AI Search Visibility: Professional Firms at Risk

Why ChatGPT does not cite your website — and who it cites instead

The direct answer is this: ChatGPT favours original sources. According to an Ahrefs analysis of millions of responses generated between March and September, PubMed rose from 64th to 4th place among the most cited domains in ChatGPT's US responses. Over the same period, WebMD — one of the most visited medical information sites in the world — fell from 18th to 79th place. Britannica, Investopedia and other general-purpose explanation sites dropped out of the top 200.

The pattern is consistent: institutional sources (FDA, CFPB, FTC), original research sites and those that carry out direct product testing are rising. Sites that summarise, aggregate or explain what others have already said are falling. Ahrefs puts it plainly: ChatGPT "appears to be prioritising original sources over sites that summarise them".

For a firm of accountants, lawyers or employment consultants, the question becomes concrete: does your website produce content that only you can produce? Regulations interpreted through your own lens? Operational templates developed in-house? Positions on specific cases? If the site is a generic showcase, AI search engines have no reason to cite you. If it contains original analysis, genuine answers to your clients' real questions and well-argued positions, you have a real chance.

What GEO means for an Italian professional firm

GEO — Generative Engine Optimization — is the discipline that studies how to structure a website's content so that AI search engines can find it, understand it and cite it in their responses. It does not replace traditional SEO: it sits alongside it, with partially different logic.

In classical SEO, the goal is to rank among the top results for a keyword. In GEO, the goal is for the AI engine, when answering a relevant question, to include a citation to your content — or to use it as a source when constructing its response. The difference is significant: in AI search there is no list of ten results to appear in. There is a single answer, and a handful of cited sources.

For an Italian professional firm, GEO translates into concrete choices: writing content that answers specific client questions (not generic ones), structuring pages with verifiable and attributable data, using correct but accessible technical language, and — an aspect often overlooked — being recognisable as an author with verifiable expertise. The professional's name, title and area of specialisation: these authority signals matter to AI models in exactly the same way as they do to a human reader deciding whom to trust.

Google Discover becomes an AI concierge: what changes for your visibility

Google Discover is the recommendation engine that surfaces content on the Chrome mobile home screen and in the Google app, without the user having searched for anything. It reaches more than 500 million active users per day, according to internal figures cited by Google itself in its job listings. For many publishing sites it is a massive source of traffic.

What is happening now, however, is of a different magnitude. A job listing for a Principal UX Lead reveals that Google wants to transform Discover from an article recommendation engine into a "proactive, hyper-personalised concierge experience". CEO Sundar Pichai has outlined the direction: in future, many informational queries will become "agentic search", completing tasks rather than retrieving information.

In parallel, Google is hiring engineers to generate and recommend AI content directly in the Discover feed, with controls on accuracy, freshness and tone. The new Discover will not simply show you an article on a topic you are interested in: it will anticipate what you need, suggest products, services and next steps, and integrate YouTube, Shopping, Chrome and Android into a single experience.

For a professional firm or SME, this means that "passive" traffic — the kind that arrived because your article appeared in Discover — will change in nature. The system will select content capable of helping the user take an action, not just read and move on. Those with a website structured around clear calls to action, practical answers and content designed to convert will have a measurable advantage over those publishing generic text.

Is traditional SEO still worth doing in 2026?

Yes — but with an important qualification. Traditional SEO is not dead, but it is no longer enough on its own. Ahrefs data show that only a minority of ChatGPT responses in free mode include a web search: AI engines still crawl the web, but they select differently compared with classical Google.

SEO remains relevant for intentional search — when a user opens Google and types "accountant in Milan specialising in start-ups". Organic ranking still matters there, and probably will for years to come. But the share of queries passing through AI is growing, and for those queries what counts is the structure of the content, its originality and the verifiability of the author.

The most robust strategy is not "traditional SEO or GEO": it is understanding what questions your potential clients are asking, answering those questions with original, well-structured content, and measuring where that visibility translates into new enquiries. The volume of content published matters less than the quality and uniqueness of what you write. A firm that publishes an in-depth analysis of a tax circular — with its own operational interpretation — is worth more than ten generic posts about "the latest tax news".

As we analysed in our article on AI Overviews and lost clicks, classic organic traffic is already under pressure. Acting now, before the transition is complete, is far more cost-effective than scrambling to catch up in two years' time.

Is your website ready for AI search engines? Check these points

• ✓
  Verifiable authorEvery piece of content has a named author, with their title and area of expertise declared.
• ✓
  Original contentKey pages contain analysis, positions or data that only your firm can produce.
• ✓
  Direct answersEvery page answers a specific question clearly within the first 100 characters.
• ✕
  Structured data (schema markup)The site uses JSON-LD markup for the organisation, the professional and the services offered.
• ✕
  Operational calls to actionEvery informational piece of content leads to a concrete action: booking, contact, download.
• ✕
  Search Console generative AI report monitoredYou regularly check how many impressions you receive in Google Discover's AI results.

How to adapt: operational steps for a professional firm

This is not about rewriting your entire website from scratch. It is about intervening with clear priorities and measuring the results. Here is a four-phase approach.

The first step is a diagnostic review: check the generative AI features report for Discover in Google Search Console. Verify which of your pages are already being cited by ChatGPT by searching your main questions manually. Identify which sources the AI engine cites in your place for the topics in which you specialise.

The second step is selecting high-value content: choose three to five thematic areas in which your firm has a genuine and distinctive competence. For each one, write a page that answers a specific client question — not an abstract keyword. Include your own interpretation, not just a summary of the regulation.

The third step concerns technical structure: add structured data markup (schema.org) to identify the organisation, the professionals and the services. Make sure every page has a declared author and a visible last-updated date. These signals help AI systems assess the reliability of the source.

The fourth step is to measure and iterate: define a simple indicator — how often, when you search your key questions on ChatGPT or Perplexity, your site is cited. Track the figure every month, after every model update. Adjust content based on what is not working. Without measurement, every effort is wasted.

Strategy matters more than volume: the principle that never changes

There is a widespread temptation: produce more content, faster, perhaps with AI. It is the wrong answer to the right problem. Ahrefs data show that sites which aggregate and summarise — even at high volume — are losing ground. What is rising are sources that produce something irreproducible.

For a professional firm, this is actually good news. Your competitive advantage — specialist expertise, experience with real cases, the professional responsibility that distinguishes you from a chatbot — is precisely what AI engines are looking for. You do not need to become a publishing agency. You need to structure what you already know more effectively.

The distinction we draw at Ovia is between tool and strategy: using ChatGPT to write generic content will not make you visible to AI search engines. Deciding what to write, why, for whom, and how to measure whether it is working — that is the strategic work that makes the difference. AI is the tool for executing it more efficiently, not a substitute for thinking.

A website designed to be found and cited by AI search engines is not a more attractive or larger site: it is a more precise one. Every page has a purpose, answers a question, names its author and facilitates an action. That is the difference between a showcase and a working tool.