01 The platform change
Google connected Gemini to 13 more named services
When an assistant can act inside a named partner service, every business outside that list is quietly routed around.
Google’s list runs from Squarespace and Webflow to apartments.com and Experian, and the assistant can now work inside them on request.
Google added 13 services to the Gemini app on September 23, in a post written by Mai Lowe, a group product manager on the Gemini team. The list spans three categories: Airtable, Linear, monday.com, PandaDoc, Wispr AI and Zoho for productivity; Adobe, Picsart, Squarespace and Webflow for creative work; apartments.com, Experian, Peloton and SeatGeek for everything else.
The apps are not really the point. The point is that a customer no longer has to leave the assistant to get something done. Google’s own framing is that “instead of switching between tabs, you can now manage projects, design creative assets, and plan your workouts all in one place.”
A person connects a service in Gemini settings, or pulls one into a conversation with an @ mention. So the assistant becomes the place the work happens, and a named service is what it reaches for when the work needs doing.
Read the real estate entry closely, because it is the shape of the thing. apartments.com is on the list and an independent brokerage is not. When somebody asks Gemini to help them find a place to live, the assistant has a named partner to work inside, and every business outside that list is reached only if the answer happens to mention it.
Nobody outside these companies decides who joins the list. The post names no application process and states no criteria for being added.
What to do about it
Check whether a service your customers actually start from is on that list of 13. If it is, your presence there stopped being a directory entry this week and became a surface an assistant acts inside. If it is not, nothing has changed for you yet, and that is worth knowing too.
Source
Google, “New connected apps in Gemini,” September 23, 2026 · blog.google
02 The consumer data
2,002 UK adults reported relatively high trust in AI answers
Customers are inclined to believe whatever an assistant says about your business, including the parts of it that are out of date.
Florence Enock and Helen Margetts found high trust in AI answers sitting alongside widespread concern, inside one nationally representative sample.
Florence Enock and Helen Margetts surveyed a nationally representative sample of 2,002 UK adults about how they use AI assistants and how far they trust them. The paper went up on September 23.
Their finding on trust is the one that matters most here. Trust in the information these systems produce is, in the authors’ own words, relatively high, and it sits alongside worry rather than instead of it. 77% of the sample said they felt enthusiastic about the benefits, while 70% said they were concerned about the risks.
Use is also drifting well past the practical. 31% of regular users said they turn to an assistant for personal and emotional support, such as talking a problem through or asking for help with a decision, and a quarter said they use one for the conversation itself.
One caveat belongs on this before anyone quotes it. The sample is British, and American attitudes are not measured here, so treat the direction as informative and the exact figures as local.
For a business the consequence is narrow and concrete. When somebody asks an assistant who to call, or whether your practice is any good, they are inclined to accept the answer they get. That inclination does not stop to check whether the answer is current.
What to do about it
Ask an assistant the question a customer would ask, in the plainest words they would use, and read what comes back about your own business. This survey says a customer tends to believe that answer. Knowing what it currently says is the cheapest thing on this page.
Source
arXiv (Florence E. Enock, Helen Z. Margetts), “Large Language Models in the UK: Public Use, Trust, and Attitudes,” September 23, 2026 · arxiv.org
03 The research
Researchers tested an AI citation tactic and counted its losses
Methods that win you citations also cost some pages the citations they had, and the sales version of this never mentions that.
Shilpa Ramakrishna and William Andreopoulos measured a technique that raised citations on benchmark documents, and reported the losing column as well as the winning one.
Shilpa Ramakrishna and William Andreopoulos posted a paper on a technique they call Query Implied Generative Engine Optimization. Generative engine optimization is the practice of getting your pages named inside AI answers. The paper went to arXiv in August, so it is context rather than news, and it is the clearest public measurement of whether this kind of work does anything.
The method reads a document, infers the questions that document could answer, and identifies material it is missing. Tested against two benchmark collections, it raised what the authors call objective scores by up to 15.9% and subjective scores by up to 17.6%.
Then there is the sentence worth reading twice. The authors report the method produced “nearly twice as many citation gains as citation losses.” Nearly twice as many gains as losses means there was a loss column. Some documents came out of the treatment cited less often than they went in.
That ratio is the honest shape of the result, and it is the part a vendor summary omits. It is also a benchmark score rather than a live-web score, which is a real limit on what it can prove about anybody’s actual website.
Forever Cited sells this category of work, so read the previous paragraph as disclosure rather than modesty. The published evidence reports a gain column and a loss column, and anyone quoting only the first is quoting half a table.
What to do about it
When anyone shows you a number for AI visibility work, ask what it was measured on and ask what went down. A method with no loss column has either not been measured properly or has not been reported honestly.
Source
arXiv (Shilpa Ramakrishna, William B. Andreopoulos), “Query Implied Generative Engine Optimization,” August 19, 2026 · arxiv.org
04 The operational change
Google documented how to name a video’s creator
If your business publishes video, there is now a documented way to tell Google which person or company made it.
Google’s update records support for the creator and author properties in video markup and clarifies which interaction counts it reads.
Google updated its video structured data documentation on September 24. Structured data is the hidden labeling on a page that tells a search engine what the page holds, and video markup is the part that describes a video.
The changelog entry states the update is there to “document support for the creator and author properties, and clarify supported interaction types for interactionStatistic in VideoObject structured data.” Put plainly, Google now documents how a page states who created a video, and which view or engagement counts it will actually read.
This is documentation rather than a ranking announcement, and Google claimed no effect on how anything performs. Treat it as a mechanism becoming official, not as a promise about visibility.
It still belongs on this page for the same reason as everything else. A video that never says who made it is a video whose credit has to be guessed at, and an insurance broker or an admissions consultant publishing explainer videos is precisely the case where the named person is the reason anyone watches at all.
What to do about it
If your business publishes video on its own website, send the updated documentation to whoever maintains your markup and ask whether the creator property is set. If you publish only on YouTube or social platforms, this one is not yours to action.
Source
Search Engine Roundtable, “Google Video Structured Data Gains Creator Property, Updates interactionStatistic,” September 24, 2026 · seroundtable.com
05 The adoption data
Google measured which occupations use AI most at work
AI is already routine inside professional work, which is why customers now expect to find your business through it.
Google’s Atlas puts computer and mathematical work at 30% of United States work-related AI use, and surveyed scientists report saving close to seven hours a week.
Google published a research project called the AI and Economy Atlas on September 15, tracking how AI is being taken up across occupations and regions. It is an open, interactive dataset rather than a product announcement.
Two figures carry it. In the United States, computer and mathematical occupations account for 30% of work-related AI usage, double the share seen across the rest of the world. In India, arts, design and media occupations account for 19%, which Google puts at 1.6 times the global average.
The most concrete number is about time. Nearly half of the scientists surveyed said they use some form of AI every day, and they reported saving almost seven hours a week. That survey covered more than 600 scientists in the United States and the United Kingdom.
Be clear about what this measures and what it does not. It measures how professionals use AI to do their own work, not how customers use it to choose who to hire. The reason it belongs here is the habit. A tool that gives a working professional most of a day back each week is a tool they will also reach for when they need to decide who to call.
What to do about it
Treat this as the denominator for the rest of the page. The habit is established and not in question; the only open item is whether your business is visible inside it.
Source
Google, “AI and Economy Atlas, September 2026,” September 15, 2026 · blog.google
