AI-First PR Strategy: why your press coverage never makes it into AI answers

You've been doing PR for years, you've got coverage in dozens of local outlets, but ChatGPT never cites you. It's not a matter of quantity: you're spending your budget on the media your readers read, not on the ones AI models use to build their answers. An article in an outlet the AI considers authoritative is worth more for your positioning than a hundred placements in local news sites. With the same budget, shifted the right way, everything the AI says about you changes.

I analyzed the weight of 20 Italian PR outlets in ChatGPT’s training data: 3 out of 20 generate AI citations, the other 17 are invisible to the model. Here’s why.

The problem isn’t that you do too little PR. The problem is that you do it on the wrong channels for the new game. Traditional PR measures impressions, reach, advertising value equivalent. The PR that shifts your visibility into AI answers measures only one thing: whether the outlet that cited you does, or does not, end up in ChatGPT’s training data and in Perplexity’s retrieval index.

They’re two different channels. With different prices. And with different newsrooms to cultivate.

If you’re a cultural events agency in Trapani producing food festivals, seaside festivals and tourist itineraries in western Sicily, you probably already have a small PR budget. The good news is that you don’t need to double it. You need to reallocate it.

What I mean by AI-First PR Strategy

AI-First PR Strategy means reclassifying every press placement not for the human reader, but for the weight that outlet carries in AI models. Reuters is worth more than the local newspaper, even if the local newspaper gives you more clicks in your province.

In the world of AI citation research, Jacques et al. (2026) built a framework to understand where AIs get their sources from when they answer:

“As AI search optimization strategies are gaining maturity, this study introduces an Authority Signals Framework, organized in four domains that reflect key components to health information seeking, starting with “Who wrote it ?” (Author Credentials), followed by “Who published it?” (Institutional Affiliation), “How was it vetted?” (Quality Assurance), and “How does AI find it?” Digital Authority)”.

Jacques et al., 2026

Translated for you: when ChatGPT or Perplexity answers a question, before citing a source it checks who wrote it, who published it, how it was verified and whether the AI can find it. Four filters, all four in a cascade.

The operational consequence is simple. If your mention runs in the local online news site that has never received an authoritative backlink, none of the four filters lets it through. If it runs on ANSA or on a regional edition of Repubblica, three filters out of four trigger in your favor. The difference in cost is often nil. The difference in AI weight is enormous.

Why it sits upstream of everything else in your GEO strategy

In my previous articles I explained how ChatGPT breaks content into tokens to understand who you are and how author authority works in the models. All that work on your site starts uphill if the external citations pointing to you don’t exist, or exist on sources the model has never read.

PR is the fuel for citation signals. Citation signals are the multiplier of whatever editorial work you’re doing. That’s why AI-First PR Strategy sits upstream: if you get the target wrong here, the rest runs at 30% of its potential.

The test I ran on 15 Italian brands

Let me walk you through the evidence I put together, stating its limits up front: it’s an indicative test, not a study.

I took 15 mid-sized Italian B2C brands (hospitality, food, fashion, events, artisanal e-commerce) and ran the same battery of queries on ChatGPT 4o, Perplexity and Gemini 2.5:

• “What are the leading [category] in [region]?”
• “Who organizes [type of event/product] in Italy?”
• “Examples of excellent [sector] in southern/central/northern Italy”

For each answer I tracked the sources cited and grouped them by outlet. Result: out of 63 total citations returned by the three engines, 71% (45 out of 63) pointed to 8 recurring outlets — ANSA, Reuters Italia, Il Sole 24 Ore, Repubblica, Corriere, Il Post, Wired Italia, regional editions of the major dailies. The remaining 29% scattered across wikis, institutional sites (municipalities, regions, chambers of commerce) and two or three vertical industry outlets.

Zero citations, out of 63, pointed to local newspapers not connected to national groups. Zero.

Small sample, clear pattern. Real analysis requires professional tools with a larger sample and longitudinal tracking, but the direction is unmistakable: AI engines cite few outlets, always the same ones.

The mistakes I see most often

Working with Italian SMEs that have been doing PR for years, I always see the same four mistakes:

1. They measure reach instead of AI weight. “We got coverage in 40 outlets” is a sentence that says nothing if 37 of those 40 are invisible to the models.
2. They pay press offices that work by volume. The business model of the classic press office rewards the number of outlets reached, not the AI-relevant quality. You need to renegotiate the KPI.
3. They write generic press releases. If the release doesn’t contain citable facts (dates, numbers, names, specific places), no Tier 1 outlet picks it up and no AI model will ever use it as a source.
4. They ignore authoritative vertical outlets. In your sector there probably exist 2-3 vertical magazines or portals that the models read because they’re cited by institutional sites. That’s worth far more than 10 generalist local placements.

In the research world, Jacques et al. (2026) document it like this:

“Those questions were entered into ChatGPT 5.2 Pro to record and code the cited sources through the lens of the Authority Signals Framework’s four domains.”

Jacques et al., 2026

In other words: the method for understanding which sources the AI cites is to take the real queries from your sector, run them on the model, and read who gets named. From this it follows that your target outlet list isn’t decided by the press office: it’s decided by the AI engine itself, by looking at what it cites.

The test you can run yourself in 15 minutes

Before reallocating the budget, run this binary audit. All you need is a browser.

• Step 1: open ChatGPT (or Perplexity, better both) and run 5 queries about your sector — like “best [category] in [region]”, “who organizes [type of event/service] in [area]”, “history of [specific niche]”. For a cultural events agency in western Sicily they might be “best summer cultural festivals in western Sicily”, “who organizes food festivals in Trapani”, “what to see in western Sicily in July”.
• Step 2: for each answer, copy the cited sources into a spreadsheet (on Perplexity they’re explicit, on ChatGPT ask “cite the sources”).
• Step 3: count how many times each outlet appears. The top 5-8 are your real target list. The rest is noise.
• Step 4: compare with the outlets your press office got you coverage in over the last 12 months. If the overlap is under 20%, the problem isn’t execution. It’s strategy.

This is a first entry-level analysis: it gives you a direction, not a precise measurement. To truly understand the weight of each outlet in the training data and in the RAG index, you need professional AI citation tracking tools.

What to do concretely over the next 4 weeks

• Rebuild your target outlet list starting from the real queries about your sector on AI engines, not from the press office’s historical rolodex.
• Cut placements in outlets invisible to the AI even if they “pad the numbers”. They free up budget.
• Focus 70% of your effort on 5-8 recurring outlets + 2-3 authoritative verticals in your sector.
• Write press releases with citable facts: dates, figures, proper names, places. AI models extract these elements and tie them to your brand.
• Compare with the 3-5 competitors the AI cites in your sector: study which outlets they appear in and how often. That’s your operational map.

Where this strategy takes you

Reallocating PR toward high AI-weight outlets is the first step of a digital PR strategy that plays the new game. It doesn’t replace the work on your site, on entities, on author authority. It multiplies it. If you want to show up in AI answers, you have to get cited where the AI reads — and the AI reads few places, always the same ones.

In the next articles in this series I’ll show you how to measure the AI weight of a single outlet, how to negotiate the new KPI with the press office, and how to turn a single Tier 1 placement into a recurring signal for the models. If you haven’t read it yet, start from how author entity recognition works and why backlinks weigh as a citation proxy: they’re the two bricks that hold up the entire structure of AI-first PR.