Still searching Google like it's 2010?
You need a supplier of food-grade silicone rubber gaskets, delivery within 48 hours, in Emilia-Romagna. What do you do?
You probably open Google, type something, scroll through the results, open four tabs, compare the sites, close three of them, go back, try another search. Twenty minutes later you still have doubts.
This is the search model we have been using for twenty years. It works. But it is getting slow.
How Google worked (and still works)
Google is a library. You walk in, give a keyword to the librarian, and they bring you a list of shelves ranked by relevance. Then you browse the books yourself.
Your search: food grade silicone gasket supplier Emilia-Romagna fast delivery
What you get:
- 10 blue links
- Some sponsored ads at the top
- Maybe a map with nearby companies
- A “People also ask” section
Google does not answer your question. It shows you where the answer might be. You have to find it yourself.
This system has a problem: it is optimised for keywords, not for intentions. If you do not know exactly what to search for, the library becomes a maze.
What changed: Google’s AI Overview
For some months now, a blue box with a direct AI-generated answer appears at the top of Google results. It is called AI Overview.
Your same search, with AI Overview:
“For food-grade silicone gaskets with fast delivery in Emilia-Romagna you can consider suppliers such as Rubbertech Bologna, Gommatech Modena, or search the Confindustria Emilia portal. Materials compliant with FDA/EU 10/2011 regulations for food use are available from specialist distributors in the Bologna and Reggio Emilia industrial areas.”
Better. But there is a problem: that answer is inferred. The AI took information from various sites, combined it, and produced a synthesis. The names it cites may be real, may be partially wrong, may be up to date or not.
You do not know what is underneath. You cannot easily verify it. And above all: does that answer talk about you or about your ideal supplier? It does not know.
The conversational model: a different kind of search
Now take the same problem and go to a conversational assistant — ChatGPT, Claude, Perplexity, Gemini.
You do not type keywords. You talk.
“I run a food company in Bologna. I need to find a supplier of silicone gaskets compliant with EU 10/2011 regulations, preferably in Emilia-Romagna, with stock availability and delivery within 48 hours. We produce around 200 pieces per month in various sizes. Do you have suggestions on how to evaluate suppliers and what to ask during the quoting stage?”
What you get is not a list of links. It is a conversation:
“For food-grade silicone gaskets EU 10/2011 in Emilia-Romagna, the main criteria to check are: material certification (request the specific compliance declaration, not a generic one), stock availability vs made-to-order (for 200 pieces/month a distributor with stock is better than a manufacturer), and quote response time (a signal of internal organisation).
Questions to ask the supplier: Do you have the EU 10/2011 compliance declaration for this specific compound? What is the lead time for urgent reorders? Do you manage documentation for food quality audits?
Would you like me to help you draft an email to send to multiple suppliers at the same time?”
The difference is substantial. The assistant did not give you a list — it helped you do the work.
Three models compared
| Traditional Google | AI Overview | Conversational assistant | |
|---|---|---|---|
| What you get | List of links | Synthetic answer | Dialogue and support |
| Who does the work | You | AI + you | AI with you |
| Personalisation | None | Low | High |
| Verifiability | High (go to the source) | Medium | Depends |
| Usefulness for complex decisions | Low | Medium | High |
| Time required | High | Medium | Low |
Why this changes something for your business too
Until now, having a website meant being found on Google. You just needed the right keywords, updated content, a few inbound links.
With conversational search the rules change.
When a potential customer asks an AI assistant “who is the best metal fabrication supplier in the province of Modena?”, the assistant does not search Google. It builds an answer from what it already knows — from what it has read, indexed, understood.
If your company’s website is not structured in a way that a machine can understand it, you do not exist in that answer. It does not matter how many keywords you have optimised.
Your website must speak to machines, not just to humans.
The current problem: not everything works well yet
Careful: the conversational model is powerful but not perfect.
AI assistants hallucinate — sometimes they invent information that sounds plausible but is false. Company names, phone numbers, certifications: everything can be wrong if the company’s website does not provide structured and verifiable data.
This creates a paradox: the more people use AI assistants to find suppliers, the more critical it becomes that your company is correctly represented in the sources these machines read.
Those who do not address this today risk being misrepresented — or not represented at all — tomorrow.
What to do now
There is no need to throw away everything you have done on Google. Traditional search still exists and will continue to exist.
But if you want to be found in the conversational model too, the first step is understanding how a machine reads your website today.
There are free tools that do this. We built one: it is called Lagotto Meter and it analyses in a few seconds how readable your site is for an AI agent.
It is not a guarantee of being found. It is the starting point for understanding where you stand.
lake8.dev — San Pietro in Casale, Bologna, July 2026
Sources
Author: Giantommaso Fogli
Publication Date: 2026-07-10
Rights and attributions
Images, logos and photographs are the property of their respective owners. Used for commentary purposes.
← Back to blog