Coverage, announcements and posts about how AI engines are reshaping brand discovery. Each entry summarises the original; follow the link for the full piece.
Engage interviews Luca Dell'Anna on the shift from lists of links to the synthesised answers of ChatGPT, Gemini and Claude. The core point: what matters is no longer where you appear, but how you are described. Also covered: dead links in AI citations, the calibrated question sets used to measure sentiment and resilience, and the Y Combinator Fall 2026 application.
Engage covers DelveDeep as a startup dedicated to analysing brand reputation inside AI models, with a seed round opened a month after launch and an application to Y Combinator's Fall 2026 batch.
Apps need to be recognised and recommended by ChatGPT, Claude, Gemini and Perplexity too. The partnership measures visibility not in the abstract but against the user's actual question and stage of the buying journey, weighing the authority and sentiment of the sources behind each answer.
Two findings from DelveDeep's analyses: many URLs the models cite lead to 404s or removed pages, including on the sites of the companies being analysed; and social platforms almost never appear among the sources, despite years of content. The argument: in the GEO era, whoever was already authoritative when the model stopped learning wins.
Luca Dell'Anna thanks Antonio Capone and Lorenzo Paletti for the episode on DelveDeep, the NextMind solution that analyses and manages the visibility of brands and retail locations inside ChatGPT, Claude, Gemini and Perplexity.
A TuttoConnesso episode with Antonio Capone and Luca Dell'Anna: for years big brands optimised their sites to reach the top of Google, and that strategy now changes with AI-mediated search. Available on Spotify, Apple Podcasts, Audible and YouTube.
Engage covers the launch: the platform analyses thousands of answers generated by the major LLMs to measure brand presence, positioning and associated attributes, with proprietary KPIs such as AI Visibility Score, Recommendation Strength and Relative Competitive Weight.
An analysis of 6,840 questions put to ChatGPT, Gemini, Perplexity and Claude about 15 Italian utility brands. The models favour stable, verifiable text sources — articles, press releases, institutional pages, Wikipedia — which makes an investment concentrated purely on social content insufficient.
The models cite pages that no longer exist: 29 broken links for Lavazza, 11 for Nespresso, 10 for Illy. The problem is not only the user who looks for the sustainability report and hits a 404 — these are the very pages the model relied on to describe the brand in the first place.
The architecture behind DelveDeep Private AI: documents stay where they are (NAS, SharePoint, Drive) and only a semantic map is built. The vector index lives on company infrastructure and holds embeddings rather than the documents themselves; answer generation happens inside the perimeter too.
Starting from research by the Politecnico di Milano Digital Innovation Observatory: many companies hold sensitive documents with no way to interrogate them, so people end up pasting them into ChatGPT or Claude. DelveDeep Private AI allows natural-language queries while staying on-premise or on private cloud.
The five NextMind lines: DelveDeep LLM for visibility inside the models, Panel for testing messages against synthetic panels, Private AI for sensitive data, Gateway DB for querying legacy databases in natural language, and Lab for bespoke work. The underlying argument: this is not about “adopting AI”, but about finding where it becomes a critical interface.
An analysis of 3,729 links cited by the models about Italy's top eight utilities (A2A, ACEA, Edison, Enel, Hera, Iren, Plenitude, Sorgenia). Of the 1,789 broken links, a third carried positive sentiment now lost, 44.6% were neutral, and 22.4% negative — criticism left unattended.
Around 150 questions across five models (ChatGPT, Gemini, Claude, Grok, DeepSeek) show what happens after a rebrand and a domain migration: the models keep citing the previous domain, now unreachable, because they do not update in real time. A problem of source continuity before it is one of communication.
The announcement of the platform built with the NextMind team to measure how AI systems present brands in their answers. At its centre is the LightHouse KPI: not only whether the brand is cited when the question is directly about it, but also in category and comparison queries — the competitive space where awareness surfaces without the brand being the explicit subject.
