Google's AI Overview — that box that now appears above all the results — is not a simple summary: it's a new layer that decides who gets read and who gets skipped, completely rewriting the rules of who makes it into the answer. If you're not already in the very top organic results, for this new feature you are invisible by definition. The problem is that many business owners are optimizing their site for the old Google, without knowing that the rules have changed. Adapting your existing strategy takes less than you might think.
I remember when Google launched Hummingbird in 2013 — it changed everything. We moved from keywords to entities, from “exact match” to the meaning of the query. Those who understood the shift won five years of visibility; those who stubbornly hammered repeated keywords disappeared.
With Gemini it’s happening a second time, but more deeply: search becomes conversation. The AI Overview is not a new box above the results — it’s a new layer that rewrites the way a business owner or a producer enters (or doesn’t enter) the answer shown to the end customer. And the interesting thing is that the prerequisite for getting in remains your traditional SEO ranking: if you’re not in the top organic results on your target queries, you’re not even selected as a source for the AI Overview. Let me explain why this changes the priorities of a producer of PDO buffalo mozzarella in Campania, but the reasoning applies to every Italian SME that lives on organic search.
What the AI Overview really is: a conversational model on top of the ranking
Google built the AI Overview on top of the Gemini family. It’s not a separate engine: it’s a generative layer that takes the already-ranked results, synthesizes them, and turns them into a direct answer for the reader. To understand how it behaves, you need to know where it comes from.
In the search world, the Gemini team described its family of models like this:
The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases.
Translated: Gemini is not a single model, it’s a family with different sizes. The largest size (Ultra) reasons over complex tasks; the small one (Nano) runs on the phone. The AI Overview you see above the Google results uses intermediate sizes, optimized for speed and volume.
The business implication is direct: the summary that appears to your end customer is not written by your site. It’s generated on the fly by a model that chooses which pages to cite among those already on the first page. If your dairy is in position 14 for “best buffalo mozzarella Caserta”, the AI Overview doesn’t see you.
Why organic ranking is the entry ticket
This is the point that escapes many business owners when they ask me “should I throw away SEO and do only GEO?”. No. The AI Overview does not replace traditional SEO — it amplifies it. Ranking is the minimum condition; citability in the generated text is what makes the difference.
In the previous articles of this series I explained how AI engines assess authority and citability: if you haven’t read them yet, start with E-E-A-T for AI and with backlinks as a citation proxy. The AI Overview is the point where these two pieces meet: solid ranking plus authority signals equals a citation in the summary.
Anonymous “about us”: the dairy has three generations of history but the page is written in the third person as if it were talking about another company
The test you can run in 20 minutes on your industry
If you produce PDO buffalo mozzarella in Campania, the test is simpler than you think. You only need your phone and Google Search Console.
First step: open Google Search Console, go to “Performance” and filter the queries that contain the name of your product + a city (example: “buffalo mozzarella Caserta”, “PDO Campania buffalo”, “Aversa dairy”). Note the 10 highest-volume queries where you’re on the first page.
Second step: take those 10 queries and enter them one by one into the Google bar from the browser in incognito mode. See whether the AI Overview appears. If it appears, check whether your domain is among the sources cited on the right. Binary rule: you’re cited or you’re not. There are no half measures.
Third step: same procedure on Gemini. Ask directly “best producers of PDO buffalo mozzarella in the province of Caserta” and see whether your brand appears in the list. If it doesn’t appear, you have an AI visibility problem that isn’t solved just by improving your Google ranking.
This is an entry-level check, useful to understand the current state. The real analysis — the one that tells you why you’re not cited and what to change — requires professional tools and a few weeks of observation.
Open Search Console and isolate the 15 main queries where you’re on the first page.
The test I ran on 15 buffalo brands
I took 15 dairies across Caserta, Aversa, Battipaglia and southern Lazio — all producers of PDO buffalo mozzarella with an active site and at least one Google first page on geo-localized queries. I submitted 12 different queries to Gemini over three days, varying the wording: “best buffalo dairy Caserta”, “where to buy real PDO mozzarella”, “historic Campania buffalo producers”, and so on.
The pattern that emerged, in summary:
- 9 brands out of 15 were cited at least once across the 12 queries
- 3 brands were cited in at least half of the queries (six or more times)
- 6 brands never appeared, despite being on Google’s first page for their reference queries
The 3 brands cited recurrently had three things in common: a populated Wikidata entry, backlinks from food-and-wine trade publications, and an “about us” page where the producer signed the dairy’s story under their own name (not as an anonymous corporate entity). A natural link to what I told you in Author-Entity Recognition and in Google Knowledge Graph entry.
Limits of the test: small sample (15 brands), short time window (3 days), Gemini only. It’s not a study, it’s an indicative test. The pattern, however, is consistent with what I’ve been observing for months across other Italian food B2C sectors.
The multimodality that’s coming (and why it matters for you)
Gemini is not just text. In the launch paper the team explained visual reasoning like this:
The visual encoding of Gemini models is inspired by our own foundational work on Flamingo, CoCa and PaLI, with the important distinction that the models are multimodal from the beginning and can natively output images using discrete image tokens.
Translated for those who don’t speak machine learning: Gemini is born multimodal. It’s not a text model to which they glued vision afterwards — it understands text and image from the very first training. The operational consequence for a buffalo mozzarella producer is very concrete: the photos of your facility, of the buffaloes at pasture, of the stretching in the lab are no longer “graphic decoration”. They become a signal. With well-written alt text, clean EXIF metadata and consistent `Product` schema markup, the image weighs in the model’s decision about who to cite.
The mistakes I see most often on Campania dairies
After following several food B2C producers over the last few years, the patterns that repeat are always the same:
- Bilingual site without hreflang: the English version competes with the Italian one and neither of the two emerges in a stable way
- Anonymous “about us”: the dairy has three generations of history but the page is written in the third person as if it were talking about another company
- Organization schema missing or incomplete: a single check with Google’s Rich Results Test is enough to see it
- No presence on Wikidata: the Wikidata entry doesn’t exist and the brand remains “invisible” to the knowledge graph
None of these mistakes is a magic factor on its own. But added together they explain why a serious dairy with an excellent product gets ignored by the AI Overview while a less structured competitor gets cited.
What to do concretely this week
If you’re a producer of PDO buffalo mozzarella (or, more generally, a food B2C SME) and you want to start moving on solid ground, here’s an operational audit in three steps:
- Open Search Console and isolate the 15 main queries where you’re on the first page. They’re your candidates for the AI Overview.
- Open Gemini and run those 15 queries, note whether you’re cited. Compare with the 3-5 competitors that the AI does cite in your sector: what do they have that you don’t?
- Check schema markup and Wikidata: first with the Rich Results Test, then verify whether your brand has an entry on Wikidata. If it doesn’t, that’s the first gap to close.
These three steps don’t solve everything — but they give you the map of where you are. From there the real work begins.
Where I’m taking you in the next articles of the series
Visibility in AI answers isn’t played out only on the AI Overview. In the next articles of this series we’ll talk about how citation patterns change on Perplexity compared to Gemini, how ChatGPT handles cited sources without having its own search index, and when it’s worth optimizing for one AI engine rather than another. The common thread always stays the same: showing up in AI answers when your end customer asks the model for advice, and doing it in a way that’s sustainable over the long term.
