumn from Acer, Asus, Dell HP, and Lenovo.

Third, Android 17 with generative UI, Gboard, Rambler voice input, deeper Chrome on Android integration, and a refreshed Material 3 expressive design across Android Auto. Fourth, Android XR glasses developed with Samsung, Warby Parker, and Gentle Monster due late 2026. The main keynote is expected to land a new Gemini model. Analysts split between 3.

2 Flash and 3.5, alongside developer tooling and the Agentic Ads framework that lands at Google Marketing Live on Wednesday. What is confirmed? The Android show announcements are on Google's own blog and developer pages.

The Tuesday keynote, the Wednesday Marketing Live slot, and the Shoreline venue are confirmed. The new Gemini model name and tier are not yet confirmed. Tech Times reported on May 17th that internal expectations are the new model, lands roughly at the level of OpenAI's GPT 5.5 and meaningfully short of Anthropic's Claude Mythos.

Treat that as one outlet's framing until the keynote. Why it matters. The Android show told operators where Google's distribution layer is going. The keynote tells operators where Google's models, developer tools, and ad products are going.

Google Marketing Live on Wednesday closes the loop on the ads infrastructure side. Three back-to-back planning days, all visible to your clients, all triggering procurement and budget questions inside 72 hours. What it means for business, three planning questions for the week. First, if your software depends on Gemini API capability, the model news on Tuesday changes your H2 inputs.

Second, if your business sells anything through paid search or shopping, the Agentic Ads Framework News on Wednesday will rewrite your account management model. Third, if your hardware roadmap depends on Android, fleet devices, in-car experiences, retail terminals, the Google Book and Aluminium OS announcements affect a Q4 procurement decision you may not have known you were about to make. What it means for marketers and agencies. Block the diary today at 6 p.

m. UK time for the i slash O keynote on YouTube, 9 p.m. UK time for the developer keynote, and 5 p.

m. UK time on Wednesday for marketing live. Brief account directors with a one-page summary by Thursday. Update client H2 plans on Friday.

The single biggest risk this week is treating these events as content rather than work. My take. The undervalued story is Aluminium OS, Google's premium laptop tier, is the real fight for who owns the next 10 years of AI on desktop. Microsoft has Copilot Plus, Apple has WWDC three weeks away.

Google now has Google Book from October. Operators making decisions about employee hardware procurement for 2027 just gained a third credible choice. The interesting question is whether agencies start running Google Book fleets to give creative teams native access to Gemini intelligence by Q1 confidence level. High on the keynote logistics and Android show announcements.

Medium on the new Gemini model tier until it lands. Speculative on whether the Agentic Ads framework arrives as a unified product or as a set of features. Next up, OpenAI switches ChatGPT ads to cost per click bidding at $3 to $5 per click. Open self-serve at a $50,000 minimum spend.

What happened? On May 5th, OpenAI announced new ways to buy ChatGPT ads, replacing its launch CPM model with cost per click bidding. CPC bids run between $3 and $5 per click. The self-serve ads manager is now in beta.

Minimum advertiser spend has dropped from $250,000 to $50,000. OpenAI is forecasting $2.5 billion in ad revenue for 2026, scaling to $11 billion by 2027 and a stated $100 billion by 2030. The pricing shift followed the launch, $60 CPM eroding to roughly $25 within 10 weeks.

What is confirmed? OpenAI's May 5th product page, the CPC range, the self-serve ads manager beta, the minimum spend reduction, and the CPM erosion path are confirmed by OpenAI's publication and corroborated by TechCrunch Digiday Pounce and Quartz. The 2026 to 2030 revenue forecast is sourced from OpenAI internal projections reported by the Next Web. Treat as directional, not audited.

Why it matters. This is the first time a major AI assistant has put a real per-click number on its inventory. At $3 to $5 per click, Chat GPT is now directly priceable against Google Search, which trades closer to $2 to $8 for most commercial categories in the UK. Performance teams now have a number they can put in a forecasting model.

The $50,000 minimum spend opens the platform to mid-market advertisers for the first time. And the CPM erosion from $60 to $25 in 10 weeks tells you that pure attention pricing is not stable inside an assistant. You have to price against intent, what it means for business. For any brand running paid search at scale, ChatGPT is now a live channel to test, not a category to monitor.

The CPC model reframes the ROI question. Instead of buying impressions inside a chat, you are buying conversion-tied clicks against a user with high purchase intent. For DTC, SAAS and B2B services brands with strong creative and clean landing pages, this could be a structurally cheaper acquisition channel than Google search for specific intents. It is also where the OpenAI revenue projection becomes a procurement signal.

$2.5 billion in 2026 implies meaningful inventory and meaningful advertiser concentration risk if a small group of categories crowd in early. What it means for marketers and agencies, three actions this week. First, audit your top branded and category queries against ChatGPT's ad surface and forecast a $3-5 CPC against your existing Google search account benchmark.

Second, your account management retainers need to include a ChatGPT line item by Q3. Design the methodology now. Third, the $50,000 minimum spend tier is realistic for many mid-market clients. Package a 90-day test plan with creative measurement and reporting as a fixed fee discovery engagement.

My take. The revenue forecast is the headline number, but the operational story is that ChatGPT just became a measurable performance channel rather than an experimental brand surface. The trap is to overtest. The right approach is a controlled intent mapped test against existing Google Search and Microsoft advertising benchmarks.

The brands that test now will own attribution data and creative learnings before competitive crowding sets in. The counterpoint is that the $3 to $5 CPC range will not hold once volume scales. Assume the inventory tightens through H2 and price your tests accordingly. Confidence level high on the product change, medium on the CPC range holding once volume grows, speculative on the revenue forecast.

Next up, Snap and Perplexity amicably end the $400 million AI search deal that was supposed to anchor Snapchat's default chat experience. What happened? On May 6th, Snap disclosed that its $400 million AI search partnership with Perplexity, originally announced in November 2025, was amicably ended during Q1 2026. Under the original terms, Perplexity's conversational answer engine was to become the default AI experience inside Snapchat's chat interface, with Perplexity paying Snap approximately $400 million in cash and equity over 12 months.

The integration never moved beyond limited testing. SNAP has excluded the partnership from 2026 guidance and is leaning on AI-sponsored Snaps and its intelligent iWare roadmap to fill the monetization gap. What is confirmed? The termination, the original deal terms, the Q1 2026 window, the exclusion from SNAP's 2026 guidance, and the pivot to in-house monetization are confirmed by TechCrunch and Gadget, Quartz, and the AI Insider plus Snap's own investor disclosure.

Both companies attribute the unwinding to product fit issues, not financial reasons. The specific internal reasons inside each company are not disclosed. Why it matters? This is the first nine-figure AI distribution deal between two named well-funded companies to publicly fail.

The strategic question it answers is whether platform scale AI partnerships can hold when both sides have overlapping ambitions. Snap wants AI to deepen Snapchat engagement and add new ad inventory. Perplexity wants distribution to grow its subscription and answer engine business. When those goals diverge, even a $400 million commitment is not enough to keep the integration alive.

What it means for business, for any operator structuring a partnership where AI capability sits on one side and distribution sits on the other, this is the cautionary deal. Diligence the integration roadmap before signing. And assume the deeper integration paths will be slower and more politically expensive than the term sheet implies. For Snap shareholders, the immediate effect is lost guidance and renewed focus on AI-sponsored SNAPs and intelligent eyewear as the in-house monetization paths.

For Perplexity, the read is that Comet browser distribution and direct enterprise sales are now its real growth levers, not platform partnerships. What it means for marketers and agencies. Two reads. First, if you have a client with a Snapchat ad strategy that assumed a perplexity-powered AI surface inside chat, drop it from the H2 plan and reweight against Snap's own ad formats, AI-sponsored Snaps in particular.

Second, for any client mid-negotiation on AI integration deals with platform partners, your structural advice should be staged commercial commitments tied to integration milestones, not lump sum capital deployment. The term sheet matters more than the press release. My take, the Snap Perplexity breakup reflects a category-wide reality. AI platform partnerships are easier to announce than to ship.

Distribution for capability deals work when the partner cares more about distribution than about its own brand. Perplexity with its standalone consumer subscription growth and comet ambitions did not need snap enough to ship the integration. Expect more of these to unwind through Q3, particularly any deal struck in late 2025 in the pre-Mythos rush. The interesting question is which deal goes next.

Confidence level high on the termination, medium on the strategic interpretation until at least one other comparable AI distribution deal unwinds. Speculative on which deal follows. Lastly, Coti hands consumer beauty content production to Pencil with an embedded team inside Coti from 1st July. What happened?

On May 4th, Coti announced a partnership with Pencil owned by BrandTech Group, sister company to Jellyfish, to build an end-to-end generative AI content system across its consumer beauty division. Brands in scope include CovGirl, Rimmel, Sally Hansen, and Max Factor. A dedicated embedded pencil team begins working inside Coti from July 1, 2026, integrated into campaign workflows across brands and markets globally. Pencil covers ideation, copy, image, and video production with enterprise commitments on governance, security, and IP ownership on Coti's side.

Pencil has publicly stated it has produced 10 million ads since 2018 and predicted performance across over $4 billion in media spend. What is confirmed? The partnership, the brands and scope, the 1st of July Team Start, and the workflow scope are confirmed by Cauti's own news page, Yahoo Finance, Beauty Packaging, Fashion Network, and Global Cosmetics News. The deal value and contract length are not disclosed.

Why it matters? This is the most credible enterprise Gen AI content deployment in Beauty so far. COTI is not a startup doing a pilot. It is a tier one CPG marketer hardwiring Gen AI into its operating model.

The embedded team model matters because it is repeatable. Expect competitive Tier 1 advertisers in Beauty, FMCG, and adjacent categories to announce similar arrangements through H2. What it means for business, three implications. First, for CPG brands competing with Coti's portfolio, the content production cost curve just dropped meaningfully, and the brands that match it within nine months will defend share.

Those that do not will lose creative velocity in measurable ways. Second, for media agencies serving CPG clients, the embedded team model is a procurement competitor. Your line items for creative production are about to be benchmarked against pencils per asset cost. Third, for any brand evaluating Gen AI creative vendors, pencil's predictive performance layer is the differentiator to interrogate.

What it means for marketers and agencies, the action this week is to know the COTI pencil model in detail. Read the announcement, brief your senior creative leads, and pressure test whether your own creative production line could survive a COTI-style overhaul. For agencies, the most defensible position is to own strategy, brand, and audience layers and treat content production as an upstream managed output. For in-house teams, this is the case study to walk a CMO through when justifying a Gen AI infrastructure budget in H2.

My take. The signal here is governance, not production. Many enterprise Gen AI deals fail on IP, security, and brand safety questions, not on creative output. COTI has signed an arrangement with a partner that puts those guardrails inside the workflow rather than as a procurement afterthought.

That is the template, and the embedded team model is how it scales. The interesting question is which beauty competitor responds first. And whether the response goes to pencil, to jellyfish, or to an in-house brand textile build-out. Confidence level, high on the partnership, medium on competitive response timing, speculative on the cost savings claim until Coti publishes its own data.

Two items for the watch list today. Google Marketing Live Wednesday, May 20th. Three core themes pre-trailered the Gemini Advantage, AI-powered campaigns, Agenc Commerce, and a new era of performance on YouTube. Marketing advisor, the Chrome AI agent, Google Ads expert, and Google Analytics expert have all been previewed.

Expect the Agentic Ads framework to land as a unified product rather than a feature set. Calendar, the keynote at 5 pm UK time, and the breakout sessions through Wednesday and Thursday. Mustafa Suleiman's 18-month timeline. Fortune, eWeek, Yahoo Finance, and others recirculated Suleiman's claim this weekend that AI will reach human level performance on most professional tasks within 12 to 18 months, naming accounting, legal, marketing, and project management as exposed.

The original quote was given to the Financial Times earlier in the year. The signal is not the timeline. It is that the Microsoft AI CEO is now actively repeating the framing in public, which seeds enterprise procurement conversations and may indicate a co-pilot positioning shift before Microsoft Build later this week. What matters most?

The pattern across today's stories is that AI is becoming the operating layer of marketing, not a feature of it. Google is hardwiring Gemini into Android Chrome and the next generation of laptops. OpenAI has made Chat GPT a measurable performance channel. Snap is learning what happens when AI distribution deals do not ship.

COTI is hardwiring Gen AI into its content production stack. Operators who plan around AI as a workflow and procurement question will end the year ahead. Operators who plan around AI announcements will keep being late. What I would do if I were running this account, brand, or agency this week.

Block 90 minutes Tuesday evening and 90 minutes Wednesday afternoon for the Google I.O. keynote and Google Marketing Live keynote. Brief the team on Thursday.

Update the H2 plan on Friday. Run a ChatGPT CPC ads forecast against your top 20 Google search keywords. Use a $3 to $5 CPC range and a $50,000 budget assumption. If the forecast is competitive, build a 90-day test plan for Q3.

If you sell into CPG or beauty clients, brief them on the Coti Pencil deal this week. Position it either as a benchmark for their own production line or as a competitive threat to plan against. For any client mid-negotiation on a large AI partnership, refer back to Snap Perplexity as the integration risk case study. Insist on staged commercial commitments tied to integration milestones.

Audit your client's H2 plan for any reliance on perplexity SNAP integration or on any unannounced AI platform partnership. Reweight against in-platform ad formats where the integration has not shipped. What I'd tell a client today the most actionable AI news of the week is ChatGPT's switch to cost per click pricing. It turns AI search advertising into a measurable performance line for the first time.

Test it. Today's Google I.O. keynote and Wednesday's Marketing Live will.

Reset your H2 plan. Treat them as work, not content. The snap perplexity breakup is a warning, not a one-off. Diligence, every AI partnership your roadmap depends on against integration risk, not just headline value.

The story of the week is not any single launch. It is the operationalization of AI inside ad platforms, content production lines, and consumer hardware. Google I.O opens that week today.

The brands and agencies that treat the next 72 hours as planning time will end the quarter ahead of the ones who treat them as headlines. That's been today's episode. See you tomorrow.