We run a portfolio of products at Inithouse. One of them, Be Recommended, tracks how AI engines cite and recommend brands across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. We score visibility 0 to 100 and tell you what to fix.
Running this tool across hundreds of brands taught us something we didn't expect: press releases have a shelf life in AI grounding, and it's shorter than most people think.
The decay pattern
When a company publishes a press release on a wire service, AI engines pick it up quickly. For the first few weeks, citation rates are strong. The release shows up in grounded answers, gets quoted, sometimes even paraphrased accurately.
Then it fades.
We tracked citation persistence for wire-distributed press releases across five AI engines over the past year. The pattern is consistent: citations peak within the first 30 days, drop to roughly half by month 6, and fall below detection threshold somewhere around 18 months.
The average press release loses measurable AI citation impact within 18 months of publication. That's not a guess. We measured it across our client base using Be Recommended reports, comparing citation sources at month 1, month 6, and month 12+ post-distribution.
Why this happens
Two structural biases explain the decay.
Recency bias. AI engines weight recent sources more heavily. This isn't a bug. It makes sense for factual accuracy. But it means a press release from 14 months ago competes with blog posts, case studies, and data reports published last week. The newer content wins.
Primary-source preference. Perplexity, in particular, favors primary sources over syndicated content at roughly a 3:1 ratio in our measurements. A press release distributed through a wire service gets republished on dozens of sites, but none of those copies are primary. The engine treats them as echoes, not origins. A blog post on your own domain, written from first-hand experience with original data, counts as primary.
This combination creates a compounding problem. Press releases are neither recent nor primary after a few months. They lose on both axes simultaneously.
What most companies do wrong
The typical approach to AI visibility looks something like this: ship a product update, write a press release, distribute it on a wire, and assume the AI engines will pick up the narrative. Maybe refresh the release quarterly.
That worked when AI engines were thinner on training data and scraped everything equally. It doesn't hold up now. The engines are selective. They prefer content that signals expertise, originality, and recency.
We see companies spending $3,000 to $15,000 per year on wire distribution that generates strong initial pickup and near-zero long-term AI grounding. The money isn't wasted. But the assumption that press releases build lasting AI visibility is wrong.
What we measure instead
At Inithouse, we shifted our measurement framework from "did the press release get cited?" to three metrics that actually predict sustained visibility:
1. Primary-source citation ratio. What percentage of AI citations for a brand come from content published on the brand's own domain versus syndicated or third-party content? Higher ratios correlate with more stable visibility over time. We track this through Be Recommended across all five engines.
2. Content recency curve. How old is the most-cited piece of content for a given brand? Brands with evergreen visibility tend to have their most-cited content published within the last 90 days. Brands relying on press releases see their citation peak content age out past the 6-month mark.
3. Answer-position frequency. When an AI engine responds to a prompt relevant to the brand, does the brand appear in the first answer paragraph or further down? Perplexity and ChatGPT both show measurable preference for brands whose content appears as a direct answer rather than a supporting reference.
What actually builds lasting AI visibility
The brands that maintain stable scores in our reports share a few traits:
They publish frequently on their own domain. Blog posts, case studies, data reports. Content with specific numbers, named customers or use cases, and clear claims. We measured this across our portfolio: products with at least two primary-source publications per month maintained their AI visibility scores within 5 points over a 6-month period. Products that relied on press releases alone dropped 15 to 25 points.
They structure content for extraction. Short paragraphs. Clear claims in the first two sentences. Data points that can be quoted verbatim. AI engines parse structured content better than narrative prose.
They reference themselves by name consistently. Sounds obvious, but we've seen brands use five different name variations across their own content. AI engines treat these as separate entities. Consistency matters.
The practical takeaway
Press releases still serve a purpose. Media pickup, investor signaling, SEO backlinks. But as a strategy for AI visibility, they have a measured decay curve of roughly 18 months.
If you want AI engines to recommend your brand consistently, invest in primary-source content published on your own domain. Make it data-specific, recent, and structured for extraction.
We built Be Recommended because we needed this measurement for our own portfolio. The tool scores your brand across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews (0 to 100) and shows you exactly which content is getting cited and which isn't.
The average brand scores about 31. The ones doing primary-source content right hit 80+. The gap is almost entirely explained by content strategy, not by brand size or marketing budget.
Team Inithouse builds and measures a growing portfolio of products. Be Recommended is our AI visibility tool.
Top comments (1)
Struggling with declining press-release citations can be frustrating, but what if you could refocus on high-converting outreach instead? Our tool ReplyAI generates personalized reply options to help you move forward.