Guglielmo Paesano AI engineer and founder of AIGROW

I build AI systems that do people's work, and I measure where they still need people.

contact@guglielmo.ioDiscord, 4K members

AIGROW

The company I founded: open research on how AI models behave, and the platform that publishes it.

AIGROW is the company I founded. It builds AI systems for businesses and publishes open research on how AI models behave: the AI Visibility Index, the AIGROW Score and the studies of AIGROW Labs. I built the platform that produces and publishes all three.

Status
In production
Role
Founder and AI engineer
Year
2025
Live at
aigrow.gg
AIGROW: aigrow.gg, filmed as it runs

What it is

AIGROW builds AI systems for businesses and publishes open research on how AI models behave. I built the platform end to end: the AI Visibility Index, a monthly open dataset of which businesses ChatGPT, Claude, Gemini and Grok recommend; the AIGROW Score, a 0 to 100 measure with a 95% interval that Claude and ChatGPT can query through an MCP connector; and AIGROW Labs, which publishes studies and benchmarks. A venture of PeakCore LLC.

Visit aigrow.gg

Built with

  • React 19
  • Vite
  • Express
  • PostgreSQL
  • OAuth 2.0 with PKCE
  • MCP

What you can check

  • AI Visibility Index, October 2026: 80,059 answers, 20 countries, 11 languages, CC BY 4.0, with a DOI
  • AIGROW Labs: 36 models on 20 office tasks; 6,937 domains tested for email spoofing
  • In NVIDIA Inception, Microsoft for Startups, the ElevenLabs grant and the Runware Startup Program

The AI Visibility Index

The AI Visibility Index is a monthly open dataset of which businesses ChatGPT, Claude, Gemini and Grok recommend when people ask them. The October 2026 edition holds 80,059 answers across 44 business types, 108 places, 20 countries and 11 languages. It is released under CC BY 4.0 with a DOI and mirrored on Hugging Face and Kaggle.

People now ask an assistant where to eat, which dentist to book and which plumber to call. What the assistant answers decides who gets the customer, and until someone records the answers nobody can say how those recommendations are distributed or how they change from one month to the next.

The system behind the dataset asks the same questions in 11 languages across 108 places, collects the answers from four assistants, and publishes them per city and business type together with the method and the neutrality rules.

The AIGROW Score, and why it has an interval

The AIGROW Score turns those answers into a measure from 0 to 100 of how AI models see one business, and it is always published with a 95% interval. An assistant does not give the same answer twice, so a single number without its uncertainty would claim a precision the measurement does not have.

The interval is what makes the score usable. Two businesses three points apart with overlapping intervals are not meaningfully ranked, and the score shows that where a bare number would hide it.

Claude and ChatGPT can query the score through an MCP connector, so an assistant can look up the measured figure for a business where it would otherwise guess.

MCP server development, as a service

AIGROW Labs: studies and a benchmark

AIGROW Labs publishes single studies with their data and method. Two so far: an email spoofing test of 6,937 local business domains, of which 80.9% could be impersonated, and a benchmark of 36 AI models on 20 office tasks, with a price index of 46 models beside it.

The benchmark asks a practical question: which office jobs can a model be left alone with today? The pilot edition ran 720 graded runs on 23 September 2026 and 515 passed, 71.5% overall. The spread says more than the mean. Models routed a request to the right team in 69 of 72 runs, and replied to a bad review without revealing private details in 8 of 36.

The spoofing study tested the email authentication records of local business domains, and it is published with a free checker so an owner can test their own domain.

All published research

The platform under it

The platform is React 19 on Vite with an Express API on PostgreSQL, prerendered to static HTML and served in 11 languages. Sign in is OAuth 2.0 with PKCE, and the site carries a public Trust Center. I built it end to end, with AI agents working in parallel on separate parts.

Prerendering is a requirement here. A research site that renders in the browser is an empty document to a crawler that does not run JavaScript, and to the retrieval systems behind AI assistants, which are the readers this research is about.

AIGROW is part of NVIDIA Inception, Microsoft for Startups and the Runware Startup Program, and has received a grant from ElevenLabs. It is a venture of PeakCore LLC.

Questions

What is the AI Visibility Index?
A monthly open dataset of which businesses ChatGPT, Claude, Gemini and Grok recommend when people ask them. The October 2026 edition holds 80,059 answers across 44 business types, 108 places, 20 countries and 11 languages.
Can the data be reused?
Yes. The Index is released under CC BY 4.0, so it can be reused and cited with attribution, and each edition has its own DOI. The October 2026 edition is DOI 10.5281/zenodo.23188442.
Why does the AIGROW Score come with an interval?
Because an assistant does not give the same answer twice. The score is a measure from 0 to 100 built from many answers, and the 95% interval shows how much it could move. Two scores with overlapping intervals should not be read as a ranking.
How do Claude and ChatGPT query the score?
Through the AIGROW MCP connector. An MCP server exposes the score as a tool, so the assistant asks for a business and gets back the measured figure with its interval.

Facts on this page verified 2026-11-10

Tell me what you need built.

Write what the system has to do, what exists already and when it has to be live. I read every brief myself and answer in writing, with a scope and a price for each milestone.

Or write to contact@guglielmo.io

The human check appears here.

No calls and no sales sequence. One written answer from me.