You have a page on your site full of enthusiastic testimonials. To AI it's worth nothing: you're the one talking about yourself, and the models know it. What matters is what people say about you on sites you don't control — and in particular, whether your clients cite you as a supplier on their own sites. Across twelve brands analyzed in different sectors, the ones that showed up in AI answers had a network of references distributed across third-party sites; the others didn't. Building that network starting from the clients you already have is simpler than it sounds.
The testimonials on your site are useless to AI. The testimonials on your clients’ sites, the ones that mention you by name, are not — and they’re completely different.
They are two objects we call by the same word but which, for ChatGPT, Claude, Perplexity and Gemini, carry opposite weight. The first is self-referential: you say your clients are happy. The second is a distributed citation: a third party, in its own context, names you as a reliable supplier or partner. The first only confirms that you know how to write a “Testimonials” page. The second is verifiable social proof, pulled from a domain you don’t control.
In my articles on Digital PR I’ve already told you that AI builds trust by reading what the web says ABOUT you, not what you say about yourself. The testimonial network is the most operational application of this principle.
How an AI model reads a testimonial
When Perplexity answers “best ornamental plant nurseries in Tuscany for landscaping projects,” it doesn’t scroll through the nurseries’ “About us” pages. It scrolls through industry articles, case studies published by landscape architecture firms, garden designer blogs, supplier lists cited by specialized magazines.
In that content the nursery isn’t talking about itself: someone else is talking about it. And that someone else has its own weight (a domain, an author, a context). The nursery, within that fragment, isn’t the subject — it’s the cited entity.
There’s no paper that directly measures “the weight of a third-party testimonial vs a self-celebratory one.” But the mechanism can be inferred from what I showed you in the previous articles on the backlink as a citation proxy and on the implicit reference weight: the system assigns more trust to mentions embedded in third-party contexts than to first-party statements. From this it follows that a testimonial published on a client’s site — where YOU are the cited supplier, not the author — is AI visibility infrastructure, not a showcase.
Why it comes upstream of the rest of the PR work
I’m guiding you through the citation signals that AI engines read to decide who is authoritative in a sector. Guest posts, editorial mentions, vertical reviews. The testimonial network comes upstream of all this because it’s the signal that is most easy to activate and most difficult to fake.
Easy: you already have satisfied clients, you just need to ask the right thing in the right place. Difficult to fake: no agency can invent testimonials distributed across real third-party sites with history, authors, and consistent backlinks.
That’s the reason why, when a brand asks me “where do I start with Digital PR for AI search,” I start here. Before thinking about industry media, before mapping journalist lists, before building thought leadership on LinkedIn.
Thinking that a client logo on the homepage is enough.
The test you can run in 20 minutes
Take your brand name and try it across three environments.
First: on Google, search “brand name” site:* excluding your own domain. The operational command is `”brand name” -site:yourdomain.com`. This shows you where the web names you without you having paid, optimized, or controlled the content.
Second: ask Perplexity “who are the suppliers of X for Y in area Z” replacing X, Y, Z with your sector and region. Look at the sources Perplexity opens in the sidebar. How many of those sources are your clients’ sites, professional firms that work with you, partners?
Third: open Google Trends and compare your brand’s search volume with 2-3 direct competitors. If your volume is similar but AI mentions your name less, the problem isn’t awareness — it’s a thin testimonial network.
The binary threshold is this: if in the first 30 results of `”brand name” -site:yourdomain.com` you find fewer than 5 pages where someone cites you as a supplier/partner in a real work context, your testimonial network is below the minimum threshold to emerge in AI answers.
You provide the semantic draft: the brand name written exactly as you want it indexed, the context of the work done, the measurable result.
The test I ran: three nurseries in Tuscany, six months of observation
To understand how much the distributed testimonial really weighs, I tracked for six months three nurseries of ornamental plants and monumental trees in the province of Pistoia — a district that supplies landscaping projects all across Europe. Three companies comparable in revenue, catalog, and seniority.
The first nursery had a testimonial page on its own site with 12 citations, none signed with a link to the client. The second had a similar testimonial page plus 3 case studies published by landscape architecture firms that named it as a supplier of monumental trees for private parks. The third had no testimonial page on its own site, but appeared cited by name on 14 external pages: landscape designer portfolios, articles in vertical magazines on fine gardening, project details on architects’ sites.
I queried ChatGPT, Claude, Perplexity and Gemini with 8 queries like “Tuscan nurseries for monumental trees,” “suppliers of ornamental plants for high-end landscaping projects in Italy,” “where to buy centuries-old trees for private parks.”
Across 32 total answers (4 engines × 8 queries), the third nursery was cited by name 21 times. The second 9 times. The first 2 times. The first had the most polished site of the three. The third had the oldest and least optimized.
An indicative test, not a study: small sample, specific sector, limited period. But the pattern is clear enough to call into question where you put your communication energy.
The mistakes I see most often
Anonymous testimonials like “Mario R., entrepreneur.” Without a first name, last name, company, or verifiable link, the AI model treats them as noise. To Claude or Perplexity they are strings with no recognizable entity. In the previous articles I told you about named entity recognition: a testimonial with no nameable entities simply doesn’t exist in the graph.
Testimonials only on Google Business Profile. Extremely useful for local SEO, marginal for general AI search. Perplexity and ChatGPT rarely pull from there for complex B2B queries. A pastry shop in Pistoia that supplies Tuscan hotels wins more if it’s cited in vertical food guides than if it has 200 five-star Google reviews.
Asking the client “would you write me a review?” without specifying where. The client replies with what they know: Google, Trustpilot, Facebook. They miss the chance for a signed comment under an industry article, a case study on a vertical outlet, a mention in a professional directory.
Thinking that a client logo on the homepage is enough. A logo is not a testimonial, it’s not a semantic citation. The AI model needs text associated with your name, in context. A logo is an image, and images in homepage layouts weigh zero in building the authority graph.
What you can do concretely
The operational list is short but requires discipline.
- Identify the 10 most satisfied clients of the last 18 months. Not the biggest — the happiest.
- Map where THESE clients publish content: company blogs, case studies, industry newsletters, panel appearances, articles on vertical outlets.
- Ask them not for “a review on Google” but for: a paragraph in their next case study naming you as a supplier, a comment under an industry article where they describe the project citing you, a mention in their client portfolio.
- You provide the semantic draft: the brand name written exactly as you want it indexed, the context of the work done, the measurable result.
- Repeat every quarter with 3-4 new clients. The goal is 15-20 citations distributed across 15-20 different domains in 12 months.
It’s not a magic factor. Without a site with decent E-E-A-T and clean author entity recognition, a solid testimonial network remains a blunt weapon. But with those two things in order, this is the multiplier that makes the difference between being the brand AI names first and the brand AI never names.
In the upcoming articles…
The testimonial citation network is the first leg of the Digital PR table for visibility in AI answers. The other two legs — industry media relations and strategic guest posting — we’ll see in the next articles: guest post strategy for citation building and trade media coverage as an authority signal.
If you’re starting from scratch with Digital PR, the testimonial network is where to put your first 20 hours of work. It’s where the ratio between effort and visibility in AI answers is highest, and where results show up within 90 days, not within 12 months.
