4 Million AI Citations Exposed: Press Releases Are Invisible to AI Search

Press release distribution services have been selling a new dream: AI visibility. ACCESS Newswire offers an "AI Visibility Checklist" for press releases. eReleases published a guide positioning press releases as tools for AI search visibility. Business Wire has written about optimizing releases for answer engine discovery.

It sounds compelling. After all, press releases contain structured data, clear facts, and brand messaging. Shouldn't AI models love them?

A new study from BuzzStream, analyzing 4 million AI citations across ChatGPT, Google AI Mode, Google AI Overviews, and Google Gemini, says no. Not even close.

BuzzStream used XOFU, a citation monitoring tool from Citation Labs, to track where AI platforms actually pull their sources. They ran 3,600 prompts across 10 industries and cataloged what gets cited — and what doesn't.

The findings are stark:

• Syndicated press releases barely register in AI-generated answers across all four platforms tested
• Editorial content — original reporting, analysis, and expert commentary — dominates AI citations
• Owned newsroom content outperformed wire-distributed versions of the same announcements on some platforms
• Earned media coverage carried significantly more weight than any form of syndicated distribution

This isn't a minor gap. Press releases are functionally invisible to the AI answer layer that increasingly mediates how audiences discover brands.

The pattern makes sense when you understand how large language models evaluate sources. AI systems are trained to identify and prioritize content that demonstrates:

1. Editorial independence — Content that reads as an independent assessment carries more authority than content that reads as self-promotion
2. Contextual depth — Press releases are thin by design: they announce, they don't analyze. AI models favor sources that explain why something matters, not just what happened
3. Source diversity — A press release syndicated across 200 identical copies on newswire sites looks like one source to an AI model, not 200. An editorial piece in TechCrunch that references your announcement looks like independent validation
4. Unique information — AI models prioritize content that adds something new to the conversation. A press release is, by definition, the same text everywhere it appears

This aligns with what we see in Avisible's AI Visibility Audits. When we measure Share of Answer across AI platforms, brands that rely heavily on press releases and syndicated content consistently underperform brands that invest in editorial-grade owned content and earn independent coverage.

For communications teams, this data demands a strategy rethink. Press releases aren't dying — they still serve regulatory, investor relations, and media-pitching functions. But they are not an AI visibility tool, despite what distribution services claim.

Here's the uncomfortable math: if your brand's primary content strategy for announcements is "write a press release, distribute via wire service, and hope AI picks it up," your announcements are functionally invisible in the fastest-growing discovery channel.

Consider the contrast:

| Content Type | AI Citation Performance | Why | |---|---|---| | Syndicated press release | Near zero | Duplicate content, no editorial authority | | Wire service pickup (verbatim) | Very low | Same content, different domain — still thin | | Original editorial coverage | High | Independent voice, contextual analysis | | Owned newsroom with analysis | Moderate-High | Unique perspective, brand-controlled narrative | | Expert commentary in third-party media | High | Source diversity + authority signals |

The brands winning AI visibility aren't abandoning press releases. They're supplementing them with content that AI actually values.

Based on the BuzzStream data and our own AI Visibility Audit findings, here's what a modern communications strategy looks like when AI discoverability is a goal:

Don't just repost press releases on your company blog. Write original analysis around your announcements. If you're launching a product, publish a deep-dive on the problem it solves, with data and customer context. This is the content AI models will cite.

A single in-depth feature in an industry publication is worth more for AI visibility than 500 syndicated press release placements. When pitching media, emphasize the story, not the announcement. Provide data, expert access, and exclusive angles that result in original editorial content.

AI models assess entity authority — whether your domain is a credible source on a given topic. Publishing consistent, expert-level content on your core topics over time builds the kind of authority that earns AI citations. One press release per quarter does not.

AI systems serve answers to questions. Press releases don't answer questions — they make announcements. Every significant brand moment should be accompanied by content that directly answers the questions your audience will ask AI about it:

• "What does [your product] do differently?"
• "How does [your company] compare to [competitor]?"
• "What are the benefits of [your approach]?"

Use tools like XOFU or conduct regular AI Visibility Audits to track which of your content assets actually appear in AI answers. You may find that a three-year-old blog post drives more AI citations than your last 50 press releases combined.

The BuzzStream study reflects a broader shift that Avisible has been tracking: in AI search, content quality is distribution.

Traditional PR operated on a volume model — get the release on as many wires and sites as possible, and visibility follows. AI search inverts this. A single piece of exceptional, original content on a trusted domain outperforms thousands of identical copies.

This doesn't mean distribution doesn't matter. It means the type of distribution matters. Earning a mention in an authoritative editorial piece is distribution. Getting your expert quoted in a relevant industry analysis is distribution. Publishing a genuinely useful guide on your own domain is distribution.

Paying to syndicate the same 400-word announcement across 200 newswire sites is not. At least not for AI.

For marketing and communications leaders evaluating their AI visibility strategy, the key metric isn't "number of press release placements." It's Share of Answer — the percentage of AI-generated responses about your category that mention your brand, cite your content, or recommend your solutions.

If your Share of Answer is low, BuzzStream's data strongly suggests that doubling down on press release distribution won't fix it. Original, authoritative, question-answering content will.