TL;DR AI + PR
A quick digest of five AI+PR updates: how AI and PR intersect, by the numbers. For past editions, click here.
1.How to make sure last year’s event doesn’t become this year’s crisis.
PRNEWS reported on a restaurant client that ran a special promotion in 2025 which they did not repeat in 2026. But AI didn’t get the memo. Several AI tools served up last year’s 2025 details as current, no timestamp, no verification, and folks showed up at locations nationwide expecting an old deal. The best way to handle this is to build expiration metadata into your materials, but if you don’t have that kind of expertise, the low-tech version still works: put the full date, including the year, on everything, press releases, promo graphics, even signage! And the moment a stale listing surfaces somewhere, move fast to get it expunged before it recirculates.
2. Reddit’s AI citations cratered, or did they?
On August 14, Reddit’s share of ChatGPT Search citations collapsed from 3.83% to 1%, an 86% drop in days, according to tracking firm Promptwatch. But when one AI SEO researcher examined the web requests ChatGPT makes behind the scenes, ChatGPT was still regularly consulting Reddit, it was simply citing it less often in the response users actually see, with brand websites picking up more of the visible credit instead. That’s a good thing. This is not the first time Reddit’s citation share has dipped and fully recovered. Bottom line, according to CMOs in the know, it’s a bad idea to rewrite search strategy every time an algorithm shifts. Visibility strategy built around chasing one platform’s citation count is at the mercy of every algorithmic hiccup, while a strategy built around genuine audience engagement is not.
3. AI search engines get the news wrong more than half the time.
We’ve all heard about AI’s lack of reliability when it comes to the information it unearths, but this puts a number on it. Researchers at Columbia University’s Tow Center for Digital Journalism tested eight AI search tools on their ability to correctly identify and cite news articles. Collectively, the tools answered incorrectly more than 60% of the time, fabricating links, misattributing quotes, and pointing to syndicated copies instead of original sources. Don’t assume your client’s earned coverage is being represented accurately just because it exists. Check what AI tools say when someone asks and correct it when they get it wrong.
4. Want AI to cite your client? Include quotes and numbers.
We all want our clients to be sourced by AI search engines. But aside from getting a ton of well-placed coverage, what else can you do? A study on AI search visibility run by researchers at Princeton, Georgia Tech, and IIT Delhi tested nine content strategies across 10,000 queries to see what moves the needle. Adding direct quotations lifted a source’s visibility in by 37%. And, adding statistics lifted it by 22%. The takeaway for bylines and press materials is clear: a real quote from a real person and a real number are essential to getting AI to name your client.
5. Wikipedia’s traffic is falling but its influence is not.
The Wikimedia Foundation reported that humans are consulting Wikipedia less than before – 8% less year over year, to be exact. That’s because AI search engines and chatbots are incorporating Wikipedia content directly into their answers without sending folks to the page itself. The moral of the story: Wikipedia is still a valued resource that AI is sourcing and surfacing in its answers. And what makes Wikipedia tick? Earned media! Per Wikipedia’s own “notability guideline for organizations and companies,” a subject qualifies only if it has been “the subject of significant coverage in multiple reliable secondary sources that are independent of the subject,” a standard explicitly written “to prevent gaming of the rules by marketing and public relations professionals.”
If you want to uplevel your AI Search, reach me here to talk about it.