You have two hundred journalist contacts in an Excel file you haven't updated in years — and you don't know how many of them still cite you, on which outlets, and how often. You're investing time in the wrong relationships while you fail to cultivate the journalists who could cite you on the most influential sources. With twenty well-chosen contacts managed the right way you build a steady citation machine — without increasing your budget.
You have 200 journalist contacts in an Excel file. Some have cited you in the last 3 years, others haven’t. Without a CRM structure you’re giving away mentions to people who are no longer active on your topic.
I’m telling you this because it’s the scene I see most often when I work with Italian SMEs that want to show up in AI answers. The file exists, the names are there, but there’s no criterion to distinguish the journalist who will cite you again from the one who will never write to you again. And without that criterion, every press release goes out blind.
Let me explain how to flip the logic: not a database of journalists, but a CRM of editorial relationships, with 20-30 names worked properly, that produce recurring mentions. Because recurring mentions are exactly what ChatGPT, Perplexity and Gemini read as a pattern of authority.
What changes when you move from “list” to “relationship CRM”
In the world of research on authority for AI models, the point that keeps coming back is that a single citation moves nothing, what moves things is the pattern. In the previous articles in this series I told you how AI models treat backlinks as a citation proxy: the single link matters less, what matters more is the recurrence with which different outlets talk about you in the same topical context.
The Journalist Relationship CRM is the operational tool to build that recurrence. It’s not an Excel file with 200 contacts. It’s a short list of 20-30 journalists, where for each one you know:
- The last time they cited you (date, outlet, topic)
- The topics they have actively written about in the last 12 months
- Their preferred channel (email, LinkedIn, WhatsApp)
- What useful thing you gave them in the last interaction
It follows that every contact becomes trackable, and PR work stops being “I sent the press release to everyone” and becomes “I closed 3 citations this quarter on the topic I want to position for AI”.
Why AI models reason on recurrence, not on volume
Large language models learn associations through the frequency with which an entity appears next to a topic in editorial sources. I told you this when I explained how author entity recognition works: the system doesn’t read “single citation yes/no”, it reads the aggregate profile of who mentions you and on what subject.
Translated into practice: if the same journalist at Il Sole 24 Ore cites you 4 times in 18 months always as a source on your topic, the AI builds a strong association. If 40 different journalists cite you once each on unrelated topics, the association stays weak.
The operational consequence is counterintuitive for anyone doing PR the traditional way. A few deep relationships beat many shallow ones. Not as a matter of “human quality”, but because it’s the signal that AI engines read best.
If the press release can’t be reused in their piece, it ends up in the trash.
The case of a PGI food producer from southern Italy
I want to tell you about a concrete case from one of my clients, a PGI food producer from southern Italy. A family business, third generation, 12 employees, revenue under 3 million. When we started, their PR file was an Excel with 180 contacts collected over 6 years: food journalists, bloggers, editors of gastronomic guides, press officers from trade fairs.
Out of those 180 names, the citations in the last 3 years had been 14. Of those 14 articles, only 6 were positioned on the topics we wanted the AI to learn: the PGI designation, the artisanal production method, the short supply chain of their territory.
First intervention: we cut the list down to 22 journalists. The criterion was binary. In the list went anyone who, in the last 18 months, had written at least 2 articles on artisanal products, the Italian food supply chain, southern PGI products or signature confectionery. Everyone else, out.
Second intervention: for each of the 22 we filled in a profile with the last article published on the topic, the preferred channel, and a line on “what useful thing I can offer them in the next 6 months”. No generic press releases: targeted proposals aligned with their editorial line.
Third intervention: rhythm. Once a month, 3-4 proactive contacts with a data point or a quick comment on an industry news story (e.g. reform of the PGI specification, new labeling rules, consumption trends in their segment).
The result after 6 months: 11 citations on Italian outlets, of which 9 positioned exactly on the target topics. When in March we asked Perplexity “best artisanal producers in their PGI segment”, the brand appeared in the aggregated sources. Before, it didn’t appear.
I’m telling you this with the limits of the case. It’s a single client, not a study. The sector is niche, and in a niche 9 target citations weigh more than in a general-interest sector. But the mechanism is replicable.
Once a month pick 3-4 journalists and offer them something useful before they ask for it: a disaggregated internal data point, a comment on an industry news story, access to a production site, an interview with a technical figure.
How to build your editorial CRM in 3 hours
Here’s the operational audit I have clients do in the first week.
Step 1 — Filter the existing list. Open your PR file. For each journalist, go to Google and search “first name last name + your topic + 2025”. If in the last 18 months they haven’t written at least 2 articles within your topical scope, they come off the list. Full stop. This first pass usually cuts 70-80% of contacts.
Step 2 — Relationship profiles. For the 20-30 survivors, fill in a profile with 5 fields: last citation (date, outlet, URL), outlets they write for today, preferred channel, last exchange with you, next possible lever. That’s all. You don’t need a 200-euro-a-month CRM, a well-built Google Sheet is enough.
Step 3 — Proactive calendar. Once a month pick 3-4 journalists and offer them something useful before they ask for it: a disaggregated internal data point, a comment on an industry news story, access to a production site, an interview with a technical figure. The goal is to become the first call on your topic.
This is a first operational reconnaissance. The real analysis of editorial authority requires professional media monitoring tools and a semantic reading of the cited content — it doesn’t fit in a spreadsheet.
The mistakes I see most often
The PR file treated as an address book. 200 contacts, zero relationships. The press release goes to everyone, nobody replies. Better 20 contacts who reply within 48 hours than 200 who ignore you.
No tracking of who cites what and when. Without a “last citation + topic” column you don’t know whether the journalist is still on your subject or has moved on to something else. You’re shooting in the dark.
Slow response times to journalists’ requests. When a journalist writes to you for a quick comment, they have a 3-4 hour window. If you reply the next day, the piece has gone out without you. For your topic they’ll call someone else. Then someone else again. And you’re out of the pattern.
Press releases that talk about the brand instead of the topic. The journalist doesn’t care that you won an internal award. They care about a data point, a trend, a case. If the press release can’t be reused in their piece, it ends up in the trash.
Why this sits upstream of all the AI work
A journalist who regularly calls you as a source is a citation machine that works for you every week, without you having to produce new content. It’s the most powerful multiplier you have for your visibility in AI answers.
I’ll say it again because it’s the thread running through this whole series: AI models don’t read your site in isolation, they read the editorial context that surrounds you. If that context is made of recurring citations from authoritative outlets on the same 2-3 topics, the AI learns to return you as a source. If it’s made of scattered, occasional mentions, you stay invisible.
In the next articles in this series I’ll tell you how to structure press kits for AI, how to handle journalists’ requests in real time and how to measure the AI return of the citations you obtain. If you haven’t read it yet, first go through E-E-A-T for AI and implicit reference weight: they are the theoretical framework on which all the operational PR work rests.
