AI Marketing News, 10 Aug 2026: Unlimited Free ChatGPT, Adobe's ChatGPT Plugin, AI Shopping at 72%
OpenAI removes free ChatGPT text limits, Adobe merges 70+ tools into one ChatGPT plugin, and 72% of high-income US shoppers now consult AI for purchases.
OpenAI removes text chat limits for free and Go ChatGPT users. Announced August 6-7, GPT-5.6 Luna replaces GPT-5.5 Instant as the default for Free and Go accounts and now carries unlimited text messages, up from a prior cap of roughly 10 to 40 messages per 3 to 5 hours. Free users also get a Think button that grants Luna more reasoning time per message. Limits remain in place for file uploads, images, and tool use. OpenAI reported internal evaluations showing factual errors down 62% for Luna and 68% for the retuned Sol model against GPT-5.5 Instant.
Importance for marketers: Removing the message cap changes how deep a session goes. A buyer who used to run out of turns mid-comparison now finishes the evaluation inside ChatGPT, so your category’s consideration stage resolves before anyone reaches a landing page. If your AI-visibility work optimizes for the single-answer query, you are aiming at the wrong shape: these are multi-turn threads where a competitor gets named in a follow-up you never see. Check which of your comparison and pricing pages a model can retrieve, because those are what get pulled into turn four.
Adobe collapses 70-plus creative and PDF tools into a single ChatGPT plugin. On August 6 Adobe launched Adobe for ChatGPT, replacing the separate app connectors it shipped in December 2025 with one unified integration spanning Express, Photoshop, Firefly, Premiere, Acrobat, Lightroom, Illustrator, InDesign, and Stock. The plugin is free, available globally through ChatGPT web and app, and requires no Creative Cloud subscription to start. Users invoke it with “@Adobe” and the plugin routes the request to the appropriate underlying tool automatically.
Importance for marketers: Adobe is betting the interface for creative production is a chat thread, not an application, which puts asset generation on the same surface where your ops team already runs prompts. For a small team that removes the handoff between brief and asset, and it gives you a credible path to generating campaign creative inside an n8n or agent workflow without stitching three APIs together. Watch provenance. Assets made this way still need C2PA marking to satisfy the EU AI Act Article 50 duties that became enforceable on August 2, and a plugin that routes silently between eight tools makes it harder to say which ones touched a given file.
PubMatic makes a written governance policy a prerequisite for agentic ad buying. On August 5 PubMatic launched a five-step guardrail architecture for AgenticOS, requiring an authorized account administrator to define governance parameters in natural language before any autonomous campaign runs. The system is dual-tier: PubMatic sets platform-wide constraints, and buyers configure budget thresholds, creative approval requirements, inventory allowlists, and audience restrictions. A drift detection layer flags targeting decisions or recommendations that fall outside configured norms for human review before execution. PubMatic said AgenticOS has run more than 80 autonomous campaigns across 100,000-plus properties, with Rise, a Quad agency, among the first partners piloting the framework.
Importance for marketers: Onboarding now produces an auditable policy document, and someone on your team writes it, defends it in review, and revises it when drift alerts land. That work sits closer to an AI governance board than to media buying, and most in-house teams have nobody holding it today. Name the owner before you sign. Skip that and PubMatic’s platform defaults become your brand safety policy by omission. This is the management layer described in paid media agents arriving as product.
KIBO ships a model-agnostic agentic commerce layer. On August 6 KIBO Commerce announced general availability of KIBO AI, consolidating the nine separate agents from its earlier agentic release into one layer spanning five functions: Engage, Configure, Explain, Analyse, and Optimise. It runs on a unified data model covering both commerce and order management, and includes a Bring Your Own Model approach KIBO says eliminates LLM lock-in. Optimise is the most autonomous of the five, continuously tuning system parameters such as inventory allocation against defined goals.
Importance for marketers: Bring Your Own Model separates the platform contract from the model contract, which gives you a lever worth using now that inference pricing moves every quarter. If you are renewing a martech or commerce platform, ask whether the model layer is substitutable and price the two lines apart. You inherit the selection problem in exchange: someone on your side decides which model runs which function, then decides again each time pricing or capability shifts.
Optimove: 72% of high-income US consumers now use AI tools for shopping ideas. The Optimove Insights Holiday Shopping Report 2026, released August 5, surveyed 648 US adults with household incomes above $75,000 and found 72% regularly or occasionally consult ChatGPT, Claude, or Gemini for product and gift recommendations. On specific features, 55% have used AI product recommendations, 41% chatbots, 31% visual search. Nearly three-quarters trust AI-generated recommendations as much as or more than other sources. Optimove’s April 2026 Mother’s Day report put the equivalent figure at 49%, giving a measured prior rather than an estimate.
Importance for marketers: A move from 49% to 72% in four months on comparable panels shows the comparison step leaving brand properties while you watch. Your product data does the persuading now, because the model reads structured attributes and third-party reviews and never sees your hero copy. Put this quarter’s hours into feed quality and review-site presence ahead of landing page conversion work, since the model samples the first and the second only matters once the shortlist is set. The B2B version of this problem is brand visibility in AI answers.
MarTech Breakthrough Awards now carry a standing Answer Engine Optimization category. Announced August 6, the 9th annual program drew nominations from more than 15 countries and included a permanent AEO category, naming Brandi AI as AI Search Visibility Platform of the Year and PR Newswire Amplify as AEO Platform of the Year. A parallel Agentic AI and Autonomous Marketing category named Zeta Global and Perion. Other winners included Klaviyo for CRM Company of the Year and impact.com for Overall MarTech Company of the Year.
Importance for marketers: Awards programs lag the market, which is what makes this one readable: analysts create a category once enough comparable vendors exist to sort. AI search visibility has crossed from experiment to named budget line, so your 2027 planning conversation covers vendor selection rather than whether to fund it. Make anyone pitching a visibility score disclose their methodology first. The category is young enough that “share of AI voice” means something different at every vendor selling it, which is why those scores stay diagnostic until they meet the bar in diagnostic vs decision-grade AI visibility metrics.
Cloudinary lets AI agents provision environments with no account and no payment method. On August 5 Cloudinary launched an Agent Experience layer allowing agents to self-provision a full visual media environment, which it calls a Claimable Cloud, with no human configuration required. Developers can take ownership of the environment later and carry it into production without losing work. The release also covers MCP server support, agent-native tooling, Skills packages, and AI-optimized SDKs, available to all customers.
Importance for marketers: Cloudinary has taken procurement out of the front of the process and left the invoice at the back, which lands on marketing ops before it reaches finance. An agent that spins up media infrastructure without a card on file still runs up usage that someone reconciles later, and nobody has owned that reconciliation until now. Set a usage ceiling and name an owner when you design the workflow, not when the Q4 true-up arrives.
Cision report names the cost of reading search, social, and AI answers in separate systems. Released August 5, The Data Fragmentation Trap argues that organizations collecting search, social, news, and AI-generated answer signals through channel-by-channel tools pay a “fragmentation tax” in blind spots and delayed reputational response. Cision cited its own Brandwatch Marketer of 2026 research finding 40% of surveyed marketers named integrating data from multiple sources as their biggest challenge. Chief Product Officer Jim Daxner framed it as a connection problem rather than a data problem.
Importance for marketers: Read this against its authorship, since Cision sells the connected alternative across Brandwatch, CisionOne, Trajaan, and PR Newswire. The structural point survives the sales pitch. AI answers are a signal type most measurement stacks have no slot for, so someone bolts on a separate dashboard and it never joins the others. What you need is a normalization layer, and a Clay table or n8n flow pulling each source into one schema does that for a fraction of a suite license.