The stories, shifts and signals from this week in AI Search, and what they could mean for your business.

    Back to Blog
    Connor Houghton··4 min read

    This Week in AI Search (w/c 06/07/26)

    Share this article

    TL;DR

    The week of 6 July 2026 came down to one theme: control over training data. A Google privacy settings change now lets it train models on photos, files and voice recordings submitted through Lens, Translate and voice search, by default, with Google Photos still excluded. OpenAI shipped GPT-5.6 Sol, Terra and Luna to general availability, with Terra matching GPT-5.5 at half the price and Luna at $1 per million input tokens, the same day Grok 4.5 launched. The New York Times and Daily News asked a federal judge to sanction OpenAI over allegedly concealed training-data searches, which OpenAI denies. OpenAI also revised its ChatGPT ad unit, now appearing in roughly 26.5% of replies globally and 49% in the US, and replaced GPT-5.3 Instant Mini with GPT-5.5 Instant Mini as the rate-limit fallback.

    Google widened AI training on search uploads this week. OpenAI shipped a new flagship model family, revised its ad format, and got accused of hiding evidence in a copyright trial. Four stories, one theme: control over training data.

    What Happened

    1. Google Widens AI Training on Search Uploads

    A privacy settings change lets Google train its models on photos, files and voice recordings submitted through Lens, Translate and voice search (TechCrunch, 2026a). The change arrived via a June customer email and applies by default. Google Photos content stays excluded (TechCrunch, 2026a).

    2. OpenAI Ships GPT-5.6 Sol, Terra and Luna

    GPT-5.6 reached general availability across ChatGPT, Codex and the API, ending a government-coordinated preview (CNBC, 2026). Terra matches GPT-5.5 performance at half the price. Luna undercuts it further, at $1 per million input tokens (OpenAI, 2026a). The launch landed the same day as Grok 4.5 (Neowin, 2026).

    3. New York Times and Daily News Seek Sanctions Against OpenAI

    The publishers asked a federal judge to sanction OpenAI, alleging it concealed training-data searches for copyrighted journalism (TechCrunch, 2026b). An April deposition allegedly showed OpenAI held 78 million de-identified ChatGPT conversations before the suit was filed. OpenAI denies the claims (TechCrunch, 2026b).

    4. OpenAI Revises ChatGPT's Ad Format

    OpenAI updated its static ad unit inside ChatGPT, the latest change since ads went live earlier in 2026 (Digiday, 2026). Ads now appear in roughly 26.5% of replies globally and 49% in the US (Tech Insider, 2026). Retail partners including Walmart, Shopify and Etsy support in-chat purchases (Digiday, 2026).

    5. OpenAI Rolls Out GPT-5.5 Instant Mini

    GPT-5.5 Instant Mini replaced GPT-5.3 Instant Mini as the fallback model shown once users hit rate limits (ReconnAI, 2026). It tracks intent more closely and produces fewer factual errors. The update does not touch the API or Codex (ReconnAI, 2026).

    Why It Matters

    Training data defines the trust gap in AI search this week. Google's opt-out default and OpenAI's discovery fight centre on one question: who controls the data feeding these models, and who agreed to it (TechCrunch, 2026a; TechCrunch, 2026b).

    OpenAI keeps compounding its lead. It shipped a cheaper flagship model, matured its ad platform, and outpaced rivals on output within one week (OpenAI, 2026a; Digiday, 2026). ChatGPT already draws 92.4% of standalone AI referral traffic (Adapt Worldwide, 2026).

    Publishers lose negotiating power each week the litigation drags on, while their own content keeps training the systems replacing their traffic (TechCrunch, 2026b). Businesses building on AI search visibility need to track legal exposure and falling model prices together. Both are moving fast. To see where your business stands inside AI search today, run your free AI Discoverability Score.

    Who Wins / Who Loses

    Winners

    OpenAI: Shipped GPT-5.6, cut flagship pricing, and matured its ad platform within days (OpenAI, 2026a; Digiday, 2026).

    Retail advertisers: Walmart, Shopify and Etsy expand chat-to-purchase flows (Digiday, 2026).

    Enterprise API buyers: Terra and Luna cut flagship-tier reasoning costs sharply (OpenAI, 2026a).

    Losers

    New York Times and Daily News: Stuck fighting over withheld evidence, with no resolution in sight (TechCrunch, 2026b).

    Google Search users: Media submitted through Lens, Translate and voice search now trains Gemini by default (TechCrunch, 2026a).

    News publishers generally: Litigation drags on while OpenAI keeps shipping and gaining share (TechCrunch, 2026b; Adapt Worldwide, 2026).

    The One Thing to Remember

    Data is this week's real battleground. Google's training opt-out, OpenAI's discovery fight, and its ad expansion trace back to one question: who owns the inputs feeding AI search. Whoever settles that first sets the terms for everyone else.

    References

    • Adapt Worldwide (2026) AI Search Roundup: The Biggest Google, ChatGPT and SEMrush Updates from June-July 2026. Available at: adaptworldwide.com (Accessed: 13 July 2026).
    • CNBC (2026) OpenAI to publicly release GPT-5.6, rolls out conversational AI models. Available at: cnbc.com (Accessed: 13 July 2026).
    • Digiday (2026) OpenAI gives ChatGPT ads a visual upgrade. Available at: digiday.com (Accessed: 13 July 2026).
    • Neowin (2026) OpenAI to release GPT-5.6 Sol, Terra and Luna on July 9. Available at: neowin.net (Accessed: 13 July 2026).
    • OpenAI (2026a) Previewing GPT-5.6 Sol: a next-generation model. Available at: openai.com (Accessed: 13 July 2026).
    • ReconnAI (2026) July 6, 2026: GPT-5.5 Instant Mini in ChatGPT. Available at: reconn-ai.com (Accessed: 13 July 2026).
    • Tech Insider (2026) ChatGPT Ads Hit 49% of US Replies, Live in 5 Nations. Available at: tech-insider.org (Accessed: 13 July 2026).
    • TechCrunch (2026a) If you use Google, you're training its AI. Here's how to opt out. Available at: techcrunch.com (Accessed: 13 July 2026).
    • TechCrunch (2026b) New York Times says OpenAI hid evidence in ChatGPT copyright trial. Available at: techcrunch.com (Accessed: 13 July 2026).

    Keep Reading

    Related Articles

    Continue reading on similar topics

    Start With Clarity, Not a Contract

    See where you show up when buyers ask AI who to hire, and who shows up instead of you.

    Book a Free Growth Assessment

    We show you the gap, you decide what to do with it.