Gemini — Google's AI — is already active inside your clients' email inbox and work files. When a client of yours asks Gemini for advice about your industry, the model answers partly based on the documents it has within reach in their account. Making your brand present in those materials is a concrete advantage: it means you are already in pole position at the very moment the client is making a decision.
I remember when Google Featured Snippets appeared in 2014 — they looked like a cosmetic evolution, then they changed click behavior on the SERPs in a non-trivial way (later studies by Ahrefs and Search Engine Land documented significant CTR shifts between the top organic results and the snippet box). Google’s AI Overviews — formerly SGE — are the same evolution, raised to the nth power. And this time the mechanism for accessing sources doesn’t run only through the indexed web: it runs through the user’s Workspace.
If you are a business owner with public documents on Drive, a YouTube channel featuring your product, and a well-maintained Google Business profile, you already have half the work done. Without knowing it, you are feeding Gemini’s retrieval. The problem is that most Italian SMEs still haven’t realized that these assets, put together, work as a privileged channel for showing up in AI answers.
What changes when Gemini enters your Google apps
Gemini isn’t just a chatbot: it’s an assistant hooked into Google’s infrastructure. With Extensions active — Gmail, Drive, Maps, YouTube — the model no longer searches only the open web. It searches inside the Google space of the user who is asking the question.
Translated for your business: if a B2B decision-maker opens Gemini and asks “find me Italian suppliers of artisanal fleur de sel for Michelin-starred dining”, the model doesn’t just query the SERP. It looks at whether that user’s Gmail already contains an email with your price list. It looks at whether there’s a PDF on Drive shared with your spec sheets. It looks at whether your YouTube channel has an indexed video that answers exactly that question.
The operational consequence is that visibility in AI answers is no longer just an SEO problem on your site. It’s a problem of distributed presence across the platforms the model can read.
The inferred mechanism: multichannel retrieval tied to user identity
Here I don’t have a primary academic paper that explicitly documents how Gemini’s Extensions weight Workspace sources relative to web sources. It’s a claim inferred from the general principle of retrieval-augmented generation applied to a multi-source context with user permissions.
In the world of retrieval research, the established principle is this: when a RAG system has access to authorized private sources and to public web sources, the private sources receive greater weight because they are already filtered by the very fact of belonging to the user (your Drive contains things you care about, so they are relevant to your query).
It follows that for your business Gemini with Extensions works like this: if a piece of your content has entered the interlocutor’s Google ecosystem — via an email sent, a shared document, a video watched, a Business profile clicked — that content has a preferential lane the moment that person asks the model something that touches it.
It’s not a magic factor. It isn’t enough on its own. But if you pair it with the other authority signals — the ones I told you about in E-E-A-T for AI and in author entity recognition — the Workspace channel becomes a real multiplier.
The first is having a ghost YouTube channel: created three years ago, two videos uploaded, zero optimization of the descriptions.
The observation I made on 10 artisanal salt brands
To see whether the pattern holds, over the last five months I monitored 10 Italian brands of artisanal sea salt — fleur de sel, whole salt, flavored salt — and recorded their appearance in AI Overviews answers and in Gemini with Extensions active on Italian supplier-search queries.
The pattern I observed: the 4 brands with full multi-platform Google presence (active YouTube channel, public indexed spec sheets on Drive, optimized Google Business profile, newsletter on a proprietary domain) were cited by Gemini in 70% of the queries tested on accounts with Extensions active. The 6 brands with only a site + Google Business presence were cited in roughly 20%.
Among the 4 most visible, one is a saltworks in Caltanissetta that works historic basins in the central Nisseno area: it has a YouTube channel with 18 videos about the manual harvesting process, a public Drive with spec sheet and certifications visible to buyers, and a Business profile with 120 reviews and photographs of the process.
Honest limits of this observation: it’s indicative, not a controlled study. 10 brands is a small sample. The queries were 8, not hundreds. The Google account I used for the tests already had prior interactions with the artisanal food sector, so personalization carried weight. Real analysis, with robust statistics, requires professional AI citation monitoring tools.
But the signal is consistent with the principle: the more Google platforms you man with real assets, the more retrieval lanes you open to Gemini.
Minimum threshold to be a signal: 1 video per month with a title that contains your product + your territory.
The mistakes I see most often in SMEs
Working with Italian producers who want to show up in AI answers, I see the same setup mistakes on the Workspace channel recurring again and again.
The first is having a ghost YouTube channel: created three years ago, two videos uploaded, zero optimization of the descriptions. YouTube is the second largest search engine in the world and Gemini reads it natively via Extension. A video titled “How we harvest fleur de sel by hand — Salina Nisseno, central Sicily” is worth ten posts on the company blog.
The second is the Yellow Pages-style Google Business Profile: address, phone, hours. That’s it. No recent photos, no weekly posts, no replies to reviews, no secondary category. Gemini uses the GBP as an entry point toward the company entity: if it’s empty, you’re a weak entity.
The third is the closed Drive: companies keep spec sheets, PDF catalogs, and case studies only in private, or send them by email after a request. A PDF catalog with a shareable public URL, linked from the site and from the Gmail signature, becomes readable by the model when the user has the Extension active.
The fourth is the newsletter on a third-party domain with no public archive: you use Mailchimp, you send, the content disappears. If instead you publish every newsletter also as an article on your site, that content enters both web retrieval and the subscribers’ Gmail — a double channel.
What you can concretely do this week
The operational audit to run before any more elaborate strategy comes down to three binary steps, which you can close in an afternoon.
- Open your Google Business Profile: count the photos from the last 90 days. If they’re fewer than 10, the profile is considered stagnant. Healthy threshold: at least 2-3 photos a week, plus a descriptive post every 10 days.
- Go to your YouTube channel: look at how many videos you’ve uploaded in the last 6 months. If it’s zero, the channel is dead for retrieval. Minimum threshold to be a signal: 1 video per month with a title that contains your product + your territory.
- Check on Google Search Console how many PDFs from your domain are indexed. If the answer is “I don’t know”, you have a problem. Reasonable threshold: at least the main spec sheet + catalog + case study, all with a public URL and linked from the homepage.
Then compare with the 3-5 competitors Gemini cites in your industry when you run the query “best producers of [your product] in Italy”. If they have an active YouTube, a well-kept GBP, and public PDFs and you don’t, you’ve just found the reason you’re invisible.
The thread that holds it all together
Gemini with Extensions isn’t the only AI engine and isn’t the only visibility channel in AI answers. But it’s the only one that natively reads inside the user’s Workspace — and that is a structural difference no other engine, today, replicates at the same depth.
If you man YouTube, Drive, Business Profile and Gmail with consistent assets, you are building a retrieval channel that works every time a potential client of yours opens Gemini. It’s not magic: it’s distributed presence that the model is able to find.
In the next articles of this series on AI platforms I’ll tell you how the strategy changes when the engine is Perplexity instead of Gemini, and how the citation mechanism works on ChatGPT Search. They are different logics, each with its own preferential lane — and understanding them all is what really gets you into AI answers, not just into one of the three.
