Google remakes Images, Apple opens Siri: July 15 brief

Google is turning image search into a personalized discovery feed, which gives strong product photography more room to find buyers before they type a precise query. Apple also put its new Siri in front of more testers, while ChatGPT made years of saved work easier to retrieve. Discovery is getting more visual and more personal.

Google turns Images into a discovery feed

Google introduced a new browsable home for Google Images on July 14. The company says the desktop experience will show a live gallery drawn from the web and tailored to each signed-in user's interests. Saved collections will appear as tabs above that gallery. The official announcement says the change will reach English-language users in the US over the coming weeks.

This matters because a customer may now encounter your product, venue, project, or recipe while browsing a topic, before making a search that names exactly what they want. Google is borrowing the feel of an inspiration feed. For a furniture shop, landscaper, restaurant, or travel business, the image can become the first sales surface.

Google is also adding image creation to AI Overviews, the generated answers that sometimes appear above normal search results. That makes original, clearly labeled photography more valuable as proof. Generic generated pictures will be easier for everyone to produce. Real work will be easier to trust.

This extends the traffic shift we covered in our guide to AI Overviews. Owners should also treat strong images as part of building a brand customers search for by name, not as decoration added at the end.

Apple puts Siri AI in more customers' hands

Apple opened the iOS 27 public beta on July 14, moving its new Siri experience beyond developer-only testing. A beta is test software released before the finished version, so bugs are expected. Apple's Siri AI overview says the assistant can use what is on the screen, search across personal messages and email, take actions in apps, and look to the web for current answers.

The business consequence sits in that last ability. Another large group of customers will soon ask an assistant where to eat, what to buy, or who can solve a local problem. Those answers may happen without a conventional list of search links.

The public test does not justify rebuilding your marketing around Siri today. It does justify a visibility check. Ask the same five buying questions a customer would ask, record which businesses appear, and inspect the facts Siri repeats. Wrong hours, vague service pages, and thin review profiles are the obvious repairs. Our explainer on how AI assistants choose businesses covers the source signals to clean up first.

ChatGPT makes stored work searchable

OpenAI rolled out one search box across ChatGPT chats, projects, images, and documents on July 14. According to the ChatGPT release notes, the feature is available worldwide on every plan across web, iOS, and Android. Filters let users narrow results by the kind of material they want.

For a small team, the immediate gain is less time hunting for an old proposal, research thread, or generated graphic. The less comfortable part is that stored work becomes much easier to surface. A useful archive can also become a messy one, full of old prices, customer details, and drafts that nobody remembers uploading.

Set a simple rule before the archive grows: decide what staff may put into ChatGPT, who owns cleanup, and when old files should be deleted. Keep passwords, payment records, and sensitive customer information out. If your team uses AI for copy, our writing-tools comparison explains where human review still belongs. Searchability saves minutes only when the material being found is safe and current.

Worth watching

• Google Images' new home starts on US desktop in English. Watch whether referral traffic begins arriving from broader browsing rather than a typed image query.
• Apple says Siri AI availability varies by device, language, and region. Test on the phones your customers actually use before drawing conclusions.
• A security researcher published a Claude memory attack that he says Anthropic has now mitigated. The lasting lesson is to keep confidential material out of assistant memory unless the business truly needs it there.