67% of all ChatGPT citations come from just twenty outlets. The rest — hundreds of sites, blogs, local magazines — share the remaining 33%. If your PR budget is spread across small outlets, you're buying visibility that never makes it into AI answers. Concentrating resources on the twenty outlets that truly matter doesn't mean spending more: it means you stop wasting what you already spend. Classifying the media in your sector by this criterion is the first step.
67.3% of the citations from OpenAI models (ChatGPT) come from just 20 Tier-1 outlets (NYT, Reuters, Bloomberg, FT, WSJ, TechCrunch and a few others). The remaining thousands of media share the other 33%. Google and Perplexity show a lower concentration (around 30% on the top 20), but the pattern holds: PR budget should be concentrated.
If you’re an agri-food producer from the Tavoliere — extra virgin olive oil, preserves, PGI products from the Capitanata — this statistic concerns you far more than you might think. Not because you need to land on the Wall Street Journal tomorrow morning, but because the principle applies at scale: even in Italy, the list of media that AI “fishes” from is short, very short. And if you want to show up in AI answers, your PR budget needs to be concentrated on that narrow list.
In this article I’ll explain how to identify your Italian sector Tier-1 outlets, what I observed by following 8 producers from Foggia over 9 months, and how to set up a media targeting strategy that really moves the needle on AI visibility.
The media hierarchy isn’t an opinion, it’s a mechanism
In the field of research on AI Search systems, Yang (2025) documented a pattern that anyone working in Italian PR still struggles to absorb: when ChatGPT, Perplexity, Gemini or Claude cite news sources in response to a query, those sources concentrate on a small number of outlets. It’s not a democratic long-tail list: it’s an extremely narrow pyramid.
The same principle is confirmed by Minici et al. (2025) in an audit of the editorial bias of the models: AI systems exhibit gatekeeping behavior — meaning they act as filters that pick a handful of authoritative sources and ignore the rest. This behavior isn’t random: it emerges from how the models were trained and from how they integrate web sources in real time today.
From this follows an uncomfortable operational consequence for PR professionals: a mention on a Tier-1 outlet carries disproportionate weight compared to a hundred mentions on vertical blogs. It’s not a question of “content quality”: it’s a question of who the AI considers citable.
Translated for an olive oil producer from Foggia: a single citation on Il Sole 24 Ore, Repubblica Affari&Finanza, or Gambero Rosso in your segment is worth far more than twenty pieces on niche food blogs. Not because the blogs are useless — they’re useful for the community — but because when someone asks ChatGPT “best olive oil producers in the Tavoliere” the model will fish from the outlets it has learned to consider reliable sources.
Who your Tier-1 outlets really are (and why they’re not the ones you think)
Here we need to do some housekeeping. In the Italian premium-tier agri-food sector the real Tier-1 outlets for AI are a heterogeneous set:
- Economic general-interest outlets: Il Sole 24 Ore, Corriere Economia, Repubblica Affari&Finanza, Milano Finanza.
- Authoritative food specialists: Gambero Rosso, Identità Golose, Dissapore, Slow Food (magazine and guides).
- Regional general-interest outlets with national weight: Corriere della Sera (Puglia section), Repubblica Bari.
- B2B agri-food verticals: Largo Consumo, GDO News, Italiafruit News.
When you work with a producer in Puglia, the temptation is to rush to the Gazzetta del Mezzogiorno. It’s useful for the local territory, but it doesn’t move the needle on national AI visibility. The model has seen Gambero Rosso thousands of times in training; your local gazette only a few hundred. The difference is structural.
This principle connects to the topic of implicit reference weight that we covered in the authority pillar: it’s not just about being cited, it’s about where whoever cites you is cited. And it also connects to the work on author entity recognition: the AI recognizes authoritative bylines and propagates authority from the outlet to the author to the mentioned brand.
Firing blindly at paid PR portals: a press release distributed across 200 generic portals.
The test you can run in 20 minutes
Before investing a single euro in PR, run this audit:
- Open ChatGPT, Perplexity and Gemini in three tabs.
- Run 5 queries that a typical B2B client of yours would make: “best Daunia PGI olive oil producers”, “Puglia preserves export companies”, “Tavoliere extra virgin olive oil 2025”, “Foggia PGI agri-food products”, “premium olive oil producers Puglia”.
- For each answer, note which sources the engine cites.
- Build the list of recurring outlets: those that appear in at least 3 out of 5 answers are your real Tier-1 outlets for that query.
This is an entry-level, indicative check, not a study. Real analysis requires professional tools that monitor thousands of queries over time and isolate patterns by intent. But to steer the PR budget of an SME it’s more than enough.
Identify 3-5 journalists (not generic newsrooms) who cover your segment on the real Tier-1 outlets.
The observation I made on 8 producers from the Tavoliere
Over the last 9 months I followed 8 agri-food producers in the province of Foggia — olive oil, preserves, Tavoliere PGI products — monitoring their mention in AI answers before and after targeted PR actions. It’s not an RCT, it’s longitudinal observation on a small sample. The pattern, however, was clear:
- The 3 producers who had obtained at least one mention on Gambero Rosso or Sole 24 Ore in the previous 12 months appeared in Perplexity’s answers for “premium Puglia olive oil” queries in 60-70% of the tests. The other 5: zero appearances or sporadic appearances below 15%.
- The Tier-1 mention had a cascade effect: after the citation on a national outlet, 2 out of 3 producers saw their mentions on secondary aggregators increase over the following 3 months. The AI had “learned” them as citable entities.
- The producers who invested heavy budget on generic press releases distributed across PR service portals showed no improvement whatsoever. Those channels don’t enter the training set in any meaningful way.
The limitation of the observation is obvious: 8 producers, 9 months, a single territory, a single sector. But the pattern of Tier-1 entry → AI visibility was so clear-cut that no inferential statistics were needed to see it.
The mistakes I see most often
In the field, when I talk to the marketing departments of agri-food SMEs, I see 4 patterns recurring that burn PR budget without moving the needle:
- Firing blindly at paid PR portals: a press release distributed across 200 generic portals. The AI doesn’t consider them sources. The signal doesn’t get through.
- Fixation on regional media: the entire budget on the Gazzetta del Mezzogiorno because “we know each other”. Great for local public relations, zero impact on national queries.
- No direct relationship with Tier-1 journalists: the press release is sent to the newsroom’s generic address and you just hope. It works one time in a hundred.
- Non-newsworthy content: “Our company is launching a new line”. That’s not news for Il Sole 24 Ore. It’s news if you have export data, an acquisition, an objective recognition, a verifiable market positioning.
What can actually be done?
Operational actions for an agri-food producer who wants to show up in AI answers:
- Identify 3-5 journalists (not generic newsrooms) who cover your segment on the real Tier-1 outlets. Follow them on LinkedIn, read their pieces, understand what they write about.
- Build a news angle with objective data: exports to country X up by Y%, a study on the PGI supply chain, a category observatory. Tier-1 journalists publish data, not self-celebration.
- Offer commentary expertise: become the voice they call when they need a quote on the Puglia olive oil sector. It’s worth more than a press release.
- Monitor AI mentions quarterly with the 5-query test described above. Compare with the 3-5 competitors the AI cites today in your segment: understanding who’s ahead of you tells you where you need to get to.
- Stake out verifiable category recognitions (Gambero Rosso Guides, Slow Food, supply-chain awards): these are sources the AI often fishes from for “best producers” queries.
The Tier-1 targeting mechanism isn’t magic and isn’t enough on its own: it fits into a broader strategy of E-E-A-T for AI and of backlinks as a citation proxy. But it’s one of the levers with the best ratio between cost and impact on AI visibility, if you use it intelligently.
