Verisca daily brief
2026-09-25 Daily Brief
Google is updating advertising, search, and AI features at the same time, broadening both how users explore and what marketers can measure.
Brief
Google is simultaneously updating advertising, search, and AI features, expanding the way users browse and the scope of performance measurement.
-
01Advertising operations deal more closely with budget, creative, and conversion connections.
Demand Gen's new feature, Google Ads API 25.2, and investment/budget recommendation feature have been released.
-
02Search performance must also include AI and image-based search.
The exposure and measurement limitations of AI Overviews are discussed, and multimodal search reporting is added to Search Console.
-
03AI is also coming into the way search and customer touchpoints are run.
This was followed by Gemini's enterprise phone feature experimentation and presentation of Live Avatar and query fanout technology.
Today’s signals
AI environment changesGoogle Ads & Commerce Blog
Google Ads adds AI and one-click visit features to Demand Gen
Google Ads introduces Business Agent, one-click landing page visits, and Google Maps Promoted pins integration to Demand Gen.
The scope of use expands when designing the flow from attracting interest in video and image advertisements to site visits and conversions in year-end campaigns.
Google Ads has released Demand Gen updates to help you engage and convert audiences on YouTube and Google's visual channels. YouTube video ads with product feeds include Business Agent, a conversational AI, while YouTube Shorts and Gmail image ads provide one-click landing pages. Promoted pins in Google Maps also support Affiliate Location Extensions. According to Google Ads, advertisers who added Gmail to their Demand Gen campaigns saw an average 40% increase in image creative conversions at the same ROI.
View source ↗AI environment changesSeer Interactive
ChatGPT-5.6 Luna searches more frequently and executes fewer queries
In the Seer Interactive analysis, ChatGPT-5.6 Luna used web searches more frequently than GPT-5.5, but executed fewer queries.
When examining the content exposed in AI answers, it is necessary to also interpret changes in the model's search method rather than just looking at the number of citations.
Seer Interactive compared the search behavior of ChatGPT-5.6 Luna and GPT-5.5 with the same 346 prompts. ChatGPT-5.6 Luna had a higher usage frequency of web searches, but executed fewer queries. The change in the number of fanout queries compared to GPT-5.5 was not significant.
View source ↗AI environment changesSearch Engine Roundtable
Google tests bottom placement of Hotel Search AI Overviews
Google is testing AI Overviews in Hotel Search by displaying them at the bottom of the hotel knowledge panel instead of at the top of the page.
Hotel and travel content may have different exposure relationships with other search elements depending on where the AI response is placed on the search screen.
Google is testing AI Overviews' positioning in hotel search results. In the observed screen, AI Overviews were displayed at the bottom of the hotel knowledge panel rather than at the top of the page. However, as this was based on limited screen observation, the scope of application and whether it will continue has not been confirmed.
View source ↗AI environment changesSearch Engine Journal
Search Console aggregates AI overviews into one block
Search Console processes the entire AI-generated overview as a single block, and traditional rank tracking is difficult to implement, according to the explanation.
When looking at AI Overview performance, it is difficult to determine actual placement or result type based on average position alone.
As Google Search Advocate John Mueller explains, Search Console treats the entire AI-generated overview as a single block. Because of this, it is difficult to usefully implement AI features that track rankings like traditional search results. Average position may appear different from actual exposure because it combines differences in search location, device, and displayed result type into a single number, and this limitation is further increased in AI Overview.
View source ↗AI environment changesSearch Engine Roundtable
Google Ads API 25.2 released
Google has released Google Ads API 25.2, including competitive benchmark metrics and Performance Max draft generation.
Teams that connect their ad operations tools to the API can incorporate new metrics and campaign settings.
Google Ads API version 25.2 has been released. This minor release includes competitive benchmark percentile metrics. There are also updates to the ability to create Performance Max drafts in Smart campaigns, asset group URL options, and Search bid adjustments.
View source ↗AI environment changesThe Verge AI
Gemini experiments with feature that calls local businesses on behalf of Pixel 11
Google is launching an Early Experiment feature on Pixel 11 that lets Gemini make calls to local businesses for you.
When a contact point is created that requests reservations and inventory confirmation through AI before a customer calls, the local business' phone response experience also comes into the scope of content and operations.
Google is introducing an early experimental feature on the Pixel 11 that lets Gemini call local businesses on your behalf. Users can rely on Gemini to make reservations, check product availability, and reschedule reservations. According to Google, users don't even need to initiate the call themselves.
View source ↗AI environment changesThe Verge AI
Gemini 3.8 Live provides Live Avatar that reacts in real time
Gemini 3.8 Live features Live Avatars, where animated AI personas react in real time.
As the screen itself that communicates with AI becomes part of the user experience, visual interactions other than text and voice are also treated as customer contact points.
The Gemini 3.8 Live update allows users to view animated AI personas and interact with models. Live Avatar matches mouth movements and shows various facial expressions during conversation. This feature is currently available on Gemini Ent… It is only available in .
View source ↗AI environment changesSearch Engine Roundtable
Google documents VideoObject’s creator/author support
Google has documented support for creator and author attributes in VideoObject structured data and has also clarified support for interactionStatistic.
Teams connecting video content to search can see the extent of markup support for creator/author information and interaction data.
Google has updated its video structured data documentation. Support for creator and author attributes in VideoObject structured data has been added to the document. Support for interactionStatistic has also been provided more clearly.
View source ↗AI environment changesSearch Engine Roundtable
Google Ads adds Holistic/Growth mode for budget reallocation
Google Ads has added budget allocation using Holistic and Growth modes to the Recommended Investment Strategy tab.
When reviewing your budget for each campaign, you can also check how spending is reallocated based on performance in the recommendation screen.
Google Ads has updated the Recommended Investment Strategy tab on the Recommendations page. The new feature is a way to strategically allocate your budget using Holistic and Growth modes. Holistic mode reallocates spend from low-performing campaigns to top-performing campaigns to increase efficiency. Details of Growth mode are only partially included in the original summary provided.
View source ↗AI environment changesSearch Engine Roundtable
Google Ads Shows Forecast-Driven Budget Change Recommendations
Google Ads is showing budget change recommendations to some advertisers with an estimated weekly conversion value.
When reviewing a campaign with budget constraints, the predictions in the Recommendations screen can serve as a basis for validating your current settings.
Google Ads is showing recommendations linked to predictions on some advertisers' screens. These recommendations come with changes you can make right on your screen. The screen will indicate that your campaign is being limited by your budget, along with an estimated weekly conversion value.
View source ↗AI environment changesSearch Engine Journal
Google announces query fanout technology R4T-Diffusion
Google announced the R4T-Diffusion query fanout framework, which combines reinforcement learning, synthetic data, and a 53.9 million parameter diffusion model.
As AI Search changes the way it breaks down a question into multiple subqueries, the search path through which your content is found may also change.
Google has released its Retrieve-for-Train-Diffusion, or R4T-Diffusion, query fanout framework. This technique distills the high-quality output behavior of a large model into smaller models to generate subqueries that take relevance, variety, and sorting into account. Google suggests a speedup of 12 to 20 times compared to autoregressive methods.
View source ↗AI environment changesAdweek
Scope3 changed its name to Apostra and converted to agent-to-agent advertising transactions.
Scope3, an adtech company, changed its name to Apostra and converted to an infrastructure layer that supports advertising buying and selling between agents.
This is an example of an ad trading infrastructure shifting from being a carbon footprint calculator for advertisers to supporting agent-to-agent transactions.
Scope3 changed its name to Apostra. The company transitioned from a carbon footprint calculator for advertisers to an infrastructure layer that supports the buying and selling of ads between agents. Brian O’Kelley reacted to possible criticism of this change in a conversation with ADWEEK.
View source ↗AI environment changesSearch Engine Journal
LLM Answers Can Clear Clues in the Investigation Process
Search Engine Journal pointed out that LLMs and answer engines may strip out path metadata they obtain during the research process in return for providing quick answers.
When using the answers provided by AI for content research, you must separately check not only the conclusion but also the amount of evidence, inconsistencies between sources, and whether it is unexplored.
LLM and its answer engine can reduce the time it takes to research and search to seconds. On the other hand, the explanation is that path metadata such as the amount of evidence obtained in the process of reaching the answer, inconsistencies between sources, and whether the topic is unexplored may also disappear. In this case, the user may reach a conclusion without any means of assessing the level of evidence behind the answer. Wharton's Shiri Melumad and Jin Ho Yun published the results of seven experiments involving 10,462 participants in PNAS Nexus in October 2025.
View source ↗UnclassifiedGoogle Search Central
Search Console adds multimodal search performance reporting
Google Search Console adds a reporting function that allows you to check and export web multimodal search data through Lens and Circle to Search.
Search inflows that start with images can also be checked with a separate filter, allowing you to examine image-based content and search performance together.
Web multimodal search reporting is being added to the search results and generative AI feature performance reports in Google Search Console. Search data from Lens, Circle to Search on Android, Google Search Image Upload, and ‘Search this image’ on Chrome can be viewed and exported in the new multimodal search type filter. This feature will be rolled out globally starting today, and you will see metrics in your performance reports for sites that receive traffic from relevant queries.
View source ↗UnclassifiedETNews
Naver provides airport, parking, and content information in celebration of Chuseok
During the Chuseok holiday, Naver provides daily life information, travel, content, and payment benefits through Search, Place, Clip, Chijijik, and Naver Pay.
This is an example of a service operation that connects search, local information, content, and payment during times when users' immediate demand for information increases, such as during the holidays.
To celebrate the Chuseok holiday, Naver provides holiday lifestyle information and travel, content, and payment benefits in its major services. Search and Place provide information on airport waiting times, hospitals and pharmacies operating during holidays, and free openings of over 9,000 public parking lots. Klip, Chijijik, and Naver Webtoon also operate holiday content and events.
View source ↗How to read this
This archive preserves the selected sources and links. WHAT CHANGED summarizes the observed development. WHY IT MATTERS explains why it deserves monitoring. Each item remains an external signal until independently verified.