Prezly

AI assistants aren't just reading newsrooms once to build their memory, they're actively querying them in real time while answering user questions. When looking at server logs across Prezly newsrooms, one pattern jumps out immediately: /search is consistently one of the most crawled paths by AI bots. Instead of just scraping static homepages or recent articles, AI agents (like ChatGPT) are using your newsroom’s internal search bar *in real time* to answer specific user queries. In fact, across a single enterprise newsroom on our platform, AI crawlers logged over 700,000 requests, with 281,000+ of those being live fetches by ChatGPT while it was actively answering someone's question. What does this mean for corporate comms teams? 1) Test your internal search yourself. Go to your newsroom search bar and search for key facts or historical announcements. If it's clunky, slow, or fails to bring up the right story, AI bots are going to struggle with it. 2) Your archive matters. AI bots aren't just looking at yesterday's press release. When someone asks a specific historical or factual question about your brand, the bot queries your entire archive, whether that’s 100 stories or 10,000. 3) Structure your key facts clearly. When an AI hits your internal search results, it needs clean, unambiguous headlines and text snippets to know which story holds the right answer. When users stop Googling and start asking AI assistants about your brand, your newsroom's search bar becomes their way in. Make sure it's accessible.

PRs – when we say "earned media", do you picture journalists and media publications, or does it stretch to Wikipedia, Reddit, guest blogs, academic research, government sources? Amanda Green raised this point in a recent post (linked in comments): when they say "earned media", PR people and the tech industry are often talking about two different things. The stats everyone is quoting, like the finding that 84% of AI citations come from earned media, are talking about the tech industry's broader definition. In the study, traditional journalism only made up about 1/4 of that figure. The rest was Wikipedia, Reddit, guest blogs, and other unpaid third-party mentions. That difference in definition could cost you your strategy. So, do you know – what does "earned media" mean to you, and what does it mean for your client?

A quick product update to round off the week: you can now put images and text side by side in your pitches and stories 🎉 Sounds like a simple update, but actually a bit of a nightmare to make work with all the email clients out there, so big kudos to Patrycja Zawrotniak and Alistair Power for shipping this one :) Have a good weekend everyone! 

Question for agencies and external PRs – do many companies ask you to report on the number of pitches you send as a key performance metric? Michelle Garrett raised this in her comment on my post earlier, and it’s got me morbidly curious. Is there a target number of emails brands want you to hit? What do you even say to that? 

All the PR posts seem to agree that spray and pray pitching is dead. And yet, every week, there's a journalist posting screenshots of pitch emails that clearly had zero research behind them. Why is there still such a disconnect? I think a big chunk of this is down to PR tools treating outreach like a volume game. Media databases in particular sell the idea of pitching a prebaked list of journalists en masse, even if they only mentioned your niche once in passing, two years ago. When your software (and its marketing) centres on how many journalist contacts it has and how many of them you can pitch all at once, of course your workflow is going to naturally lean towards mass distribution. But if you want better response rates, stop looking for bigger lists. Instead, prune your media list so that you know who you’re targeting and why, and treat your outreach like a conversation. The numbers reflect this – looking at the aggregate stats from all the email campaigns sent through Prezly this past year, we can see that clickthrough plummets as your mailing list grows. That's because PR should be less about numbers and more about authority and trust. So the question is, how do we build that trust?

A customer got in touch recently to ask how Prezly handles AI scrapers, so I wanted to clarify what we do – and don't – open up for AI to crawl. ✅ Public stories and newsrooms = ARE searchable and crawlable by search engines and AI agents (like ChatGPT, Perplexity, and Claude) ❌ Private stories and newsrooms = are NOT searchable and crawlable by AI agents and search engines If you want to check if AI is surfacing your public content, you can look at your site analytics for referral traffic coming from AI agents (or ask us to look into it for you). However, there is a big caveat here: if you don't see AI referral traffic, it doesn't mean AI isn't referencing your newsroom. It just means users aren't clicking through. Now, that might be because they are getting enough info from the AI answer that they don't feel the need to click through, or it may be that your brand isn't showing up in the answers. That's not something you can tell from traffic figures. A few things you can do to make your story work better for AI: - Include key facts, like date, location, story points, up top - Use structured, unambiguous language - Explicitly attribute quotes and data - Always include a clear boilerplate - Format images with descriptive alt text Luckily, most of this already comes part and parcel with standard press release structure. You don't have to write your story specifically for AI. But structuring your content well ensures the machines don't mangle it when they pass it along to journalists and everyone else. --- What did I miss? Is there something else you're including in your press releases to help AI sit up and take notice? Let me know below :)

Question for PRs: Do you want AI to train on your press releases? Right now, our newsrooms are crawlable by AI agents seeking to answer human queries – so, when someone asks ChatGPT "what's the best mic for studio recording", it's possible for the AI to then go and find your story as a source. Prezly does not block those. What we DO block are requests from models looking to train on your data. (That said, there's only so much we can do here – our Cloudflare data suggests AI agents don't always respect the robots.txt "Do Not Enter" sign. If you're curious, I can dig into those numbers and share them in a follow-up post.) The reason we block training AI requests is largely due to consent. We host content for thousands of brands. We can't assume everyone is morally or legally comfortable with tech giants hoovering up their IP to train proprietary models. There's also the issue of bandwidth. Unregulated scraping bots aggressively hammer servers, consuming massive bandwidth and computing power, which can slow down newsroom load times for actual humans. BUT... blocking training data isn't as simple as "good v bad". There are potential benefits to letting AI train on newsroom data: 1. Better brand representation. If an AI model is trained on your client's official, factual press releases, it’s far less likely to hallucinate false information 2. Long-term discoverability. As search engines evolve into conversational AI tools, being part of the foundational training data ensures your client stays in the AI's memory cache, rather than relying on a live web search 3. Industry authority. Press releases are structured, high-quality, verified factual data, which is the exact type of clean data that AI companies need. Having your content in those training sets can position your client as a category authority So, PR community, I want to hear from you. Do you view AI training as unauthorized IP theft, or as the ultimate form of media distribution? If Prezly gave you a simple toggle switch in your settings to allow or block AI training scrapers on your newsrooms, would you turn it on? Let me know in the comments.