Live AI search news feed
Live from Momentic Studio, this news feed is curated by the Momentic team so you can stay up to date on AI search, SEO, GEO, AEO, AI search, frontier AI models, and relevant tools.
2 new image-related Google Search announcements:
New home for Google Images: images.google.com with new features for signed-in users:
- Tailored to your interests
- Save ideas to collections, which become tabs above the gallery
Nano Banana is coming to AI Overviews (already available in AI Mode). Rolling out over the coming weeks.
AI performance insights are rolling out in Google Merchant Center Analytics. Here's the help document with details on the metrics, scope filters, and "optimization opportunity scorecards".
I checked a bunch of the GMC dashboards we have access to and found it in one of them, so keep your eyes peeled!
When available it's found under Analytics → Products

Apparently there's a hack you can use to force your platform performance report (for YT, insta, FB, or X) in GSC to show generative AI performance: just append /ai to the report URL.
But it doesn't work on regular GSC performance reports (I tried it). If you've got any of your social platforms hooked up in GSC already, give it a shot!
OpenAI is sunsetting their ChatGPT Atlas browser because the new ChatGPT desktop app has a built-in browser. Atlas will be deprecated August 9, 2026 (it arrived on the scene in October 2025). They also updated their Chrome extension so you can use ChatGPT directly in Chrome's sidebar.
But wait there's more: ChatGPT Work is a new agent inside ChatGPT that can do stuff for you, not just answer questions. It's powered by GPT-5.6. Lots of use case examples in the announcement.
> ChatGPT Work starts rolling out today on web and mobile, beginning with Pro, Enterprise, and Edu users and expanding to Plus and Business users over the next few days. The updated ChatGPT desktop app is available globally today for Mac and Windows, with Chat, Work, and Codex available to users on every plan, including Free.
Anthropic created a "reflection dashboard" (now in beta) to let you track, visualize, reflect, and refine how you use Claude:
- Review past Claude chat activity over 1, 3, 6, or 12 months - how often you used it, and what you spent time on
- See a report summarizing your Claude activity across the "4D AI Fluency Framework": Delegation, Description, Discernment, Diligence
- Schedule "quiet hours" and breaks when Claude is off
- Get a nudge when your session runs long
- Periodically see questions surfaced, like "What's one thing you want to keep doing yourself, even if Claude could do it faster?"
- Receive practical suggestions like "Start a Project instead of re-explaining this context" for ongoing work
Incognito chats are excluded from the reflection dashboard, as are files from connected tools.
To use it, memory has to be turned on. You can try it out by going to Claude Settings. Coming next for Cowork conversations.
GPT-5.6 is here, and there are 3 models in this family:
- Sol (flagship)
- Terra (balanced for everyday work)
- Luna (fastest & most cost-efficient)
The announcement includes a video, and a ton more detail about each model.
Ask YouTube is now available on desktop - you have to be signed in and aged 13 or older. Standard YT search results are still available, you'll just see a new :sparkles:Ask YouTube button in the search bar
Google Merchant Center Next replaced Google Merchant Center in March 2023.
Now they are switching it back to Google Merchant Center & scrubbing "next" from GMC branding & documentation. Just a name change, no functionality changes.
GPT-Live is rolling out now:
> Starting today, when you tap the Voice button to talk with ChatGPT, you’ll get an improved experience powered by GPT‑Live—with more natural conversations, smarter answers, better listening, and visual responses.
Cloudflare and OpenAI paired up for a research pilot. The goal is to find out how websites in Cloudflare's network (about 20% of the total) can help AI search engines discover & index web content better.
Google added *category* to merchant listing structured data:
> *What*: Updated the Merchant listing documentation to detail how the `Product.category` property can be used with both `Text` and `CategoryCode` types. This aligns with the Google Merchant Center feed specifications for the product_type and google_product_category attributes.
>
> *Why*: To help merchants provide both merchant-defined and Google-defined category information within their schema.org markup, enhancing product information for Google Search and Shopping.
Google also clarified how to specify sale price effective dates in Product structured data:
> *What*: Added a new section on "Sale duration" to the Merchant listing guide. This explains how to use the `validFrom`, `validThrough`, and `priceValidUntil` schema.org properties to set the effective range for sale prices, including best practices and examples for placement on either `Offer` or `PriceSpecification` nodes.
>
> *Why*: This aligns schema.org usage with the Merchant Center feed attribute sale_price_effective_date, providing clear instructions and best practices for merchants using structured data.
Unconfirmed experiment people are noticing on Google SERPs: search result links use google.com/goto as a passthrough URL before redirecting you to the actual web page you wanted.
Alex Greenland posits it's to make it harder for AI scrapers to grab links from SERPs and that sounds reasonable to me. Haven't seen this in the wild myself but good to be aware of in case you come across it or a client brings it up!
Potentially big issue for those who use the Google & YouTube Shopify app to sync products from Shopify to GMC. Because the Google Merchant API is replacing the Content API on 8/18, when you migrate the app it will change all your product IDs. That means all historical data will be gone.
There's a workaround (if you do it before August 18), and also the Google Ads Liaison has said "we're looking into this" so it's possible Google will come up with a solution.
Wil Reynoldes/Seer shared an open-source GEO strategy tool. It's a 2x2 matrix for collaborative "thinking through" of all the various AI visibility levers to help everyone understand the speed, risk, impact, and "doableness" of each tactic
Google Search Console update: new documentation on platform properties in GSC
You can add a supported property on the Search Console welcome screen, or with the property selector dropdown in the sidebar → Add property.
Supported platforms: YouTube, Instagram, TikTok, X
It's rolling out gradually, but once you have access you can view reports to see how your other platform content performs in Google Search/News/Discover (if it appears in those places).
> Platform properties only show how your content performs on Google surfaces. They don’t track when people see your content on the platform itself (for example, they won’t show how many times your video appeared on TikTok).
You'll be able to find total clicks, impressions, CTR & average search position (same metrics as the standard Performance report).
> To help you understand your metrics, here are some examples of how Search Console handles specific types of platform property content and search interactions:
> • Instagram stories:When your Instagram stories show up in Google search they’re counted as impressions. If a story appears in search results and a user clicks it, it’s counted as a click.
> • Videos played on Google:When your video content appears in search results or Discover, it counts as an impression. If a user clicks it, it counts as a click. Note that even if that click opens the video inside the Google viewer, a click will be added in Search Console.
Bill Hartzer made a free CLS tool to make it faster & easier to diagnose what's causing the problem after PageSpeed insights shows a poor CLS score
Need some data to show that improving CWV improves business outcomes? John Mueller shared this case study on LinkedIn.
Cloudflare added more nuance to their AI bot controls to give users more ways to manage AI traffic:
- Search - Bots that scan your site to help it appear in search engine results
- Training - Crawlers that scrape your content to train AI models
- Agent - Bots that retrieve info from your site to answer user questions
The options across all 3 use cases are: Allow, Block, or Block on pages with ads. The reasoning is that ads are meant to be seen by people, and this monetization funds the business that publishes the content.
Starting September 15, 2026 there will be new default settings for these controls:
- Training & Agent will be blocked by default on pages that display ads
- Search will be allowed by default
- Multipurpose crawlers that combine Search + Training (e.g. Googlebot, Applebot, Bingbot) will be blocked if the user blocks Training
They're also creating bot management capability based on content use, with three settings:
- Immediate - go ahead and interact, but you can't store or reuse anything
- Reference (default) - index, excerpt, and reference back
- Full - go ahead and summarize/reproduce my content
These content use values can be combined with the bot classifications to refine the rules to your liking.
Cloudflare is currently testing a new `use` signal that will live in your robots.txt but it will only signal a preference, it won't block things.
There's more in this article too, about verified bots and "transitive trust" and things for the future they are thinking about.
Google updated AMP (accelerated mobile pages) documentation yesterday:
> *What*: Simplified our AMP documentation by removing outdated references to the AMP viewer, AMP Cache, and signed exchange.
> *Why*: Starting today, Google Search is updating how it connects users to AMP pages, and will now take users directly to the publisher's AMP host pages. This change simplifies and reduces maintenance efforts for publishers who are creating AMP content, as they no longer need to update the AMP cache or configure signed exchanges. AMP content will continue to rank just like any other web page.
If you noticed an uptick in total citations/avg. cited pages in Bing WMT's AI Performance report around the beginning of June (I sure did, across multiple accounts) I'm sorry to say it's not because your website started getting more AI visibility. :pensive:
Microsoft Bing's Krishna Madhavan said it's just data backfilling. Which makes sense since there's been a "heads up" since day one in this report that "Results may be refined as additional data is processed"
In a separate reply he clarified: "we are constantly backfilling data. This is not about completeness. Just regular data processing pipeline artifact."
And in another reply: "trends is what I would focus on"
Would've been nice to get an annotation or alert inside Bing WMT instead of a month-delayed acknowledgement, but at least we know now. I guess put a sticky note on your monitor because there isn't a way to add your own annotations in this report. Thank you to Barry Schwartz for summing all this up!

Today I noticed PageSpeed Insights has an Agentic Browsing section and there's documentation for Lighthouse agentic browsing scoring, too.

Schema App has a new tool called Knowledge Assistant. Haven't tried it, but if it works as well as they describe, it sounds really helpful.
- Search/chat tool for your website
- Powered by Microsoft's NLWeb
- Users ask natural-language questions & get answers grounded in the website's Content Knowledge Graph (a knowledge layer that can govern multiple websites & other digital properties!)
- If the answer doesn't exist, it says so instead of hallucinating
- Can also be used by employees to find answers
- Lets you identify content gaps through its database of customer questions (super helpful)
There's a video on the page and a link to try a demo. I tried it out (I assume) in SchemaApp's own site search tool and it's pretty okay. It takes a little bit to give you an answer, which ends up being an overview summary statement, list of key facts, and previews of web pages it drew information from. Cool to know about.

Also Claude Sonnet 5 is now the new default model for Free & Pro (upgrade from Sonnet 4.6). They say it hallucinates less, isn't as sycophantic, and has more safeguards on by default.
