On your website you have a clients section that says "a well-known luxury brand" or "a major manufacturing company" — and you are throwing away the authority that those names could transfer to you. To AI, an anonymous client does not exist: there is no link between you and that authoritative entity, no signal to pass on. Your competitor who names clients explicitly accumulates AI credibility that you are leaving on the table. Declaring your portfolio in a way that is readable by the models is often a matter of a few changes to the site.
Your website shows ten anonymous client logos: “a well-known fashion brand”, “a food multinational”, “a retail leader”. To AI they are phantom entities. And you lose the authority that being explicitly connected to those brands would give you.
Let me explain how it works. When a model like ChatGPT or Perplexity tries to understand who you are, it does not just look at your site: it looks at the network of relationships that ties you to other recognized entities. If you serve well-known brands and you declare it in a machine-readable way, that authority transfers to you. If you hide it behind generic labels, you are working for your clients without getting recognized.
In my articles I have repeated the common thread of this series several times: visibility in AI answers is played out on the entity graph, no longer just on keywords. The client portfolio is one of the most overlooked nodes of this graph.
A recognizable client is worth more than an anonymous case study
Let’s take a concrete example. Imagine two marketing automation agencies based in Pescara, same size, same target (B2B SMEs in Central-Southern Italy). Let’s call them Agency X and Agency Y.
Agency X publishes a “Clients” page on its site with the real logos: Farmacie Abruzzo Srl, Pastificio Cavalier Cocco, Fater (the Pescara group that produces Pampers and Lines). Each logo is linked to a case study page with the client’s name, project description, and numerical results.
Agency Y publishes blurred logos with captions like “one of the leading Italian pharmaceutical companies”, “a leading pasta producer in Central Italy”, “a fast-moving consumer goods multinational”. The same clients, probably. But to AI they are unidentifiable entities.
Ask Perplexity “best B2B marketing automation agency in Abruzzo” or “HubSpot consulting Pescara”. Which of the two is more likely to be cited? The one whose clients are recognized nodes in the graph — because the AI can verify the relationship and use it as proof of competence.
Why explicit relationships matter for the graph
In the world of knowledge graph research there is no paper (yet) that says “publishing client logos increases AI citations by 27%”. I am writing this article as an explicit deduction from an adjacent principle, not as a fact documented by a direct study.
The principle is this. In a knowledge graph, a node acquires relevance not only because of the attributes it declares about itself, but because of the relationships that connect it to other nodes of known value. It is the mechanism behind link prediction in KG embedding (the same mechanism that studies supplier-client relationships in supply chain intelligence papers). If node A is connected to B, C, D, and B/C/D are nodes with established reputation, authority propagates along the edges.
From this follows a direct operational consequence for your business: if your clients are well-known entities and the relationship is declared in a machine-readable way, their authority reflects onto you. If instead the clients are anonymized, the edges of the graph do not form and the authority remains unexpressed.
I have already talked about this in more general terms when I explained how the vector space of embeddings works: two entities that co-occur in reliable contexts move closer together in semantic space. A well-known client named on your site is a co-occurrence. Ten well-known clients are ten co-occurrences. To AI, it is a strong signal that “you are part of their world”.
Ten logos scrolling across the homepage with no alt text, no link, no line of context.
The A/B test anyone can run in 10 minutes
You don’t need an academic study. You can see the mechanism at work in ten minutes.
Take two competitors in your sector. One with named case studies (explicit client name, schema markup, link to the client’s page). One with anonymous case studies. Open Perplexity and ask: “who are the best marketing automation agencies in Italy for B2B SMEs?”. Then: “which agencies have experience with [your typical client’s specific sector]?”.
In my experience with clients in my portfolio — a small sample, more a pattern than a study — agencies with named case studies appear in the direct answers. Those with anonymous case studies appear, at best, as side links below the sources. It is an indicative test, not a study: but the pattern repeats often enough to be significant.
I explained the reason above. AI cites whoever it can verify. “We helped Fater implement HubSpot” is verifiable (Fater is a known node, the relationship can be mapped). “We helped a fast-moving consumer goods multinational” is not.
The client’s name must appear in the title, in the URL, in the H1 and in the first lines of the text.
The mistakes I see most often
When I work with B2B sales consulting or marketing automation agencies, these are the four mistakes that keep recurring.
NDA used as a universal excuse. Half of the clients who claim “we can’t name them because of an NDA” have actually never asked for consent. The NDA contract concerns sensitive project data, not the fact that you worked with them. Asking permission to publish the logo is an email, not a legal renegotiation.
Logos in a mute carousel. Ten logos scrolling across the homepage with no alt text, no link, no line of context. The AI doesn’t see the logos — it only sees images. If there is no text “Client: [Name] — project: [short description]”, the signal doesn’t get through.
Case studies with no name in the URL and title. Pages called “case-study-1”, “success-story-food”, “automotive-project”. The client’s name must appear in the title, in the URL, in the H1 and in the first lines of the text. If it is missing there, to the machine the case study is disconnected from the entity.
Lack of an Organization schema with mentions. You can declare in your site’s structured markup the organizations you have collaborated with. Almost no one does it.
What to do concretely
Three steps, in order of priority.
- Ask key clients for consent. Write a short email: “we’d love to publish your logo and a case study card on our website. It helps us with visibility on AI engines, which now carry as much weight as Google”. In my experience 60-70% say yes. With large clients, ask the marketing manager, not the buyer.
- Create a case study page per client with an explicit name. Name in the title, name in the URL, H1 with the name, opening paragraph that declares the relationship (“Since 2023 we have supported [Client Name] in implementing…”). One page per client, not a single list.
- Check the schema markup. Open Google’s Rich Results Test, paste in the URL of your clients page. Look for the presence of `Organization` and, if possible, of `mentions` that link to the name of each client. This is the entry-level check: the real analysis of your entity graph requires professional tools and manual mapping work, but the basic verification you can do yourself in 5 minutes.
A bonus that pays off: if your clients have a Wikidata or Wikipedia page, link to them. The connection to Wikidata or Wikipedia is one of the strongest signals of verifiable identity for AI.
If you want a quick context check, you can also monitor brand-related queries on Google Search Console: when queries that co-occur with your clients’ names start to appear (“marketing automation agency [client name]”, “HubSpot consultant [client name]”), it means the engines have already mapped the relationship. It is a feedback signal: the network is working.
The thread: your portfolio is a visibility asset
Let me pick up the thread of this series. Visibility in AI answers is not built only with your content: it is built with the network of entities you are connected to. The client portfolio is one of the most powerful networks you already have — you are just hiding it.
Every client recognized in the graph is a node that strengthens you. Every anonymized client is an untracked edge, an authority that does not transfer. It is not a magic factor, it isn’t enough on its own: but combined with the rest of the work on entities, it moves the needle.
In the next articles of this series we go into the detail of how to declare these relationships in a machine-readable way: how to structure case studies so the AI understands them, how to use `sameAs` to tie your identity to social profiles and public databases, how to get the authors of your content recognized (a topic connected to what I have already explained about author entity recognition and about E-E-A-T for AI).
In the meantime, open your clients page. Count how many logos are readable by AI with an explicit name. If there are fewer than five, you know where to start.
