One human consultant. Six lightning-fast AI agents. One bench.
Being chosen today means appearing in the answers generated by artificial intelligence. Most searches and choices now start there.
See right now, for free, how AI sees your page.
No access to your siteA few minutesNo commitment
Got your result? A senior consultant reads it through with you, and tells you what is worth doing.
Book the conversationWho reads your site
A traditional program runs a checklist written months earlier and sees nothing else. An agent is given a goal, decides where to look, puts the clues together and forms a judgement. It is the difference between a metal detector and an expert appraiser. There are six because a single agent asked to look at everything looks at everything badly. They work in parallel and pass on what they find: when Agent Scout discovers a section closed to crawlers (the programs search engines send out to read websites), the other agents know at once and take it into account. And they work for the human consultant, not in their place: the consultant opens the job, chooses what matters and signs.
Moves through the site the way an engine would: follows the links, reads the instructions left for crawlers, checks that every page can be reached, that redirects don't loop, that the certificate and security headers are in order. If Agent Scout can't read a page, neither can Google, and neither can AI.
Loads the site and times it: how long the content takes to appear, how much elements shift while the page settles, how quickly it responds to a tap. These are the Core Web Vitals, the three measures Google really uses to decide rankings. It also watches for the errors a site produces silently while loading, such as 404 errors: nobody notices them, and they break the page for visitors and engines alike.
Classic SEO: titles, descriptions, content hierarchy, usability on a phone, and structured data, the invisible labels that explain to the machine what it is reading.
Checks whether the site is written so a machine can lift an answer out of it: definitions, question-and-answer blocks, the brand declared as an entity. It is the area where most sites are weak.
Looks at the site through the eyes of ChatGPT and Perplexity: text that can be extracted or is buried in scripts, cited sources, a brand recognisable beyond the site, AI crawlers admitted or blocked unknowingly, freshness signals, the presence of an llms.txt file.
Gathers the problems found by the other five, compares them with your competitors and rewrites them as actions in order of value: what to do, with what return, at what effort, at what cost.
An inspection that would cost a person weeks is finished in minutes. An audit that costs weeks is done once and grows old; one that costs minutes is repeated after every correction, and becomes a working tool.
The first step
Half an hour with a senior consultant, no commitment. Your website is all we need: with the six agents at work, they go through it and tell you what is worth doing, and what is not.
Everything will be decided with you: no fix without your go-ahead. They write to you to set a day and time. Or drop a line to colloquio@labottega.app.
Who decides
At Bottega we call them masters, the Florentine way, because they open and close every job and answer for it with their signature. They are senior consultants only, with decades of work in visibility: professionals who have taken businesses in the United States and Europe to results that changed their history. They know search engines from the inside, from the written rules to the ones you only learn in the field, and today they read the answers of artificial intelligence with the same precision they bring to Google's first page.
The agents are fast. The consultant knows where to look.
What the machines see
This is an extract from a real analysis of the home page of a major national newspaper. A huge newsroom, unquestionable authority, and a 62 on answer engines.
The method, in brief
The market offers you numbers or a document. We offer you a consultant who listens and decides, six agents who do weeks of work in minutes, and proof of every correction. It is the reason we exist.
Before any analysis, a person. They understand what you sell, to whom, and what has to happen for your business to grow.
Selling airline tickets and selling preserves are not the same trade. Machines don't know that. The consultant does.
Six agents open your site together and within minutes report what Google, the answer engines and ChatGPT see. The consultant chooses what matters.
Three grades, eighteen areas, every fix with its impact and effort stated. And the candour to tell you what is not worth doing.
You authorise, one fix at a time. The agents carry out the technical corrections; the consultant reopens the code and shows you the line.
Copy and design stay yours. We measure again, and the grade tells you how much has changed.
No machine decides alone. The consultant steers, the agents run, the code proves it.
Why everything changed
For twenty years Google showed a list of ten links: being on it was enough. Then it started answering on its own, in a box at the top of the page, citing two or three sources. Today artificial intelligence writes the answer and names someone, often just one. Search engines no longer read your site: they question it. It is no longer a ranking. It is a selection.
Ten blue links, in order of score. Whoever made the first screen had a chance. For twenty years it was the only goal.
What it took: keywords, links from other sites, speed.
Google answers on its own, at the top of the page, and cites two or three sources. The rest of the list scrolls by underneath, often without a click. Since 2024 the box is written by AI.
What it takes: the answer written in a form the machine can lift out.
ChatGPT, Gemini and Perplexity read dozens of sources, write a paragraph and name someone. Whoever is missing, to the person who asked, does not exist.
What it takes: a recognisable brand, extractable text, sources, AI crawlers allowed in.
Being the name AI puts forward is the new first place.
The three disciplines
They are three different disciplines competing for the same website. A fix that improves the first may not move the second or third by a millimetre. That is why we measure them separately, each with a score from 0 to 100 and a letter from A to F, and beneath each grade the areas that make it up.
The classic trade, the one that has existed for twenty years. When someone searches on Google, the engine lines pages up by a score: how the page is built, which words it uses and where, how fast it is, whether it reads well on a phone, and how many other sites cite it, and which. If your page appears on the first screen you have a good chance of being chosen. On the second, almost none.
It still matters, and it is the foundation the other two rest on: a site Google cannot read, no artificial intelligence can read either.
For a couple of years now, Google has been answering on its own. Above the results sits a box, the AI Overview, that summarises the answer and cites two or three sources; voice assistants do the same, out loud. Whoever lands in there takes almost all the attention; the rest of the list scrolls by underneath, often without a click.
The box goes to whoever wrote the answer in a form a machine can lift out: the definition up front, question-and-answer blocks, the label that declares "this is an FAQ", the brand name, that is the name of your company or product, declared as an entity rather than a mere word. It is the area where most sites are weakest, because until recently it wasn't needed.
Millions of people no longer open a search engine at all: they put the question straight to ChatGPT, Gemini, Perplexity or Claude. These systems show no list. They read many sources, write a paragraph and name very few. Being among those names is the new first place, and it is not won with the same tools as SEO.
Artificial intelligence rewards text that can be extracted (not buried in images or scripts), declared sources and citations, a brand recognisable beyond the site, fresh updates and, quite simply, permission to enter: many sites block AI crawlers without knowing it, and vanish from the answers.
Customers no longer reach you by picking you out of a very long list. They come because a machine chose you from a shortlist of very few names. Our job is to make sure yours is on it.
The five steps
The consultant opens and closes every job. In between, the agents run. At every step you know who does what, and who answers for it.
A conversation, before anything else: your site, your trade, the competitors you want to reach. If you have already taken the free first look, we start from there.
The consultantSix agents read the whole site, in parallel and from the outside, as an engine does: no access, no installations. Within minutes, three grades and eighteen areas, every problem explained.
The AI agentsThey read what the agents found, compare it with your competitors and turn it into a plan in order of value: impact, effort, steps adapted to your platform, and who does them, you, your developer, or us. They clear up doubts, take your objections on board, and shape the plan to the way you see things.
The consultantYou authorise every fix, one by one. The agents carry out the technical corrections: structured data, crawler directives, llms.txt, performance. Copy and design are not touched.
The AI agentsThey reopen the code and check that every change is really there, with the line cited. Then we measure again. The grade tells you how much has changed.
The consultantAnd the free first look, the one you start yourself by pasting in an address? It is the agents' calling card: within minutes, the three grades and the problems found. The real work begins with the consultant.
The document
Not an endless list of anomalies, nor a PDF of generic advice. Every recommendation arrives like this.
The site is famous to a person and ambiguous to a machine: neither answer engines nor generative engines find a declaration of who you are. This directly limits citation.
No question-and-answer block found. It is the single biggest obstacle to being cited by AI.
Why us
The market sells two things: a tool that measures, or hours of consulting. We sell the third: a result proven in the code, decided by a person, obtained at machine speed.
Semrush, Ahrefs, Screaming Frog measure well. Then they leave you four hundred lines of anomalies and nobody to tell you which one matters, in what order, and whether it is worth it.
A hand-made audit costs weeks and photographs a single day. Six months later it describes a site that no longer exists, and nobody has gone back to check the code.
A consultant who listens to you and signs, six agents who do weeks of work in minutes, and proof in the code of every correction, repeated every cycle. Of some thirty Italian and international price lists we compared, none puts the three together.
Show me a website and I will tell you what the machines see.
| What you get | Analysis software | Traditional agency | Bottega |
|---|---|---|---|
| Separate SEO, AEO and GEO analysis | Almost always SEO only | Rarely AEO and GEO | Three distinct scores |
| Prioritised by value | Hundreds of lines, no hierarchy | Yes, in the document | Impact and effort on every action |
| Comparison with cited competitors | Only on request, at extra cost | Sometimes | Included |
| Steps adapted to your platform | Generic advice | Yes | Yes, with who does them |
| Who carries out the corrections | You, on your own | The agency, by the hour | The agents, after your sign-off |
| Verification in the code, with proof attached | No | Usually a tick in the report | Line cited, every time |
| Re-measurement after corrections | If you run it yourself | At the next contract | Every cycle, even daily |
| A person who answers for the decisions | No | Yes | Yes |
The written method
A written method that forbids us from flattering you. Language models have a well-known flaw: they tend to agree with whoever is questioning them, present assumptions as facts, and always find something positive to say. Applied to an audit, that flaw produces polite, useless documents.
The Rules of Bottega are the answer: not a value hung on the wall, but instructions the agents load before forming any judgement, versioned like code and corrected when they get things wrong. Like the statutes of an old Florentine guild, written for whoever sits at the bench.
Convincing answers are everywhere.
Reliable ones have to be built.
Pricing
The first look is free. What you pay for is the bench: the consultant who listens to you, the comparison with competitors, the agents' work, and the verification that it really reached its destination.
Single-language site, up to fifty pages
SMEs and professional firms, structured site, possibly a second language
Multi-site, multilingual or multi-brand organisations
The quote comes after we have seen the site, because the amount of work determines the price. A price list drawn up blind would be a convenient lie. The first look costs nothing.
Questions
Senior consultants with decades of experience in visibility, who have taken businesses in the United States and Europe to first-rate results. At Bottega we call them masters: the consultant who takes on your job opens and closes it, signs every decision and verifies every correction in the code. The agents work for them, not in their place.
No, quite the opposite. The consultant is a person: they open and close every job and sign every decision. The agents work for them: they do in minutes what would take weeks by hand, so the person's time goes where it matters, to understanding your business and deciding.
Yes. Paste in your address and within minutes you receive the three grades, SEO, AEO and GEO, with the list of problems found. You pay only for the work at the bench: the conversation with the consultant, the plan, the fixes and the verification.
Not for the first look: we view the site from the outside, exactly as a search engine does. Access is needed only if you authorise us to carry out the corrections.
No. We work on the technical packaging: code, structured data, text invisible to users, crawler directives, performance. Content and design remain yours.
You decide: you can do them yourself, your developer can follow our brief, or we can do them. If you authorise us, the agents carry out the technical work and the consultant verifies every correction in the code.
After every correction we reopen the page's code and check that the change is really there, citing the line. Not a tick on a spreadsheet: proof.
No, and anyone who promises one is not being honest: it depends on third-party systems that change without notice. We guarantee the method, the verification of every intervention, and the candour to tell you when something is not worth doing.
The first look is free: three grades and the problems explained, in minutes. Then, if you wish, the consultant.
Then a senior consultant goes through it with you.
Book the conversation