AI Visuals Just Shipped

This was a massive week for AI-generated visuals. OpenAI shipped ChatGPT Images 2.0 with reasoning baked into the model, and Anthropic dropped Claude Design, a full-blown design tool that has people drawing comparisons to Figma and Canva. Neither one renders exactly what you picture every time, but the gap between "AI slop" and "usable marketing asset" just got a lot smaller.

Two years ago, an AI couldn't spell "margarita." This week's models are laying out manga pages, rendering QR codes that actually scan, and building interactive prototypes from a screenshot. I also ran a live test of Opus 4.7's new vision capabilities at this month's alumni meetup, pitting four frontier models against the same landing page design challenge. Walk-through in the learning resources below.

Let's get into it.

🔥 This Week's Hot Takes: AI Marketing Edition

Anthropic Launches Claude Design to Compete With Figma and Canva (Anthropic) – Anthropic Labs rolled out Claude Design, a new product that turns prompts into prototypes, slides, one-pagers, and marketing assets. It reads your brand from your codebase or design files, applies your system automatically, and exports to Canva, PDF, PPTX, or standalone HTML. Available now for Pro, Max, Team, and Enterprise subscribers at no extra charge.

A Non-Designer Spent a Day With Claude Design and Said This (Pen With Paper) – The reviewer, who isn't a designer, called Claude Design a fourth option beyond hiring someone, learning Figma, or using cookie-cutter templates. Figma's stock reportedly dropped seven percent the day it launched. Worth noting: Canva's design engine is part of what's underneath, so exports flow straight into Canva for further editing.

OpenAI Ships ChatGPT Images 2.0 With Reasoning Built In (TechCrunch) – The new model (gpt-image-2 in the API) finally spells. It can search the web, reason through layout before rendering, generate up to eight consistent images from a single prompt, and double-check its own work. Free users get Instant mode. Plus, Pro, and Business tiers get Thinking mode.

ChatGPT Images 2.0 Nails Multilingual Text, Maps, and Manga (VentureBeat) – Early hands-on tests show clean typography in Japanese, Korean, Hindi, and Bengali, plus working QR codes embedded inside fully branded posters. Aspect ratios run from 3:1 banners to 1:3 vertical stories at up to 2K resolution. The release collapses a generator, a layout app, and a typography editor into one prompt.

OpenAI Positions Images 2.0 as a Visual Production System, Not a Toy (OpenAI) – OpenAI is pitching the new model as infrastructure for real workflows: localized advertising, infographics, UI mockups, educational diagrams. The company frames images as "a language, not decoration," signaling a shift from one-off art generation toward repeatable asset production.

Google Ships Deep Research Max and Takes the Top Spot on Key Benchmarks (OfficeChai) – Google's new autonomous research agent, built on Gemini 3.1 Pro, scored 93.3% on DeepSearchQA and 85.9% on BrowseComp, outperforming GPT 5.4 Thinking and Claude Opus 4.6 on both. It now connects to third-party data via MCP (FactSet, S&P, and PitchBook are first partners) and generates native charts and infographics inline.

Before You Buy Another AI Tool, Answer These 5 Questions (MarTech) – Tonya Walker makes the case that most AI tool failures are operational, not technical. Her five questions cover data readiness, stack integration, decision ownership, failure modes at scale, and true operating cost. Her warning: buying AI without the operational infrastructure creates "AI debt" the team pays back later in broken workflows.

Your AI Assistant Is Quietly Becoming a Salesman (The Economist) – Ads are showing up inside chatbot conversations, and the two biggest platforms are taking different approaches.

The Execution Gap Keeps Holding AI Marketing Teams Back (Forbes) – Rhett Power argues most marketing teams are buying AI faster than they can operationalize it. Adoption is near-universal, but the data, systems, and governance needed to actually move outcomes haven't caught up.

Anthropic and Amazon Sign a $100B+ Compute Deal (Anthropic) – Anthropic committed over $100 billion to AWS over ten years, securing up to 5 gigawatts of new capacity for Claude. The consumer and enterprise demand spike is real, and the hardware scramble is on.

🧪 The Alchemist's Lab: Tools I've Been Testing

Claude Design Function: Prompt-to-visual design tool that turns natural language into prototypes, slides, landing pages, and marketing assets. Refine through chat, inline comments, direct text edits, or custom sliders Claude generates on the fly. Exports to Canva, PDF, PPTX, or handoff to Claude Code. Marketing application: Rapid first drafts of campaign pages, social assets, one-pagers, pitch decks, and multi-asset campaigns with your brand system applied automatically. Pricing: Included with Claude Pro, Max, Team, and Enterprise subscriptions (uses your existing subscription limits).

ChatGPT Images 2.0 Function: OpenAI's new image model (gpt-image-2) with reasoning, web search, up to eight consistent images per prompt, aspect ratios from 3:1 to 1:3, and 2K resolution. First OpenAI image model that renders legible text across English and non-Latin scripts. Marketing application: Infographics with real data labels, multilingual campaign assets, storyboards with character consistency, product shots with accurate packaging, and QR-coded posters that actually scan. Pricing: Instant mode is free to all ChatGPT users. Thinking mode is limited to Plus ($20/mo), Pro ($200/mo), Business, and Enterprise.

🎓 AI Marketing Master Class Moment

The Opus 4.7 Landing Page Design Challenge (Advanced AI for Content Marketing Alumni Meetup, April 2026)

In this month's meetup, I ran a live vision test across four frontier models (Opus 4.7, Opus 4.6, ChatGPT, and Gemini), asking each one to audit HubSpot's homepage and then design a competitive Salesforce landing page from scratch. Gemini produced the strongest strategic mockup, with copy that went head-to-head against HubSpot's positioning. Claude won at turning that mockup into working HTML. The vision test and landing page challenge start around the 50-minute mark, with a live Canva Magic Layers demo shortly after. If you want to see Opus 4.7's new vision capabilities tested against real marketing work and a walk-through of when to reach for which model for which part of the job, this is the segment.

Join My LIVE Lightening Lesson: How to Audit Your Content Strategy with AI 

Most content marketers are publishing without ever testing whether their work resonates with their audience. And most have never audited what they've already built to find out what's actually performing. These are two of the highest-value activities in content marketing, and AI makes both of them possible right now. You'll walk away knowing your own content better than you do today.

My free Lightning Lesson on May 8 walks through building an AI companion for content feedback, running a data-driven audit with agentic browsers, and turning both into a practical action plan. If you've been cranking out content without ever asking whether it's actually working, this is the fastest way to find out.

🤖 AI Whispers (That You'll Want to Shout About)

AI-generated visuals crossed a real threshold this week. What used to be novelty output can now plausibly ship as production assets. Two years ago, the joke was "churiros" and "enchuita" on fake restaurant menus. This week, ChatGPT Images 2.0 is generating working QR codes embedded in branded posters, and Claude Design is exporting full marketing landing pages to Canva.

For marketers, the visual content bottleneck is cracking open. The work that used to need a designer, a copywriter, a layout tool, and a QR generator now fits inside one prompt. Smaller teams can produce brand-consistent multilingual creative without an outside agency. Agency teams can test three times as many concepts before picking one.

The catch is familiar. Everyone gets access to these tools now. Not everyone knows what to do with them. The ones who get real value already know what good looks like, can spot when the model is off-brand or off-strategy, and have a workflow for getting from "first draft" to "shippable." Everyone else will flood feeds with 2K-resolution slop.

🏆 My Take: What This Means for Your Marketing

My biggest takeaway this week is that AI visual generation has officially moved from ideation to production, and that shifts where your time should go.

For the last two years, "getting a decent image" was a meaningful bottleneck. You'd iterate on prompts, accept something 70% right, and move on. That phase is ending. With Claude Design and ChatGPT Images 2.0 in your toolkit, the limiting factor is no longer whether the model can render it. It's whether you know what you want the image to do.

Your action item for the week: pick one piece of visual content you're about to create. A social graphic, a landing page hero, a pitch deck slide, a multilingual campaign asset. Before you prompt anything, write down three things in plain English: who this is for, what you want them to feel, and what the single most important word or number on the image should be. Then run it through both Claude Design and ChatGPT Images 2.0. Compare the outputs against your intent, not against each other.

You'll learn more in that one exercise than in ten hours of reading launch coverage.

📊Before You Go: Hit reply and tell me - what's your biggest AI marketing challenge right now? I read every response and it helps shape future newsletters!

That's a wrap on this week's AI marketing alchemy! Keep experimenting, keep transforming, and remember - true magic happens when we blend AI efficiency with human creativity.

Crafted with passion (and a dash of AI alchemy) by Lisa Peyton | AI Practitioner, Professor, and Pioneer

Let's connect in the digital realm: https://linktr.ee/lisapeyton

🎓 AI for Strategic Content Marketing – 50% Off for Newsletter Subscribers

My on-demand course, AI for Strategic Content Marketing, is 50% off right now for newsletter subscribers.

It's 55 lessons and 3 hands-on projects covering everything from AI-powered market research and competitor analysis to building audience profiles, developing a full content strategy, and measuring results. You also get 6 advanced bonus sessions on agents, automation, and lead gen, plus a live session with me on the current state of GenAI for content marketing.

You get lifetime access to all the content, a community of peers doing this same work, and a certificate of completion you can add to LinkedIn.

The course is $297. With your code, it's $148.50.

Use code maven50 at checkout.

→ Enroll here: https://maven.com/lisa-peyton/ai-for-strategic-content-marketing?promoCode=maven50

If you've been following along in the newsletter or showing up to the meetups and wondering where to start building your own AI content practice, this is it.