This Week in AI Search (w/c 03/08/26)

OpenAI made its default model free and unlimited this week, then confirmed the end of its Atlas browser (OpenAI, 2026a). Perplexity beat Amazon in federal court over AI shopping agents (Engadget, 2026). Brussels started enforcing new AI transparency rules (European Commission, 2026).

What Happened

1. OpenAI Makes GPT-5.6 Luna Free and Unlimited

OpenAI made GPT-5.6 Luna the default for Free and Go users on 6 August, with no text chat limits (OpenAI, 2026a). Paid users got an upgraded GPT-5.6 Sol with a reasoning slider and better factual accuracy, and a Think button reaches free users the following week (MacRumors, 2026). Rate limits remain on file uploads and image generation.

2. OpenAI Confirms Atlas Browser Shutdown

ChatGPT Atlas stops working on 9 August, less than a year after launch (OpenAI, 2026b). Browser features move into a Chrome extension, the desktop app, and ChatGPT itself. Bookmarks do not transfer automatically, so users must export them as an HTML file before the deadline (9to5Mac, 2026).

3. Perplexity Defeats Amazon Over AI Shopping Agents

The Ninth Circuit overturned the injunction blocking Perplexity's AI agents from shopping on Amazon on 4 August (US News, 2026). The court found Amazon unlikely to succeed under the Computer Fraud and Abuse Act, because it is the user directing the agent who accesses Amazon's computers, not Perplexity (Engadget, 2026). It is the first federal appeals ruling on whether AI agents acting for users can legally access online platforms. The underlying case continues in San Francisco.

4. EU Starts Enforcing AI Transparency Rules

Brussels began enforcing the AI Act's transparency requirements on 2 August (European Commission, 2026). Chatbots must identify themselves as automated systems, AI-generated or edited content must carry machine-readable marks, and deepfakes need labels. Non-compliance risks fines up to €15 million or 3% of worldwide turnover. Generative systems already on the market have until 2 December to meet the marking obligation.

5. Study Finds Topical Depth Beats Breadth for Brand Mentions

New research across roughly 1,000 categories found brands get cited outside their expertise, but not named (Search Engine Land, 2026). In distant topic categories only 25% of appearances are named brand mentions, against 50% that are citations. Shallow presence across many categories correlates with fewer mentions, and legal and healthcare brands need repeat mentions before AI recommends them by name.

Why It Matters

OpenAI is folding every surface into one product. Free unlimited chat removes the reason to leave ChatGPT for a rival (OpenAI, 2026a). Killing Atlas concedes the standalone browser bet failed, but the agentic ambition moves inside ChatGPT rather than away (OpenAI, 2026b). Fewer entry points mean OpenAI controls more of the daily search habit.

Courts and regulators are drawing lines the platforms could not agree themselves. The Ninth Circuit treats an AI agent as an extension of the user, not of the company that built it (Engadget, 2026). That logic protects Perplexity now and will shape every agentic commerce dispute that follows. Brussels moved the other way, forcing AI systems to announce themselves rather than blend into human interaction (European Commission, 2026).

The citation research shows where the practical work sits. Getting cited outside your niche is achievable; getting named is not (Search Engine Land, 2026). A thin spread across many topics buys citations without recommendations, which is exactly the outcome that looks like progress in a dashboard and produces no enquiries. Depth inside the categories you actually serve is what earns the named mention. To see which of the two you currently get, run your free AI Discoverability Score.

Who Wins / Who Loses

Winners

Losers

The One Thing to Remember

Control is consolidating into fewer interfaces, and courts now decide who answers for what an agent does. OpenAI wants one app for everything. Perplexity just won the right to act on a user's behalf inside someone else's platform. Meanwhile the research says being cited everywhere is not the same as being recommended anywhere. Depth in the categories you actually serve is the work that pays.

References

• 9to5Mac (2026) OpenAI explains what will happen when ChatGPT Atlas shuts down this weekend. Available at: 9to5mac.com (Accessed: 10 August 2026).
• Engadget (2026) Perplexity has successfully overturned Amazon's injunction on its AI shopping bot. Available at: engadget.com (Accessed: 10 August 2026).
• European Commission (2026) Commission starts enforcing AI Act rules and new transparency requirements on 2 August. Available at: digital-strategy.ec.europa.eu (Accessed: 10 August 2026).
• MacRumors (2026) Free ChatGPT Users Get Unlimited Text Chats and GPT-5.6 Luna. Available at: macrumors.com (Accessed: 10 August 2026).
• OpenAI (2026a) Improving GPT-5.6 Sol in ChatGPT and expanding access to GPT-5.6 Luna for free users. Available at: openai.com (Accessed: 10 August 2026).
• OpenAI (2026b) Evolving Atlas into ChatGPT for browser-based agentic work. Available at: help.openai.com (Accessed: 10 August 2026).
• Search Engine Land (2026) Does topical focus make your brand more visible? Available at: searchengineland.com (Accessed: 10 August 2026).
• US News (2026) Amazon Loses US Court Ban on Perplexity's AI Shopping Tools. Available at: usnews.com (Accessed: 10 August 2026).